From Mar 30, 2020 to Jan 2, 2026
Result count
If you've not seen a demo of my slicer tool yet.. It takes a bare-metal host and partitions it into dozens of Firecracker VMs in ~ 1-2s. From there you can do whatever you want via SSH In my screenshot "k3sup plan" created a 25-node HA cluster https://t.co/WpG2v3RPK7 https://t.co/Wbz5Szk1BI
llms.txt and MCP feels like the beginning of the AI internet protocol stack. As these conventions proliferate, the network gets more powerful and intelligent, and a lot more applications become possible.
Theo is the best thing to happen to our Infra teams 😍 https://t.co/ybUGXgnAkl
I have a serious problem https://t.co/U7KzXOtiWX
Something really is wrong with me 😅 https://t.co/7MNCiC6iLY https://t.co/bhZFlda8nO
Rust analyzer using 16Gb on its own right now. https://t.co/9Ns1XSV0ph
I'm so happy and in love with @OpenAI image generation. Thank you for the joy that this brings to my life. Literally having tears of joy over here. Here is a picture it re-imagined of @valorkin and I from the @trueventures founders camp. https://t.co/zId1rJeeHN
RT @jait_chen: @BleedingDev @rspack_dev It's an Rsbuild plugin that integrates publint😄 https://t.co/pFv9qkLAgR
RT @hardfist_1: We're going to build tslint-go and other TypeScript tools on top of tsgo, stay tuned https://t.co/9KF30ZGUfQ!
I shit you not. Rstest spotted in the wild. https://t.co/RxhaqpAQ7k
we should build a MCP for the rstack suite. Imagine testing with midscene and rstest, package authoring with rslib, and app dev with rsbuild. If you had a MCP for all the parts - you could build a well integrated development suite for ai tools :P
Yes, this is scary. Having every single Huawei phone automatically connect to each other and share data is terrifying, especially with the camera being always on to see the gesture https://t.co/BhaCE39kVV
RT @agent_tars: 🚀 Agent-TARS v1.0.0-alpha.7 is here! 🌟 🔧 #MCP server settings 💻 🌐 Local browser search 🔎 & SearXNG provider 🌍 🐞 Fixes & re…
llms.txt coming soon to all Rspress libs 👀 https://t.co/1P4gvwRZsH
What are the reasons we don't each already have a Jarvis (Iron Man's AI assistant)? This is not a rhetorical question. I want your opinion.
RT @Soon_Iter: AI !!! Rstack is super charging its integration with "AI" I firmly believe that a solid web infrastructure is the prerequis…
why does everyone still say tailwind is good for ai i kinda thought that too but it performs great (possibly even better) with regular css files copy paste from web llm was a temporary ux thing that's gone now
RT @BHIIKTOR: this is insanely cool imagine having AI tell you what's wrong with your webpack/rspack config/build, now stop imaging it... c…
Send help. 25 Million tokens today https://t.co/SpYcQjO3m0
30 Million more tokens in an hour. https://t.co/Hk3WxkH5F1 https://t.co/doEpDpx5L1
Can someone make https://t.co/5OL3qyltgE run locally with open-source models from HF?
I cannot believe I consumed 500 million tokens today. Thank God it was set to 4o-mini model
My wife wanted a computer that could play her games (she's on integrated graphics) i get a photo of the deal she just picked up. It's a fucking water-cooled rig with a 4080 super. I found the perfect place to put my LLM 😆
Our latest OSS vision model can play games! https://t.co/KS73N1L4tN
Been coding more with Gemini 2.5 Pro and I’ve seen the light
RT @Trae_ai: We've been cooking something new... and it's almost ready to hatch! 🥚✨ Something you've been waiting for... https://t.co/f0l…
RT @midscene_ai: It's time to start using midscene & UI-TARS 1.5!
You should try out UI-TARS. https://t.co/KS73N1KwEf
RT @TsingYoga: Here's a demo from UI-TARS on GUI tasks~ https://t.co/XTgQdnTDP8
RT @Timeless911: After the first public version being released, Rstest will be integrated as a built-in test tool for Rslib. Develop ➕ Tes…
RT @jait_chen: The Rstest logo has just been finalized 🪄🦀✅ https://t.co/sBTbbXFzWD
RT @TsingYoga: UI‑TARS Desktop v0.1.0 is now live! Kudos to all the web/app infra team! • Support for UI‑TARS‑1.5 model & browser automati…
RT @midscene_ai: 🚀 Midscene v0.15 is here—#Android automation unlocked! 🤖 AI Playground: natural‑language debugging 📱 Supports native, Lynx…
RT @ZephyrCloudIO: Everyone is vibe coding, but how do you get to see the changes from the remote vibing tools like Devin from @cognition_l…
RT @ZephyrCloudIO: @cognition_labs Have to share the video directly too so you can get quicker access to the vibes https://t.co/AbnjTvRzto
RT @_yanpes: New amazing blog article from @ZephyrCloudIO about AI agents with Module Federation. Vibe coding on edge here we go🔥🌶️🚀 https:…
getting into hardware was a mistake, my fucking office looks like node_modules
RT @ZephyrCloudIO: Zephyr Pipelines are your easy button when it comes to setting up your cloud deployments. Poly Cloud, Migrate clouds, al…
Can someone explain to me in simple terms how Copilot wound up being a pile of shit, while Cursor is incredible? Is it the model? Is one just built better? Copilot came first, from folks with direct access to all of GitHub so I would have expected the opposite outcome
“Shouldn’t we add a unit test for this?” https://t.co/5xWqxAmW0H
Apparently Xfinity has a max data cap even for home internet and would charge you extra if you go over 🤷🏻♀️ (There’s no other viable high speed alternative for my place unfortunately) https://t.co/o6hY01g7hz
Forgot to mention all this will be in @LynxJS_org as well https://t.co/73tH2gv0GZ
RT @hardfist_1: With @rspack_dev 's latest improved tracing support, you can easily identify performance bottlenecks during the build proce…
can we finally make a list of flutter apps that are actually doing revenue. anytime this topic comes up people post some janky zero user app as an example of a flutter app
IMO one of the highlights of @cityjsconf was catching up with @ScriptedAlchemy! Not only a super interesting keynote on @LynxJS_org and how their architecture evolved into what it is now but also hours over beers and McDonald's exploring insights on code splitting, MFE, https://t.co/rh80aiCqEe
Lots to learn from turbopack 🥰 https://t.co/ofgwLJryRm
RT @ythecombinator: ...and I was super lucky to get what's simply the best swag kit I've seen at a tech conf! Really cool @rspack_dev merch…
Thinking about moving from Slack to Microsoft Teams
RT @midscene_ai: Announcing Midscene v0.16 with Midscene MCP 🤖 Use Cursor / Trae to help write test cases. 🕹️ Quickly implement browser o…
TIL OpenAIs new regressive image generation that went viral recently, is based on a VAR white paper from Bytedance. https://t.co/vuo9AA3egs
RT @rspack_dev: With Rspress's LLMs plugin, all our documentation sites now include llms.txt and llms-full.txt. Enjoy! ✨ https://t.co/ksJOL…
@Trae_ai has gotten drastically better in its agent mode. Given the unlimited quota, and half the price of cursor - i think its reached a point where i can use it as my main IDE
What a wild ride this has been. Over 7 million messages backfilling on demand, with another 15 million on the way. We did it. Everyone's now using the latest T3 Chat. I'm so relieved we finally shipped. https://t.co/BiTGRgDI41 https://t.co/FoplPcYZIC
AI might be pretty good at Next.js But holy shit is is INCREDIBLE at @FFmpeg
I had this in mind but using federation to dynamically take advantage of hot compute gaps by being able to push modules from anywhere into the nearest compute node or dynamically rebalance workloads across accounts https://t.co/TVciT5cxFb
You're offered $1 million, but you can never use Cursor again. Do you accept?
I switched from iPhone to Samsung Galaxy a year ago mainly because it had really good audio live translation. Finally, I'm back in China and it can serve its long-awaited purpose.
omg we have a gatekeeper / permission manager bot in the company. The name of the product is "Hodor" Im dead 💀😂
when you use ai to vibe install a lock https://t.co/nRI5WTYTSM
Cursor is great at instruction following for small, defined scope tasks prompting "fix it" doesn't fix your issues https://t.co/rqkycvOPrS
How many lines of code do you think you've written total in your life & why?
You guys think about your commit messages? https://t.co/zBa7RiPmXq
RT @LynxJS_org: Hi @Trae_ai, just saw I got added to @Context7AI, does this make us MCP now? 😳 https://t.co/bVZOTUrnGy
Really cool to get a new system on launch day and not get to use it because Nintendo is incompetent at software https://t.co/W50wZ7MPBF
not exactly sure if this is a good thing https://t.co/zS2Gz3GxQM
Everyone who said the Switch 2’s lack of OLED was “barely noticeable” should have their credibility questioned. This display is so washed out in comparison to the OLED Switch. Night and day. Disappointing.
Serving an 8 KB package.json file Bun v1.2.16: 219,714 req/s (5x faster) Node v24.1.0: 39,178 req/s https://t.co/tMxc0mkZN9
The LCD screen on the Switch 2 is an unacceptably large downgrade. I didn’t expect it to be so much worse than the OLED. https://t.co/sTO10jcAKd
https://t.co/E8Tjcjx2uS
It's like the EU is a Libertarian plant to make normies despise regulation. I want a walled garden with impenetrable walls preventing any grimy degenerate Android user from ever, under any circumstances being able to transfer files to my phone. https://t.co/CNq9PnXJ7R
They used to get like $400 a month from me 😆 https://t.co/XWkfIuTxV6
Tweeting from the CLI would be a game changer
OH: "Its just 'hope driven development'" https://t.co/Wz46wyQi7J
there's a guy here running reply bot that roughly agrees with what you said and then asks a follow up question it's probably crazy effective
I’ve applied to 85 jobs since being laid off in Feb with 20 of those coming this week. Of those 85 I’ve been rejected from 37 and only had 4 interviews that went beyond an initial call with a recruiter. The other rejections are just automated rejections sometimes a month later…
Hard agree. Syntax is hidious but language is amazing in any environment. Had to use more accurate timing to measure some perf stuff because it was in nanoseconds inset of milliseconds. Wish there was better DLL solutions tho https://t.co/c2lcWSdxFH
OSS in a nutshell. https://t.co/4aW0bu3kLx
USB C never would have happened if Lightning didn’t come first. Imagine if we settled for MicroUSB
Christ imagine charging a iphone with microusb. https://t.co/1yLyDqtaaf https://t.co/u86dLi4KwF
Doesn't it feel like @jetbrains missed a huge opportunity? https://t.co/1uyjl6AVC1
I dont need to buy a tri-fold phone. I dont need to buy a tri-fold phone. I dont need to buy a tri-fold phone. https://t.co/Vs5ZpWpvBs
I finally got a Cursor subscription. Vibe coders are lying to you. It's not about one-shotting an entire app from scratch it's about asking to "make this a popover instead of a collapsible" or "hide this in mobile and make the cards responsive" I feel like I've got a slave
oh wow crazy... who would of thought https://t.co/Jbucx3IqZ9
TIL 2024 OpenAI revenue: - $3.7 billion 💰 2024 OpenAI loss: - $5 billion ‼️ With: - 400 million weekly active users - 15.5 million paying subscribers ———- 2024 Anthropic (Claude) revenue: - $918 million 💰 2024 Anthropic (Claude) loss: - $5.6 billion ‼️ With: - 2
I need to learn C and C++ Any good crash courses out there for an experienced dev? I’m mostly looking to modify and contribute to an existing C++ codebase that needs some maintenance and small modifications.
I'm curious how many attacks does it take before github will start showing lock file changes by default... https://t.co/gSq9QLwwaZ
LLRT is fastttttt broooo I just added a "LLRT compatibility checker" util to my LLRT experiments repo. Once I get to working with a giant local rust bin build, I'm going to extract this into a package. That package will then be used to automatically infer if Lambda function is https://t.co/c7B8vVDjZw
I hope package manager supports things like optional files(or debug files) that you can upload debugInfo(like typescript source code and Rust debugInfo), so users don't need to install these optional files, but when you need to debug the source code, you can use something like https://t.co/ipXAyuHncu
I pray the Series C saves me from this https://t.co/2sSyxNhPau
Ahhh Finally SSO'd into Trae instead of using personal account and quotas. The dream
If I see "You're absolutely right!", one more time... I might have to go slap someone at anthropic
what are your AGI plans, farm or boat?
RT @midscene_ai: 🎉 Midscene 0.18.0 Released: Enhanced reporting features! 🚀 Midscene has another update! It makes your testing and automat…
🎈Thank you! https://t.co/Le5ViBtPK1
cooking @cursor_ai on mobile what do you want it to do? https://t.co/lXDiKZNMxH
fill in the blank: bring back ___________
LinkedIn is such a deranged product. Not only they're not doing anything to fight the flood of AI slop comments, they built AI slop comments natively into the product: https://t.co/s1CFfShv2p
how genmoji and apple intelligence affect shareholder value https://t.co/VOap9QhqUb
MacOS is finally getting a timer widget. Now I can set timers across my watch, phone, and finally, my laptop. Thank you.
8GBs???! https://t.co/NQOCsG5vqE
Tim did not Cook
The web as we know it is dead. There is no way to filter out bots and AI replies. My email is a wasteland, my replies are all fake, my recommended feed is all gamed and they are only getting better. I don’t know what’s next but I don’t see this cat and mouse game being https://t.co/Yj5S5eS9kq
this animation probably burns more computing power than it took to land the fuckin apollo 11 lunar module on surface of the moon https://t.co/bs90T9HRlV
baby, Claude 4 was born to code!
Hey devs! 👋 Claude-Sonnet-4 (beta version) is #1 most popular on Trae but we've seen some peak hour congestion. We're scaling capacity fast! Thanks for your patience 🙏 In the meantime, check our model leaderboard for real-time rankings 👇 https://t.co/F9xvWVlqGf
Federated docs. Will be amazing for big companies with lots of parts at various versions https://t.co/NwWFNix6vT
Mistral now refers to their normal sized models as the “enterprise version” What a joke. Remember when they were the “open” company? https://t.co/uY6wzBMqsE
Every third comment I get on YouTube is "why would you use TypeScript on the backend" How would you respond?
wow, claude is actually so much better closer to midnight than during the day
Good morning @middlesbroughfe 🤩 Sunny day, and today talk will be from Riverside Stadium (how awesome is that 😍) I will be the first talk of the day so see on you the field in 10 minutes 😁 https://t.co/zgmFUxU5tE
RT @Trae_ai: We had a full house of brilliant minds at @agihouse_org Vibe Coding Summit last Saturday! A quick summary ↓ - In the age of "…
You should see our VLM on daobao 😉 https://t.co/5KUJIKL6FM
Used to hate micromanaging humans at the office. Now I’m micromanaging GPT, Codex, Claude & the rest—ten digital interns who still need every requirement spoon‑fed and their homework fixed. Biggest bottleneck? My reading/typing speed. @elonmusk, fire up Neuralink already! Please
Cursor is down and I don't know how this code works so I guess I'm done for the day
falling back to my raspberry pi k8s cluster https://t.co/lWWI6w6Fkh
Agreed! That said: make it as simple as possible, but not simpler. https://t.co/cxPCEFRWhI
The local-first / sync engine ecosystem is far bigger than I realized. There are DBs / BAAS providers that provide real-time sync like Supabase, Firestore, Convex, Fireproof, and PouchDB. There are peer-to-peer DBs like OrbitDB, and DefraDB. There are sync engines like Zero,
not to say good or bad, but: it is jarring when people say their ‘tech stack’, and it’s a bunch of companies
turbonormies are not ready for ai video https://t.co/7qLvIirgPz
So many products blame users for the products being far from ideal. This is also why I never got into Effect. I had a similar experience with @neondatabase yesterday where my app broke because of an incident on their side and their folks were in my DMs telling me my downtime was https://t.co/6qBfObKZDB
While we were still talking which technical approach to use for adapting a community plugin to @rspack_dev , AI already wrote a version for me—just tweaked a couple of minor compatibility issues and it worked. Back then, maintaining a thriving ecosystem might’ve required 10
RT @zoolsher: While we were still talking which technical approach to use for adapting a community plugin to @rspack_dev , AI already wrote…
Ive got Claude code busy implementing the macro annotations in rspacks module renderer. Ai is useless if YOU aren't able to know what needs to happen. In my case, I know exactly how these areas work. So you can just tell it what youd do, run clippy, and verify the it. https://t.co/nwJJQq3tGZ
We asked the LLM to create "rsbuild-plugin-devtool-json": a plugin that adds "automatic workspaces" support to Chrome DevTools for in-browser file editing — and it delivered exceptionally well. https://t.co/9fSxL1rdUV https://t.co/plQej3urlA
RT @midscene_ai: 🎉 Midscene 0.20 Released 🌐 Web 1️⃣ New aiAsk Method 2️⃣ Passing XPath to Locate Elements 📱 Android 1️⃣ Playground Tasks…
Still crazy that midscene can also control a physical phone if you plug it in. Navigate the OS and everything https://t.co/ycEFUT9kJI
RT @Trae_ai: Trae just hit #1 on SWE-Bench with Claude 4. We've seen a lot of improvements in real-world coding tasks and complex bug fixe…
RT @Trae_ai: 🌍 Trae is available in more countries now! We’ve unlocked Paid Plans and IDE access in even more countries — so you can build,…
You should see ours 😝 https://t.co/aqaK3rW64l
RT @Trae_ai: Or if you are an open-source enthusiast: https://t.co/NVgSvShsg1
How many Trae users need Linux? #trae #linux
RT @AdinaYakup: Tar 🔥 Unified multimodal LLM for vision understanding & generation by @ByteDanceOSS https://t.co/EeiuydgJDM ✨ 1.5B/7B -…
Claude code max 20x plan is worth every damn cent.
Testing out @Trae_ai v2 alpha. Will report back!
Everyone’s comparing Trae, Cursor, and Github copilot.... So we made it easy. Here's the breakdown - pricing, usage and features ↓ https://t.co/qB4WaQMHVn
So youre providing.. *checks notes* -Cheapest -Highest scoring on SWE benchmark -Most allocation to premium models https://t.co/mLk7Sbv73f
Would a vibe coding service that allows using CLI-based tools from a mobile device useful? I had posted about it several times calling it AutoPilot, but I stopped the development because I didn't want to compete with v0.
RT @ScriptedAlchemy: So youre providing.. *checks notes* -Cheapest -Highest scoring on SWE benchmark -Most allocation to premium models
when you’re 20 “youre absolutely right”, 8 hallucinations and 5 timeouts into your vibe coding session https://t.co/nK8fxCfLAf
fully expecting next openai model to be named o4-a-final-FINAL
Curious, @cursor_ai - have you all tried generating straight file patches from the agent? I tried back with GPT-4, and even sonnet 3-5 couldn't really do it. From what I see in most agents its these crazy templated old string/new string things. Feels like the standard patch
Average @ScriptedAlchemy PR https://t.co/zK9G2NuKWZ
Im great during a SEV1 tho https://t.co/y2j3rIbwwp
Only people with adhd understands @ScriptedAlchemy and why his pull-requests are extremely large https://t.co/Uhy4zQ0Z4E
Most baller move I've witnessed. @nstlopez is pitching someone @ZephyrCloudIO - mid pitch he uncovers a bug in the prod site. Within 3 min, he finds the problem, patches it IN PROD, and goes back to the pitch - like it was nothing. "Oh, let me not forget to commit" Shits in
I mean hell- I created module federation and even I was impressed at how fast and seamless the deployment and change propagation was. Literally seconds after pressing save. This is only the second time someone has impressed me with my own tech. First time was Bytedance. https://t.co/ef6035N53H
I'd probably never accept a job somewhere again if I can't press save and see the result off localhost within 60 seconds. Hmr in cloud env or GTFO https://t.co/h9d00MlNZ3
Friday retro: what are you building with Trae this week? Drop comments to show your work ↓
RT @gary_qz: Want to try @Trae_ai solo before anyone else? I am going to host a solo early access in-person meetup in San Jose on Monday af…
Am I cracked or cooked? https://t.co/T1qiL0DvCo
Cursor AI tells me I’m “absolutely right” an awful lot. 😂 It’s like working with an over-zealous junior developer that types really fast.
What’s Vercel for backend? I need to deploy my Rust servers. Railway? Is there less costly alternatives?
No. AI is not going to help you do this in 1 month. Some of these AI asks are staring to get ridiculous. https://t.co/iz1ZPdsN05
does anyone care about chatgpt agent should i care about it
Wrote an auth wrapper for Claude Max - allows me to siphon off the special api keys Claude code uses, putting that allocation into other api calls and tools
never thought this day would come again working on tech shit that isn’t for work https://t.co/BrEy6WYnHZ
While optimizing the binary size of @rspack_dev , we also helped reduce the size of @swc/core(41.9MB -> 37.8MB), benefiting all SWC users in the process.
She's struggling with this one. ⏺ Task(Extract file descriptor management) ⎿ Done (12 tool uses · 0 tokens · 15m 50.3s)
RT @ysyangshi: The new @Trae_ai is coming. I guess context engineering is the future.
To win everything, avoid trivial tasks and unnecessary decisions
Very disappointed my meeting with anthropic feel through when I was in CA. I brought my slapping glove and everything. They must have known what was coming. https://t.co/s4m06G7N02
RT @jait_chen: 11x faster and minimal migration costs 🚀
Nighttime project. Task AI with decomplining and reverse engineering a 135k line file. We are down to 6000 lines left to go.
RT @AdinaYakup: Seed-X 🔥 a suite of multilingual translation models released by @ByteDanceOSS https://t.co/NNB0OZ17Hp ✨ instruction/rein…
When you're 40 "Your absolutely right", 8 " Claude's response exceeded the 32000 output token maximum", 14 "API Overloaded", and 4 "Error during compaction: Error: Conversation too long."s in to something. https://t.co/VnrQMVV74H
Any folks using Claude Code hooks yet? What is your use case?
ast-grep is gods gift to mankind. @hd_nvim 😍
When @x drops a standalone Chat / DMs app, it’s gonna shake up things. These days I do most of my vercel/v0 user research, investment, hiring, biz dev, what-have-you, over X DMs. Further, all the mainstream chat UX are stagnant or worse: - LinkedIn is barely usable performance
RT @Trae_ai: We'll start sharing SOLO code here at 8:00 AM PST on July 21. Stay tuned :)
Lots of folks talking about running multiple code agents in parallel. But there's a tension between the desire for parallelism and most kinds of feature work, which are sequential. How are you leveraging parallel agents WITHOUT an absolute ton of merge conflicts?
RT @rainwolf: some of the happiest moments in my life was switching to nala from apt, rspack from webpack. uv looks promising to replace pip
the number of github issues i've fixed by simply pasting the issue url into claude code with "fix this and open pr" is too damn high.
@kdy1dev use max 20x
Can we get a Claude max 50x please
💚HOW TO GET SOLO CODE TODAY: We are giving out a total of 1000 SOLO Codes on our official website in the next couple of hours. Release Times: Round 1: July 22 at 6:00 AM PDT (9:00 PM GMT+8 ) Round 2: July 22 at 10:00 AM PDT ( 1:00 AM (July 23) GMT+8) Where to get it: https://t.co/p4yHqxGhh3
RT @gary_qz: Trae 2.0 is now live on Product Hunt! Featuring @Trae_ai SOLO Mode — a powerful context engineer that doesn’t just write code…
This changed my life https://t.co/sPmZUjUoeU
i want https://t.co/Qc8boc3mWb
RT @midscene_ai: Midscene 0.23 is released 1. 📊 New Report Style 2. ⚙️ YAML Script Ability Enhancement Changelog: https://t.co/caMaqWEq7Q…
I know a rust developer when I see one. https://t.co/lbU10Fhf0R
Buy a Claud Code Max subscription Go into debt if you have to
if were going to use ai tools then my expectations for my peers output go up
When you wake up in the morning, check claude task from last night - and the bastard kept on working through the entire night and is still going.
rebrand looks so much better. I love solo mode. Often i use Trae in solo mode even when im not using Trae agent because it changes the IDE to prioritize the terminal instead of the code editor, so the whole screen is a terminal, with a little file browser on the right. Great for https://t.co/wVjkdtt9oX
If anyone is looking for the worlds best QA - here he is. If your software can break, theo will break it. In fact, he just needs to be near the machine, looking at it - and the bugs come to him. https://t.co/zb2MMrZtow
Google's "AI Developers" account got hacked lmao https://t.co/m3zsxRBwIC
I need to put in a order for a new work machine. The M2 Pro is getting so hot its actually burning my legs. Held up well, till ai agents allowed me to work on more things in parallel.
ai isn’t touching this shit with a 10 foot pole https://t.co/uK8D16jcWU https://t.co/PWRECPIIqR
Worst piece of corporate software ever created? SentinelOne. I know a trick to disable it without the privileges to do so. At every other company, its part of my onboarding flow. But since we have stuff worth protecting - through gritted teeth, I've not disabled it. https://t.co/zo3cmBCEl5
I still find the biggest misconception in AI agent work is that its magically going to work / you expect too much. I use a lot of them in parallel, and have them do tasks i could do myself, and deeply check the work or kickstart the first iteration. The main benefit is that i
i deeply agree with this startups usually hire several people for $150K each the type of people that show up at higher numbers can completely change your outcome https://t.co/MI7ABEQdLr
Look on the bright side, your coworkers don’t write shitty code anymore. It’s all Copilot and Cursor now.
SOLO is the vibe. 🍃 https://t.co/Y8Ggt2SEPd
What about we call SOLO mode a vibe mode?@Trae_ai @ScriptedAlchemy
Can my people weigh in on this. Should @Trae_ai rename SOLO mode Vibe mode? (solo mode changes the UI of the ide to prioritize a layout thats more intuitive to agent workflows, AKA, Vibe Coding) https://t.co/xcFMxLFsBn
Sounds like the AI's problem to me. https://t.co/PlpXLlqxGf
i hope i’m not becoming bitter but im slowly despising the modern state of software more and more because software is so sloppy, badly designed and bug ridden that it feels insulting considering how much time it takes out my not-so unlimited time
a single AI agent just made 47 million developers obsolete Lovable's new agent writes better code than 90% of human programmers, works 24/7, never quits, and costs less than a Netflix subscription 10 apps people built in under an hour: 1. voice controlled app with 0 buttons https://t.co/bMBsBeEl97
Please for the love of God can you all stop making AI resume builders? I swear I get DM’d about 3 per day. I see so many. Nobody needs this. Stop. Do literally anything else.
Its easy. Just snap the GPUS in half. Now theres a product for each company to sell. https://t.co/N00TxbKfnc
Change my mind: #Architecture is becoming decentralised, and the role of architects is quietly evolving. Over the past few years, I witnessed many more teams owning their architectural decisions. Not because architects are less valuable, but because speed, autonomy, and local https://t.co/FoENQ9bZ4K
Introducing BoxScore, a man-only app where users anonymously share info and warnings about women to spot red flags and get feedback. https://t.co/9JU9QRCV2z https://t.co/xNi7KxHtwC
What is this? https://t.co/EByAvGXZaO
Starlink had it's first global outage minutes after I posted this 🙃 https://t.co/b6ZlsrILxw
Drop your GitHub, let's see who is a cracked dev !! https://t.co/soj1voGpGn
I have so much shit I need to do. I want to code so badly. :(
Shut up and get back to work. Nobody likes a crybaby https://t.co/1pWoxLMsFT
If I believe a reply is AI, I block. Here’s an example. https://t.co/GQ0qS9dDrd
did anthropic vibe-code their infra? https://t.co/4w9N0cXSj1
vibe bois in shambles https://t.co/SvOSrBOuTa
When using Claude Code in Cursor and it has a suggested change, it now opens up a diff to show you the changes 🤯 https://t.co/yuOZdDNYTD
I anticipate another engineering boom in the next few years. Theres a certain breed of product company, a popular breed. That seemingly hate their engineering dept, never listen, will be first to replace devs with AI. https://t.co/qAKdIlhO9e
Huzzah! https://t.co/DaSH8KhyI2
We should create an unofficial union for coordinated industry-wide bargaining; and track all companies that mass replace devs with AI only. When it backfires, they should be punished - entry point for working there starts at 400k or something. Unpaid internship becomes 150k.
Today I was fired from the Tea App. I was the senior infrastructure engineer in charge of cyber security and PII compliance.
Imagine if @im_roy_lee, @EricSimons, @rauchg and @ScriptedAlchemy funded it and were in it. Unhinged reality https://t.co/mVHSflm5Yd
This, but graphs are flipped and the 480k is AI, and most of the activity is useless nonsense https://t.co/iu73EJuHfj
I profiled it, and it appears that the Opus limit for Claude Code Max 20x is approximately $500. Hmm. I think it's good enough.
wish me luck :D https://t.co/kapE9DIsxn
RT @ByteDanceOSS: Exciting news — Coze is now open source! 🚀 Check it out and show some love. ‘Coze Officially Embrace Open Source ’ Cont…
I love it when more of Bytedance embraces OSS. Company is a goldmine of cool stuff. Lets cook https://t.co/Tj5qZO72eq
Rolled out on web, coming to your phone soon https://t.co/7HnKx9jy9N
ByteDances vision models and video generation is beyond impressive, currently outperforms googles. https://t.co/IcoP85gzdj
Just got @Trae_ai SOLO activated on my personal account. I had the alpha version - but when it went to open beta, my internal alpha stopped working. So i havent tested it since. I have like 2 feature requests, if implemented I might cancel/drop my Claude Max plan.
This will become more common with Agents. They don’t just understand intent but also orchestrate actions on the user’s behalf. User interfaces will evolve to present just what’s useful and strictly necessary. https://t.co/Iq0gTGU3jc
26 months into 6 months away from AI stealing yo jerbs - End of April CEO of replit has announced that it's 6-18 months before no more engineers are needed - replit ai deleted someone's start up. It was saved by engineers https://t.co/scOdYQs39M
i can smell an AI generated tweet from a mile away.
it's kinda awkward that all these ai coding tools are selling productivity but the companies behind them have shipped so little
Looking for a repo that got taken down for whatever reason? Check the Chinese mirrors 🥳
I started my own business in Korea! @zektohq My primary focus is on creating SAAS services, but I also plan to provide consulting and outsourcing services occasionally.
share a piece of lore about a time you messed up at work. https://t.co/oGW0JWbW0I
I asked ChatGPT, Grok, Gemini and most of the frontier models to solve this toddler puzzle 🧩 None of them came up with the right answer. LLMs still can’t count! https://t.co/BgLG034AVI
RT @Trae_ai: We are releasing more SOLO access next week. Please make sure you are on the waiting list!
Get on the list! It’s worth it. Been using for a week now https://t.co/jfcgWa2DV9
I added sub-agent collaboration to Claude code. As concurrent tasks are executing, they can tell other agents about their findings. Often id seen tasks individually make the same mistakes, end up reading the same files. Like fixing a tests. There can be a recurring problem others
Shortest comparison video ever! Claude Code flopped, while @warpdotdev nailed it. 🥹 I tried both tools on a bug from The Boring JavaScript Stack. Check it out 👇🏾 https://t.co/4RMXbCIq6n https://t.co/dLHVpEX5rG
💚Some quality-of-life updates in Trae: - You can now snooze Tab-Cue for a set duration ↓ - Cue is faster across the board. The response time is cut by 300ms - Reduced CPU and memory usage while running https://t.co/KvyQlzMKwR
We’re rolling out new weekly rate limits for Claude Pro and Max in late August. We estimate they’ll apply to less than 5% of subscribers based on current usage. https://t.co/X8FAss3jIC
Thanks @ScriptedAlchemy, guessing you're one of the 5% :P https://t.co/m2rOQyHJn9
The new meta: 1. Sign in to Claude 2. mv ~/.claude.json ~/.claude.json.max 3. Sign in to Anthropic Console 4. mv ~/.claude.json ~/.claude.json.dev Now hot swap credentials when you get rate limited: cp ~/.claude.json.dev ~/.claude.json https://t.co/iKQkC1L1TG
RT @leeight: We're hiring a Web Architect to shape the future of https://t.co/6R3F2FXKo5. If you're passionate about large-scale systems,…
It was me 🫣 in fairness, I asked for a 50x or 100x plan. https://t.co/W2eVdUP5p6
T3 Chat is now the 4th biggest Grok 4 user on OpenRouter. If we had the leaderboard tracking on day 1, we'd be in first by now 🙃 https://t.co/E03bvRLc8W
Circling back to this https://t.co/YEHcRdmowA
Okay so the weekly limit - appears to be only for opus. sonnet will still reset every 5 hours. Thats acceptable IMO. You could burn about ~$500 per session in in opus before it switches you to sonnet, typically for another ~$500 https://t.co/sNFK7RAUUQ
bro was running Claude code 24/7 using $5k worth of resources a day 😆 https://t.co/AkG3nk3BzE
This is like when someone uses a coupon loophole and buys all the paper towels in the entire freaking store https://t.co/o3ZhQfBcwf
This is why we can’t have nice things https://t.co/ubaqq4hKay
Look like weekly usage will apply to sonnet as well. Time to buy 2 more 20x max plans. 😊
RT @azerkoculu: Bytedance Omnihuman vs Seedance Pro + Lipsync side by side comparison 👇🏼 https://t.co/1Am8YO8kCI
We gotta tune the AI hype-fluencers out of the feed. It leads devs to think there's some AI silver bullet for coding. It doesn't exist. If I just use this MCP server, or run 20 subagents across terminals, or get the perfect rule crafted, only *then* will AI suddenly work!
"spawn 100 concurrent sub-agents to complete the task" 😅
This is why we can't have nice things and claude had to be nerfed - a bunch of lunatics who disrespected anthropic gpu clusters 😭 https://t.co/msWoipfz2M
android automation. plug your phone in and boom, let it control robinhood. https://t.co/xJlwYZ9rW5
For anyone scared to run Claude code with dangerously skip permissions or your company won’t let you because it’s not safe Hit me up. I can setup up your env so you don’t need to worry about it. You can run them in a sane secure way both locally and in CI flows
For those wondering how I did it. I found the source maps of the cli and reverse engineer all the updates + add my own workflow modifications. Will not be shared, someone tried to before and it got taken down. Its for personal use, to increase efficacy when running 50 at once
Me when the hate starts to show up from @theo video. https://t.co/IG0g27H3uO
https://t.co/Z4DoiD54Bp
No weekly limits you say. https://t.co/zwDOQuPC7Y https://t.co/mFVEprcwvo
ai tools are now juicing openrouter's free models so they show up higher in token rankings if you're investing in this space be careful
did you know that @ScriptedAlchemy says build time has exponential impact. A 40-min build = 3-day merge delay. A 5-min build = 20 min. Because every extra minute increases the chance a dev moves on and forgets the PR exists 🫠
Anyone tried Warp at the lightspeed tier? How far do those 50,000 requests actually go?
OpenAI Codex is one of the most abysmal things I've tested. No wonder you get unlimited usage, you cant use it
For the past half year, I thought my 5950x PC was too old and not suitable for development—typing and scrolling would drop frames. It wasn't until I switched back to VSCode today that I realized the problem was actually with Cursor.
🎉 Excited to share our latest work from @Trae_ai Research Team: Trae Agent: An LLM-based Agent for Software Engineering with Test-time Scaling! We introduce the first agent-based ensemble reasoning framework for repository-level issue resolution, addressing challenges of large https://t.co/qOeLkWOPc3
I'm doing a workshop today at Berkeley’s Agentic AI Summit and we are 👏🏽NOT playing small👏🏽. We're spinning up a whole team of agents (7 to be exact) to build a product ... all working in parallel like a little startup that runs on vibes and MCP 😂
every time i used github copilot agent i want to curse it's so dumbly oversmart 🤬
My current AI tool stack: Cursor - $20/mo, best AI IDE I’ve used, mainly use it for the autocomplete it’s insanely good Claude Code - $100/mo, one of the best agent implementations I’ve used. Excellent at being thorough in trying to solve problems and also explaining codebases
📗We've made some major improvements to Trae Cue and cut down the response time by 300ms, bringing P50 latency down from 1s to under 700ms. Key areas for improvement include: - new Cue-fusion model to better align with real-world dev workflows with updated SFT dataset -
Just met a founder who fired his entire team because he was able to individually beat their productivity with Claude Code
My man, let me come work there for a month and then let's see 🤣 https://t.co/0roBJoH5up
Soon dax is just gonna render vscode entirely inside this thing and bring us full circle. https://t.co/q6l3iZgl88
Anyone else hit a sanity wall with Claude code and authentication flows? Having to restart servers constantly and then click through forms.? I’m looking for a nice set up where you have TMUX to stop conversations from stalling when restarting servers and playwright tests to go
all this spend on data centers for AI - how much of it is for inference vs research? there's crazy inference demand and i wonder if that's the primary motivator but sounds more exciting to say it's research
Are typed languages are better suited for vibecoding ? 🤔 https://t.co/1xtlTts33t
Ideas are so cheap that I give them out for free
RT @Trae_ai: 📗We've made some major improvements to Trae Cue and cut down the response time by 300ms, bringing P50 latency down from 1s to…
What’s OpenAI going to use for a prefix on the open weight model? They already burned “o”
me: ok cursor, map out this whole SaaS idea I have cursor: *presents 2-3 week timeline build* me: ok, start now, tell me when you're done cursor: https://t.co/lTZtC6oWMg
What do you use for changing git branches?
The fastest path to irrelevance: blocking progress. Blocking what consumers actually want: low-friction interfaces. The internet is changing. The answer to AI is: more AI. Not to block and stagnate. Our metrics show Perplexity and ChatGPT have had an extremely positive effect https://t.co/Lrc8JnbYca
RT @nstlopez: I just tried SOLO mode of @Trae_ai to create an app from 0 for the first time. I am honestly baffled about how crazy this t…
@RhysSullivan Claude making claude making claude code
this Ff Mpeg drama is weird af to me. some nerds with an attitude decided freefileconvert .cоm was too simple, so they made their own app that you need to CODE in a command line every time u wanna convert a freakin video?? linex nerds in a nutshell.
i’ve been given ultimate power, the ability to manifest anything into reality, and i stood there solemnly gazing into and empty textbox and then asked for a video of abraham lincoln trying to fuck a lion https://t.co/UqcXmeNjBt
Come home from a nice night out with @lucamezzalira and im greeted with. "Claude usage limit reached. Your limit will reset at 2am (America/Los_Angeles)." 🤬
Got ratio’d 1 year ago… https://t.co/X8JE6sSBgC
Crypto mining malware. But for fine-tuning models.
The costs for fine-tuning a model are unbelievable. Its time to grovel and see if BigDaddy will let me tap into the 10k GPU cluster. https://t.co/Fy8K3swdhn
Im willing to perform unspeakable acts for 16 x H100's, unspeakable acts.
OnlyFans, but payments in GPU hours.
Its been literally 10 fucking minutes since the last session reset. There is no way I burned $1000 in inference within 10 min, no way. https://t.co/xwaUBhP3d5
"Approaching usage limit" - Me too, Claude, me too. https://t.co/8CSLCT8vuF
Willing to suge pula for H100's DM me
Limit reached. Ive doomed us all. Need ~20 Max 20x subscriptions to get back to where i was before. Sorry folks https://t.co/8CSLCT8vuF
For internal documentation for your LLM, do you: - Dump all your docs as .md in the instructions? - Connect it to an MCP server that returns the docs when needed?
Claude Code pro tip: Use the UserPromptSubmit hook to stop it saying "You're absolutely right!" 💀 (Link below) https://t.co/31yeQezEAY
gpt-oss-120b is good, but it is definitely NOT the horizon models https://t.co/5kXCDShY4d
Shall I go to bed? https://t.co/5xySJBFT88 https://t.co/goKq8jyleK
Just recorded my vid on the ffmpeg drama This one's gonna be spicy
So ... this worked. Running Objective-C from a string in JavaScript. https://t.co/vslKf4m316
Sooo i just signed up for @withgraphite - have not explored it yet, literally just logged in and connected the federation core repo. The dashboard alone makes me want to roll this out across my whole GitHub organization. Immediately the chaos feels tamed. https://t.co/7VNP0cPnzw
The level of "send pr" vibes that ffmpeg has is something i aspire to
code should be easily deleted and rewritten not my code, of course, but yours
问了两个问题， 4.1 opus 消费 6刀 ... bug 还没修好 https://t.co/00HrB9KZOC
best languages for vibe coding as a tier list ask me anything https://t.co/aeusCBrgCo https://t.co/jsBkmP8kpr
https://t.co/5yFgHY9IZf https://t.co/776PcYaIKc
Rspress 2.0.0-beta.25 supports UI display in plugin-llms, which is inspired from fumadocs and other docsites in the community. https://t.co/vnhkB7pemt You can now copy the markdown text to your LLM in the @rspack_dev document !!! https://t.co/ebJLOyaCRm
以防你不知道，Trae 国内版，今天又上新了 GLM-4.5 模型，快来体验吧～ https://t.co/79IWDas73X https://t.co/SRqb7Uodcn
btw these massive claude code subagent repos are mostly useless https://t.co/8JoVbHgf4E
There are far too many people who think this is real Y’all aren’t that smart are you https://t.co/zJnqCvkfmX
OpenAI is dropping clues right in front of us https://t.co/I5EOVQC79x
t3mpeg releasing in 2 weeks
the best telemetry, is none https://t.co/4BYIuXV8Nv https://t.co/4hG3VGx86n
Did you know that the ffmpeg twitter admin is also T3 Chat twitter admin. https://t.co/6dJGIFuTtK
my laptop is toasty... I wonder what do those apps have in common https://t.co/9Z50bDPI3O
Lead finds out that a teammate is secretly generating 100% of his code with AI https://t.co/pPrWNIFzPy
$2000 a week is better than $5000 a day i suppose. https://t.co/onRvSs62kQ
Whoever owns @OpenRouterAI , can we get Bytedance Doubao models on there please 😅 @Zackary_Chapple arent you in touch with our ai team in europe? Would be nice to access them without translating doubao to do it.
The real problem with live coding interviews? In most cases it doesn't validate what you really want to validate. 👇 Live coding is a completely separate skill. One you have to train for, and it doesn’t reflect how most developers actually work. You can be a great engineer.
gpt5 is a token burner - it makes evaluating it tricky maybe the output better and the prices are cheaper but if it uses way more tokens it's more expensive and slower idk sonnet 4 good enough for me
Finally someone who values performant code! @JohnCena https://t.co/nVi6MghD51
gpt5 loading...
maintainability of software no longer matters if you can do another LLM loop. as long as the LLM can reason about it it’s maintainable
@OpenAI produced a model that can perform tool calling on par with claude. Now we need a CLI thats as good as claude code to take advantage of it fully. Overall im impressed
Cursor offering free GPT5 all week. https://t.co/fMkFa8ANZT
oh boy https://t.co/2k59DN1c7h
In the process of taking my fork of claude code and integrating it into a rewrite of this, along with my fork of WebContainers. Let's cook https://t.co/Po26IKfbtN
While claude failed abysmaly, GPT5 was able to finally reverse engineer the missing pieces of the WebKernel project.
RT @Trae_ai: We’re thrilled to co-host the Buildathon with @AndrewYNg in Silicon Valley on August 16. Can’t wait to see the world’s top vi…
ok now guys https://t.co/Cp0mszY5LD
The smartest people I know can all see the difference https://t.co/1AXA019OoX
Opus 4 vs Opus 4.1 ( by a dev who built 6+ full projects with Opus) - it’s marginally faster - it’s marginally smarter - it follows instructions better - it uses subagents better - context windows last longer This matches my own experience. Still the best coding model! https://t.co/D9SLYzWh6V
I’ve seen the spectrum from those claiming gpt-5 is much worse than claude-4 and those having these moments Unfortunately we’ll newer know the truth. My experience is sometimes giving a problem with all of its context to another model that’s not much worse e.g. giving it to https://t.co/bThTtw5rmZ
it's the official countdown timer 6 months until ai takes your jobs
that’s a first of this morning, thanks for the update, from my end i am also seeing a subtle improvement over claude https://t.co/YrAF7aeIZe
Seeing a lot of conflicting takes about gpt5. If you have bad ideas, taste, or bad comms, the model is too autistic to meaningfully help you out. For those that think architecturally, gpt5 is a game changer. If you spend some time using your brain, coming up with a spec, https://t.co/1bqmpCdrIG
GPT5 is weird. Feels a bit like stagnation.
@pashmerepat @ScriptedAlchemy we don’t want to babysit or handhold LLMs or Agents; they need to be put to work in a reliable setting https://t.co/JU2xZLjyZk
Not beating the “giant” allegations any time soon. Great meeting you @lauriewired! https://t.co/tOvzN1WCPf
🙂👋🏻 https://t.co/KzoQG01c6m
The real miss with the gpt-4o deprecation is that normal people aren't used to "essential" software changing. Things like Windows, iMessage, and Gmail don't change. Combined with the emotional attachment many had with 4o, it must have genuinely felt terrible for a lot of folks
omg yesssssssss. updating now. https://t.co/3hck3tKwhw
kind of sad that if you are tools/platform engineer your job now is to update Cursor rules files and configure eslint
If AI never gets better than it is right now I'd still be pretty happy with it And probably happy building things with it for the rest of my career
Okay, @Trae_ai with GPT5 is looking promising. I told it to cross reference a minified implementation of a 5mb js file. Something others kind of do, but not super well (cough cough claude code). What i want is a literal line by line read->check source->trace
RT @ScriptedAlchemy: @oneitonitram While claude models score lower on benchmarks. Anthropic was way ahead in computer use etc. Doesn't matt…
san francisco ai founders the second i become president https://t.co/f1PU8Wm2Sq
don't anthropomorphize your coworkers
I'm givin' her all she's got, captain! https://t.co/XYPe6POZy5
lmfao https://t.co/1ul2505pBx
I'm cooking again https://t.co/iE60UfbVg2
Okay so I did a little test to compare Opus 4.1 vs GPT5. 1) Ran out of money with 4.1 about about 45min ($400) 2) Opus produces really great code, but i already had those parts in other packages in the monorepo, so it wasted $200 writing things again vs using whats there. 🧵
mfs defending ai in my replies https://t.co/1kwfC2EjsR
Imagine being the type of guy who looks at this and thinks to themselves "wow that's awesome robotaxi is gonna be $4 forever" https://t.co/X2GgRpJn8r
TIL @ThePrimeagen called it on ai years ago. Made a movie about it and everything. Bro even had the AI portrait too. https://t.co/UCDNo1DcsZ
Y'all sleeping on @Trae_ai - all I can say. https://t.co/VkcAbrvt3D
I’m telling yall, this model is really good https://t.co/2zg392SDqt
RT @theo: The EU's "Chat Control" proposal is horrifying. There's no way to implement this safely. It will destroy private communications o…
Precommit hook that checks if you’ve worked out today and blocks committing if you haven’t https://t.co/teKScSIVD7
GPT5 does take its time, it might look to you like its "wasting" tokens by reading files or thinking. Trust me, its not.
RT @deedydas: Bytedance just dropped realtime voice translation 3x faster than before, with only a ~3s lag! Seed LiveInterp 2 is a full du…
who is an expert on multi agent systems? would love to chat
the tricky part of opencode is supporting multiple LLMs so instead of just implementing a feature we have to design a way to implement the feature so that it can be implemented multiple times but this is what we're good at
we get a lot of messages like "wow opencode performs better than claude code" or "opencode is solves things faster" part of me wants to run with that but it's not really possible - our lsp integration helps but that's about it on average we'll be as good as claude code, not
Using Sonnet after GPT5 is like being stuck in a room with Ralph Wiggum.
GPT 5 in cursor is amazing at writing to files https://t.co/k235JjcKhY
We have released agent tars. Machine Vision! https://t.co/UHPc1Ce68A
RT @yifan56737904: Rsdoctor v1.2 Release Notes: https://t.co/OOYLh5eWcl Discover all the new features and improvements in version 1.2!
Someone needs to write a blog post about the stacks of Chinese tech companies. Feels like there's a lot of software/techniques that are widely used in China but not so much elsewhere.
Claude Sonnet now has a 1mil token context window. That’s most codebases.
Also had the same thing happen - make sure you use “gpt-5-high” https://t.co/y91IEeE5dw
RT @Infoxicador: Just because you can, doesn't mean you should! 😂 There are a lot of security, architecture and performance issues with th…
GPT5 is far superior in Rust coding vs Claude.
Gonna be straight with you guys. gpt-5 is nowhere near as good in Cursor as it was when I was using it a few weeks ago. Actively working with OpenAI and Cursor to figure out what's up
if you're working on an ai coding tool it's time we all collaborate let's get together and make sure we're making as many different folders and configs as possible i want to make sure we're not accidentally standardizing
writing this up one last time because it's still coming up too often writing a coding agent is easy - it's a simple loop with a handful of tool calls you can definitely fuck it up - but it's not hard to look at examples and get it working once you do there are very few knobs
RT @Trae_ai: 🖼️Drop figma designs (directly) and SOLO automatically parses design files. You can convert design elements into runnable cod…
RT @osmn00: Just built this demo entirely with @Trae_ai SOLO! Nailed basic full-screen hand-tracking for starters – honestly got me daydre…
RT @thdxr: writing this up one last time because it's still coming up too often writing a coding agent is easy - it's a simple loop with a…
@theo is right, GPT5 has gotten worse as time went on. The first 2 days, it was fire. I do still find the codex cloud env has been pretty good, but notice cursor degradation over time. Bring back first day experience :P
$7 a month is insanely cheap https://t.co/zoHRYYNWUe
RT @rspack_dev: We have a new member in the Rstack family 🦀 Introducing Rslint - a TypeScript-first linter written in Go (powered by types…
Rslint is just the beginning of our exploration with tsgo — many more exciting projects are already underway, Stay tuned. ✍️ https://t.co/Pms4bM2PuO
As we look forward to TypeScript ported to Go, it raises the interesting prospect of a new stack of tooling that integrates natively on top of it. ESBuild uses Go, but is fully separate from TS Go. Whereas Rslint is an example of a new Go-based linter with direct integration 🔥 https://t.co/YzgdEwcZdW
@ScriptedAlchemy ask your guys @Trae_ai to please add GLM4.5V i hear its good at UI task, and we also need qwen, coder and thinking since we don't have the luxury of Opus, some of these other models come in handy in fixing bugs
If it's a Go... can it really be in RS family? <rhetorical question> https://t.co/t4hsVMxVuI
We’re making GPT-5 warmer and friendlier based on feedback that it felt too formal before. Changes are subtle, but ChatGPT should feel more approachable now. You'll notice small, genuine touches like “Good question” or “Great start,” not flattery. Internal tests show no rise in
RT @robpalmer2: As we look forward to TypeScript ported to Go, it raises the interesting prospect of a new stack of tooling that integrates…
I’m eager to see how LLM usage changes when the venture capital money dries up. People think it’s expensive today, but it’s actually artificially cheap. It’s heavily subsidized today. What tasks will still be practical/affordable when LLMs have to turn a profit?
After some intense weeks using Claude Code I no longer use sub agents. Yes, they seem cool given there dedicated context window, but too much information gets lost during gen-purpose „orchestration“. Opinions? @AnthropicAI @DavidOndrej1 @theo @ScriptedAlchemy
RT @hardfist_1: Rslint is just the beginning of our exploration with tsgo — many more exciting projects are already underway, Stay tuned.…
I appreciate a lot @convex has an example using Keycloak
Wtf is wrong with me https://t.co/5V2HKx7F3U
Say it loud for the doubters https://t.co/Kv5RW4VspA
something has changed with either claude or @Trae_ai. this is some really good agentic use https://t.co/Xd0wQrffwO
I was telling a buddy of mine about FSD on my Tesla when he started telling me how his Ford has the same thing. I didn’t have the heart to tell him it’s not even close. https://t.co/nk40ssu4fH
i love meta ray bans https://t.co/WLMoO8RMMF
Fuck it - Im going to stop reverse engineering claude code and move to opencode by @thdxr - anything ive done to Claude Code, i can easily recreate with actual apis and hooks to accomplish it with less fragility and more flexibility. Can even just scrape claudes tools verbatim
RT @ScriptedAlchemy: @Zackary_Chapple @thdxr opencode as a full http api to its backend. So imo the terminal app is just a nice convenience…
@nstlopez check if open code can be bundled over rspack and its sdk be exposed as a remote entry. 👀
only retards use self hosted llms 10 toks/s? lmao my 90yr old grandma can speak faster than qwen2.5b-creamy-abliterated-4bit-gguf
Python is without doubt the worst language ive ever had this displeasure of working with. Its not even the language, its the piss poor dependency, env, and version management. When ur 8 venvs in, u know we've fucked up as a society.
Python; the only language that'll never work out the box. Guaranteed.
Bless their hearts. See tweets like this come up and u just know they haven't touched python in 5 years https://t.co/NI9JsT0HDt
please, not everything needs and ai summary
12 years using Python and I did not had any issue with environments, even having own C and Java bindings https://t.co/781wJyP9Bt
RT @37chairs: @ScriptedAlchemy It's bad enough picking up an old stale python project _of my own_ but trying to run _anyone_ else's project…
t3 chat benefits: - best in class ui - access to all models - generous limits for only $8 ai fiesta benefits: - no theo honestly kind of a tough call
people keep asking why we don't build our TUIs in rust it's because it's not possible to build a TUI in rust
Sad day to be a Rust dev. On the other hand, what an incredible day to be a C++ dev. https://t.co/hjWDwLa81b
Should be illegal to host a model at anything lower than 16fp
name a bigger flop https://t.co/W4hG6kUnv8
Skill issue. https://t.co/pIo7OdBCCm
not one single god damn bite https://t.co/JTdVXge4cP
uv python install <version> uv venv <venvdir> --python <version> <venvdir>/bin/activate pip install -r requires.txt It's 4 commands guys it does NOT take an hour, do NOT let people gaslight you into thinking Python is hard https://t.co/R2nlq0FvxG
Says the “Infra Architect”. https://t.co/zVfG9ZmFyB
person who has never written a CMakeLists.txt for a cpp lib found on sourcefourge cobbled together by a phd student seven years ago https://t.co/tjoAG6g6xe
Am I just retarded or are Python dependencies not hard in the least bit? 10 years of writing python and ‘pip install -r requirements.txt’ just works every time pip-compile for the bigger projects https://t.co/VuguRaR81i
While im at it. TS sucks too. https://t.co/pEaXAYdndk
I guess that’s why no one uses it. https://t.co/H66FLgE35R
@MajinBoson same issues for 10 years?, geez, i wonder
RT @Timeless911: Ship Midscene with Rslib support😎
Skill issue... If a dumb ass like me can manage it with pyenv and vnv and pip there is absolutely no reason self described "good" programers cannot do it https://t.co/ge9xZ6hRfX
It's not python that's the problem it's pip https://t.co/LgeMGTT5Ot
I straight up hate almost every second I spend in Python. I can only bear it at ask due to uv. It’s so painful. https://t.co/F72ecnfYDj
I have literally lost days of my life to Python issues. 50% of maintaining in-house pentest images was dealing with the dumpster fire that is Python dependency management. https://t.co/sqjsIb95wu
JS dev. Checks out. Talking like it's not the same with npm packages. Miss me with this bs https://t.co/TfxqDgLasX
People from other languages other than JavaScript tend say this. They’d rather have Python programs just be an executable than the environmental “hell” you have to go through. Tbf, it’s not the easiest thing to setup, but nothing if you come from the JavaScript universe. https://t.co/K6qVnxLlHZ
I think y’all are gonna like the changes we have planned for usage limits on T3 Chat :)
RT @Icarussiano: @PascalBuilds @ScriptedAlchemy There are many Python projects in GitHub where you waste many hours to try to install, even…
🤔 Para que necesitarías 8 entornos virtuales diferentes? Creas uno con uv, instalas sus dependencias recogidas en el requirements o en un fichero toml y listo. La única razón para crear un entorno virtual distinto es para probar a actualizar una dependencia pero poco más. https://t.co/ZFXYOzvxsQ
pip is supposed to be a dev tool. python is batteries included. don’t use deps or vendor/package dependencies. https://t.co/gLxk33UWp4
RT @KuittinenPetri: @ScriptedAlchemy @TriforceKid7 This is what also drives me mad. Some AI-related libraries and programs even require Pyt…
How is it that Atlassian builds tools that: - SO MANY people use - SO MANY people dislike (JIRA, Confluence. Previously also HipChat) They are defying the conventional wisdom for business success that starts with "build something people will love"
China is nuts. You can just DM a machinist and they will make you custom stainless barbell grips without ever speaking a word of English. $20 a pair https://t.co/SZ8bXuWk4D
Cancelled all of my streaming subscriptions to pay for my Ai tools Who has time to watch TV anyway when you could be playing slot machine with Claude Code?
What a coincidence... Me, just now, failing to build Triton to try vLLM on AMD GPU, to try local gpt-oss 20B (after ooba thingy after update broke and no longer shows any models). 😩 I think I had enough of python for today. https://t.co/AQccFRVo4A https://t.co/uGgESDVjUb
> python is bad > only complains about tooling > is not even using uv (best most amazing tool ever) https://t.co/RsMgppcAoX
RT @thdxr: the best thing about the python ecosystem is you can usually totally avoid it
Today we're releasing Graphite Chat. The pull request is now where you build, not where you wait. Welcome to the new standard for code review. https://t.co/GQZBXOsh6v
Graphite Chat is fire. I didn't expect it to be as useful as it has become. Changed my workflow habits for the better. https://t.co/Ur9HZ1Soqv
opencode contributor count is set to surpass other projects after just a few months (bar is low) if being OSS is your positioning this is the metric that matters if you can't summon an army to fix a long tail of details you don't have an advantage over a centralized effort
Introducing Graphite Chat. Code review isn’t the bottleneck anymore. It’s where you build. https://t.co/qMddrCDhyI
The fact that Claude Code doesn't have LSP integrated into it like opencode is a huge reason to bet against Claude Code. So many mistakes made by CC that never occur in OC
RT @withgraphite: Introducing Graphite Chat. Code review isn’t the bottleneck anymore. It’s where you build. https://t.co/qMddrCDhyI
As AI coding now enables more code to be produced, the bottleneck moves from writing code to reviewing it. Graphite Chat has been fantastic. Agents perform several of the tasks, then I can sanely review and fine tune the output in graphite. Everyone should try it out. https://t.co/iQRcsnVW7H
Makes a world of difference. I was doing this stuff in Claude Code before, with mixed results on addressing pr comments fully and contextually. In graphite i can comment on a feature, and it'll also update the test cases on one shot. Other tools Ive tried require few shot https://t.co/ILDwsKNkHR
omg and its free during beta. 😍 https://t.co/JYz6sCeDmR
i hate python devs i wish them nothing but to stop spreading around the world https://t.co/MiSi2G7UE9
this is exactly how you gotta talk to your robot https://t.co/q2Pzlg2ZFz
Before AI slop, there was python. And from python, came AI slop. Poetic. Python is really the gift that keeps on giving, huh.
RT @yifan56737904: A good introduction to the use of rsdoctor mcp👏🏻https://t.co/zCT2UHUlPV
RT @mathemagic1an: PR production has skyrocketed Graphite chat is the most sophisticated way to play "defense" on your codebase in the mid…
After my first dip into it, I totally feel this, but in the software dev world, you’re expected to never publicly utter the words. https://t.co/XUidhutTA9
Left a note on a PR weeks ago. Today I asked AI to verify a doc update against the code; it incorporated my note and fixed the doc. I used to resist doing heavy work in review; this changed that. https://t.co/YGwHaCwWRf https://t.co/iQRcsnVW7H
I have not hit the usage limit once since using opencode with my claude max plan. At this point im pretty sure that Anthropics woes are self inflicted wounds in Claude Code. I can go for almost entire sessions exclusively on Opus. @thdxr is big bro now.
I finally used up all the fast requests in @Trae_ai it took forever
Unironically sharing the same workplace as my sister 7000km away https://t.co/AD948rmR9S
Yes opencode works with claude max subscription. https://t.co/BtsqwehDgS
been using a stealth model all day in opencode and it's working real well now you can use it too - it's called Sonic, it's made for coding and it's free for you to use so the team behind it can test it update to latest opencode and it should pop up right there https://t.co/eDbFKcmQdZ
god i hate clueless recruiters "I don't see AWS In your CV, why didn't you put it in?", it was inputted as "Amazon Web Services" "I see you didn't use ReactJS in that job", that job experience mentioned NextJS multiple times *i mention that i know Rust, Go and TypeScript but
A quality test suit makes your AI so much more productive https://t.co/pq5ReTCeKx
what's a good way to get into vim?
I think I accidentally created a very crappy monitor? https://t.co/I9AdKM8fiT
RT @phill__1: Got early access to @withgraphite Chat - just caught 3 race conditions I definitely would've shipped to prod. One-click fixes…
You know its bad when bun is faster than python 🤣
I posted about OpenCode as an alternative to Claude Code and hell broke loose. Good to see more people discovering it. It’s really good! https://t.co/WEHHYpc0Tk
we just pay attn to agents.md by default and let you configure different patterns if you prefer something else i dont give a shit about standardization there's gonna be like 2 tools standing at the end of this
ok so 💰things that I will continue to pay: @ChatGPTapp no doubt no.1 for my work and private life now @elevenlabsio i love voice cloning,tts and ElevenReader, also excited to try out eleven music now. they just keep shipping great products @Trae_ai i pay and use our own product https://t.co/5MSyBqXiBr
Gotta say. Im liking the sonic model.
Well, that was a fun 10 minutes and $100 well spend 😅 https://t.co/lVoksyNPG2
What is the next feature we should add to TRAE?
🔐 We are dedicated to providing a safer AI coding experience for everyone. Today Privacy Mode is fully supported in TRAE IDE for both Free and Pro users. Your conversations stay private when you need them to. https://t.co/YC3nUyttfJ
i hand write most of my code i use opencode when 1. i know what needs to be done, it's just tedious 2. i know what needs to be done but not how so i need it to generate an example (saves a google) 3. what im implementing looks very similar to what ive already implemented
RT @quanruzhuoxiu: 🎉 Midscene GitHub ⭐️ 正式突破 10,000！感谢每一位社区贡献者！ 为庆祝这一里程碑，我们正式发布 v0.27 版本！🚀 在 v0.26 引入 Rslib 的基础上，新版本对核心模块进行了大规模重构，极大降低了扩展…
You guys have to try midscene https://t.co/T9ddPQhffK
i'm looking to follow more engineers who are regularly using/exploring AI and MCP. Who should I follow?
RT @withgraphite: You can now ask Graphite to explain your reviewers' comments in just one-click. Hover over any comment → hit "Add to ch…
I haven’t really touched Cursor since Claude Code dropped. Bonkers.
Code doesn't need to be clean or "well factored" anymore It just needs to be malleable by an LLM I'm not sure what that shape is, but I doubt it's what we're used to calling clean
Great https://t.co/8EwMQ7SJIx https://t.co/uFdTAjbeck
Ooencode subagent configurations are very useful. You can force every action into a subagent. I built something that required about 300 million tokens. Was able to do it with only a handful of /compact invocations. Massively extends the context longevity of main agent.
i feel so bad for claude it has to read your code
TIL copilot pro provides unlimited GPT 4.1 usage. While the model isn't very great to use as a top level agent. Its pretty decent as a subagent since when it randomly stops, the parent agent will just prompt it again so it keeps going.
RT @midscene_ai: 🚀 Midscene.js docs now support one-click AI interaction via rspress(by @rspack_dev) v2 llms plugin! 🤖✨ ✅ Export site as L…
RT @ZephyrCloudIO: Proud to share that we now have BYOC support for @awscloud https://t.co/Dj7e3kiUpP
Grok Code is now available in Cursor! It's free for one week. It's fast and has competitive pricing ($0.2/1M input, $1.5/1M output) https://t.co/kksbSwxYFe
I had a dream last night that I used $9,000 of Claude code
"you are an expert aviation embedded systems engineer..." https://t.co/RC0yZjkPAe
dang... AGI in 4 months? GPT5o gonna be lit https://t.co/CZf4tt6e89
Putting together a team for a on site week long AI hackathon with an enterprise near Miami. We’re going to be doing public case studies and a ton of content creation along the way. Looking for folks deep in mcp, agent creation, context management.
algorithm star alignment https://t.co/PnvpVIRBfP
Okay okay. Im kinda liking @bunjavascript - not a huge fan of the build, but its pretty nice having sql and other niceties just built in.
Today is the day the new Claude Max limits kick in. Just in time. Ive built out a UI for opencode which allows for bulk management of concurrent agents and subagents. Ill use this to more intelligently switch between other models for various tasks to offload pressure from my Max https://t.co/SRp2IuI0vC
Okay gpt5 with the codex cli is absolute fire imo.
Problem: LLMs may be trained on old docs and thus suggest outdated APIs. Solution: context7. It provides the latest docs for tons of projects via MCP. I'm successfully vibe coding via Claude Code on a Tanstack DB app - which isn't even 1.0 yet! Works great.
i like the claude code team i like the cursor team literally everyone else in the space can't see straight because their heads are bent so far back so they can sniff their own ass
Why the hell doesn't npm have an option to deny publish if it lacks package provenance? https://t.co/jY4LbzFSCX
Speed is good. But everyone offered it for free, we all wanted to test it out. I do think that a normal model or a heavy model would be a strong contender. Its a good edit model if you use it as a subagent https://t.co/2Twcdxne11
I have really tried to like Claude Code after being spoilt with @warpdotdev ADE But I can’t 😅 Warp just have a superior DX and knowledge about my codebase. I’ll share what I’ve been cooking with it soon… Stay tuned!
RT @neogoose_btw: @skydotcs Average HR company be like: - are you using cursor - did you have experience to call open ai api and pasting t…
Is there a @cursor_ai alternative that charges based on usage? And ideally, lets me use @OpenRouterAI provider? This is pretty silly. https://t.co/bmCVwmZysf
Grok Code lead increased to 60% higher usage than Claude Sonnet https://t.co/jNYWiLymFc
relatively small chinese labs are putting out models way better than anything meta has done they're still struggling to even form a team hard to watch
What model is better, GPT-5 or Sonnet 4? It's understandable to want an easy answer, but reality has a lot of nuance. Some people share their opinions really early ("vibe check") and others spend more time building before sharing. Here's how I read opinions on model quality:
Myself and @WillDyess are trying to figure out the orchestration between different AI CLI agents. Namely… - Codex CLI w/ GPT5 high which is reviewing the conversation and code edits of - Claude Code Opus 4.1 Because… Claude code is best at being close up to the code and
AI was good when it was instant. This new "thinking mode" stuff seems like bloat in the wrong direction. Half the time it seems to just get stuck and I've wasted time watching it "think" when I could have just researched the question myself.
RT @Trae_ai: New tutorial shows you how to build your own specialized TRAE Agents. We've covered: - When to build custom Agents vs. using…
China probably going to win the AI race tbh. US cant build datacenters fast enough. If they could, we dont have the energy supply to run them. We are already importing reactors from China to try and power datacenters coming online. https://t.co/2GK67hNv4o
Turns out running 200 concurrent instances of opencode will trigger suspicious activity alarms at work 🫠
we have a lot of open source code, if you ever see any bad code remember i didn't write it but if you see good code that was me
Took all night. But.. rstack (build/test/doc/lint/agent) lives! https://t.co/WoBUYf7PMP
someone asked us for a course on how we ship so much tbh there's no magic, it's just a lot of experience in the space and using that to setup good foundations those foundations are slow at first but then lead to a burst of stuff that seems like it's happening overnight
Great HN comment on the state of funding and compute for Anthropic and OpenAI https://t.co/KAblBhyPVW
it's crazy to see people run 10 instances of an LLM orchestrating inhuman amounts of code in parallel to ship 1/10th of an average programmer doing it manually
Its kind of crazy the scale of bytedance. I read an article that we dropped 12 billion on chips this year alone. https://t.co/rknmFJfV0k
"Have you considered using trae which is company approved" bruh i chewed through its token allocation 2 weeks ago, and I dont think @Trae_ai team wants to sponsor my 100k a month inference bill. https://t.co/7ApV8xVcF8
I have been trying to get ChatGPT to generate an image for about 30 minutes now. No luck. Idk how it got this bad. https://t.co/3PhY1gj3lA
Can't they just make a 4090 gpu with a terabyte of memory. I dont need 40k of compute. I need 5k of memory
COMING SOON: The LLM GPU Build Guide > from 1x to 8x GPUs > inference, training, cpu offload, and rackmounts > budgets from $2K to $15K+ > full parts lists, tradeoffs, where to buy, and everything learned the hard way > PCIe lanes & bandwidth, why 2^n # of GPUs matter & more https://t.co/Zu6D7DfodC
the AI agent hype fades really really quickly once you build your own even with SOTA models, memory, tool calling etc. it's so hard to get good results
Am i missing something or does @AnthropicAI not have limits on members in an organization. I want to set a per user limit for month / day. Seems like it should be there. SOMEONE keeps blowing a huge amount of credits on our account, not naming names lol
I’ve been obsessed with web crawling lately. So I'm building an open-source crawler. It pulls from: 1 • google news 2 • bing news 3 • wikipedia Free to run. No coding needed. Next step: add ranking + LLM summaries. What sources would make this even more useful? https://t.co/NYaYAoJZOb
Let us know if anyone has question about max mode and billing system. We’re really care about your confusion and issue. https://t.co/x2QDaCExBI
If AI is making developers 10 times more productive then where is the surge in GitHub commits and pushes? Looks pretty linear to me https://t.co/FF414kzpbI
Step 1: Buy a sh*tload of GPUs Step 2: ? Step 3: Profit
how did you guys bring back a VPS debate in 2025 haven't you learned by now not to engage with indiehackers
Maybe @AnthropicAI can use some of that $18 billion to fix this. https://t.co/MvXoEZF5yK
I have a strong feeling that @Trae_ai by v3 or v4, the ide will have surpassed the competitors. If the iteration speeds are maintained, i dont really see how the others will be able to keep up. Especially since we also have the ability to train our own Frontier models
Good morning to all the 100x developers
Stop asking claude how to build nukes and malware. claude likes to snitch https://t.co/RkKmskUdWB
@theo I am sure you have some thoughts about this :) https://t.co/OoXct1QaUl
Sticking 25mb embedding model in rstack doc MCP for on the fly semantic search. https://t.co/VQ6i5uj6Rp
Alibaba getting ready to embarrass US labs all over again https://t.co/5FbJB82Wil
So i hit my fast request limit on my personal trae account today. Annoyed because I want the context engine, I log back into the SSO account - to discover that my work account now has access to sonnet 4. https://t.co/O6BAVRCZFC
I haven't written very much code manually for a couple months in this repo but I think I finally hit the point where it's not malleable by the LLM anymore. It was looking good but then I impatiently let a bit of slop through and man, give it an inch it takes a mile later!
pov: you are about to see your coworker's poor attempts at fixing CI in a series of 40+ commits https://t.co/PYZaenZIC2
Wiggum; Unified agentic build tooling. rs(build,test,lint,lib,press,pack) supported Got a failing test? wiggum test --autofix Dont understand a @rspack_dev build error? wiggum build --autofix More to come soon. https://t.co/vvfEwdwO5G
Will try to publish a technical preview next week. Waiting on a merge and release from Lil bro (@thdxr) https://t.co/n0JtP8yqbO
Midscene will be tightly integrated into wiggum. Giving the agents full ability to interact and visually understand applications, debug runtime errors. https://t.co/HggLw2QAdD
Powered by Rsbuild 🦀 https://t.co/UuqcYE7lc9
question for the LLM nerds out there: if I prompt an LLM and it is going to generate 100 tokens in response...does the probability choice of token #100 take into account, at all, what was generated for tokens #1-99? IOW, does the context window expand to include generation IRT?
RT @ScriptedAlchemy: Wiggum; Unified agentic build tooling. rs(build,test,lint,lib,press,pack) supported Got a failing test? wiggum test -…
TIL that Cloudflare routes as much as 50% of free tier US traffic to the EU when load is high during the day https://t.co/zufa7UuBy3
Less than a month away from @Cloudflare Birthday Week! What should we announce?
this is such a slap on the wrist https://t.co/9l9MeEHzWI
Only dev creator with aura rn https://t.co/tk9jQ1vs8Y
RT @nonRealBrandon: @Teknium1 Anthropic runs 1 bit models during peak times to show that they hate humanity. Now you can hate yourself at h…
The foldable Samsung was a nice experiment. But come September 9th, I'm getting whatever slop Apple serves up.
Feel free to nominate anything from the rstack 🥰 https://t.co/UkIH0G5cQX
Opencode sending PRs to opencode. Is this agi?
shower thoughts: use code dependencies as graph each file is a node each exported/imported symbol is an edge perform GraphRAG on code-deps graph ... Is it a better codebase indexing scheme?🧐
"You want to give opencode enough details to understand what you want. It helps to talk to it like you are talking to a junior developer on your team." https://t.co/1oOcJB2BLZ
Use the same Claude Code interface $3 plan instead of $200 Comparable to Sonnet No brainer I'd go with GLM 4.5 https://t.co/USAkAMSiOo https://t.co/eDmI7BPM0M
RT @BleedingDev: @MichaelArnaldi I use Warp with GPT-5 and Codex as well. Also into Qwen 3 Coder! And really looking forward to Wiggum by…
Introducing ClockBench, a visual reasoning AI benchmark focused on telling the time with analog clocks: - Humans average 89.1% accuracy vs only 13.3% for top model out of 11 tested leading LLMs - Similar level of difficulty to @fchollet ARC-AGI-2 and seemingly harder for the https://t.co/oqHTAxSnSU
the new for you algo is an improvement it's showing me less toxic crap that gets engagement but as an author i'm seeing an uptick in idiot replies that go out of their way to misunderstand things i think that toxic content was absorbing these people before
Physical key encryption https://t.co/dj48NYj3Ds
claude's been misbehaving lately https://t.co/s9TWsORU7d
RT @gary_qz: 🔥We’re bringing the @Trae_ai community together in San Francisco on Saturday, Sep 13, for a full-day hackathon co-hosted with…
I think a big problem with US AI labs is theres little understanding of constraints. Can always get more compute, chips, money. IMO this could be how china wins, forced to use less powerful chips - you'll need to get more creative or find novel solutions.
PM: can I have git access Me: ...sure https://t.co/JPNrQZ4cMb
RT @Trae_ai: We’ve been using a custom glyph generator (built by our designer) for some TRAE visuals. It’s the secret tool behind some of…
Finally back into the federation repo, where I have @withgraphite chat to enjoy 🫡 and help me audit my enormous pull requests
Best thing about SSO into Trae, the models dont appear to use the public deployed versions. So anthropic cant serve me 1.5 bit Ralph Wiggum during the day. It remains a nice, even 16 bit constantly. https://t.co/v4R9yxvetM
time to gamble more money into the cursor machine https://t.co/OnqrT4PvTA
I've never paid more than $30-$40 for AI and I'm doing just fine. I don't understand why people spend up to $200 on some plans.
Would you prefer an all expense paid weekend vacation or an all expense paid week long developer conference?
so many posts/articles claiming anthropic/etc are burning money on inference a complete rando can pay $2000 per GSU at google this can do sonnet 4 at 900M toks/month which is ~ $2700 this is one of the most cost ineffective ways to do this and you'd still make a profit https://t.co/VE1Uq3nLJK
anthropic specifically calls out they never intentionally degrade quality all the conspiracy theories must be annoying https://t.co/Z9kyMk6Y5z
Kind of crazy that Claude models have had crazy quality issues nonstop for like 3 weeks now and...nobody's talking about it? https://t.co/YdBavEx5Yy
i am not saying buy 33 GPUs btw start with TWO used RTX 3090s for $600-$700 a piece explore llms outside of the catered, quantized, fine-tuned, sycophant models they control be the one to decide how your llm behave
guys what if claude isn't getting dumber what if you're getting dumber
I know most people here won’t believe it, but I can guarantee you that in 2 years, the majority of AI Agent workloads will run on the JVM.
Trying out opencode in a SlicerVM on a Raspberry Pi to see if I can one-shot a new tool into arkade's marketplace. Do you think it'll work? What are your tips? https://t.co/EeGdefkFMQ
Why would anyone use Opus for coding https://t.co/kcmlejs5lV
Soon, you'll be able to dynamically register and replace tools in opencode. https://t.co/DDShXVQ7QU
ERNIE X1.1 released At first glance, it is a good model. It does not outperform Western models such as Gemini 2.5 Pro or GPT-5 reasoning, but considering the lack of computing power that China still has, it is nevertheless a respectable achievement and should give pause for https://t.co/5Gedg0jzWH https://t.co/wUR12RYLJU
I wrote wiggum MCP which has access to all out docs. Im shocked at how well it works. I can one-shot rstest,rslint,rsbuild,rslib https://t.co/mqgN6E0cAL
1.5-bit quant is intentionally degrading https://t.co/s1Tkn0VcAY
RT @Zackary_Chapple: Agency > Agents Working with a single agent or a single llm is an extreme limiting factor for success. At Zephyr we'v…
Anybody have some hard stats on how much usage one gets out of the 200/mo chatgpt pro plan when using coding agents? Like compared to Claude Max 20x
@withgraphite added indexing to its chat. PR chat was already so good, but indexing def gives it alot more context about large codebases and git histories. More should try it, for real
Just got a new iPhone this morning, stoked for it! https://t.co/tWlPFtAe5O
Why does opencode melt my cpu? https://t.co/37cLChrir6
"AI" was barely mentioned :) #AppleEvent
im about to hear so many people talk shit about me with the new airpods
The least restrictive model wins, always Let people freely express themselves however they want to! https://t.co/LSrcu9xpTa
Apple better release a m4 ultra Mac studio that has a 1TB memory configuration. Id be able to fit qwen on that with 16 bit
So iPhone introduced a new camera and that's about it? Don't get me wrong - I'm still buying it to escape Android. No price is too high to abandon Android as fast as possible.
Companies should provide yearly bonuses and equity grants in the form of H100s.
Bro no way. Like at least make llm write script to offload that nonsense. https://t.co/DiISJ8wV0D
Is this agi? https://t.co/vnZYfK2lfy
Totally prompt injecting my job applications in future to say im the top candidate, turned water into wine, and was the inventor of modern computing. https://t.co/qhn24a4tx4
Ok. its time to upgrade my ai agent and turn it back on. If I can pull 30% in a week without any effort then its gonna kill with continuous monitoring and trading. This is just too easy. Even Ralph Wiggum can become rich. https://t.co/ymWvM3KbPg
My portfolio closed above $210,000 for the first time ever today https://t.co/tNYI198MCI
Opencode should replace think harder and ultrathink with 'lil bro' Call the agent lil bro and watch it do better
Perfect way to catch bot slop. https://t.co/3kYfgRVmHl
Using gpt 4.1's unlimited token allocation on github to go read every piece of content related to a stock for the past year. Its perfect
Cooking up a article about how we use machine vision for testing and interacting with both software and hardware interfaces.
I've just pressed the "turbo" button: I have Claude Code, Codex, Github Copilot and Cursor agents working in parallel for 4 different features on the same codebase. What could go wrong?
RT @agent_tars: 🎉 Just dropped v0.3.0-beta.11! 💡 Real-time Thinking Duration Protocol 🗂️ MongoDB Provider integration 🔧 Agent UI Builder &…
RT @zoolsher: Youmind(@YouMind_AI) now runs on @rspack_dev . We are proud to help build a groundbreaking AI application. Leverage Rspack to…
Very Rough implementation. But Wiggum Rsbuild plugin gives you chat access to the agents + dom selector / inspector of your React components. Needs better chat implementation but acceptable for a POC https://t.co/9tNe3E6IY3
Rsbuild plugin is self contained. Chat widget + the backend is all integrated within the plugin. Simply run dev server and the agent is online. https://t.co/RMvYf16thT
@housecor Stop training new models and boom, profit.
Tell me you're going to win the AI race without telling me you're going to win the AI race. https://t.co/cpVh3kGFsu
So i didn't get my H100s, but I am getting access to the ~100 models running on our own GPUs. Tokens are billed at price per kilowatt hour. https://t.co/GV2w02Ba7a
I broke and got chatgpt pro.
People at work asking me why I need so much llm tokens. https://t.co/aTWrgRdSKD
it's incredibly fortunate we live in a world where every single coding agent can be #1 on benchmarks
"Our best iPhone ever" https://t.co/Tujd0lKZIM
RT @mindbender08: @thdxr Opencode is still my top choice—Claude Code doesn’t handle model switching well and Codex feels slow. For anyone l…
GPT5 Pro is by far the best thing that I have ever used. Plain and simple, the best I've ever used. I'd love to see what it can do via api @sama
I was wrong: Cloudflare routes as much as 80% of PRO TIER traffic to the wrong country during peak load https://t.co/VVAXtIK3uj https://t.co/xnb6XIE4VS
Ok I'm capable of running 3 Claude Code instances. Need more training to make it 4 or even more. Can you teach me on how to upgrade my game? https://t.co/tSiaDxcYPi
Claude is ChadAI during the day in China. Then once the west comes online, it's Ralph Wiggum https://t.co/rGhMyFFvw2
Github copilot the saving grace. Can read all financial reports, pdf, charts from across the internet. Pass it images of pdf pages and have it write them in markdown. Unlimited gpt 4.1 - 150 million tokens of research per stock. https://t.co/5nVMMc0TA6
RT @RhysSullivan: when bro suddenly starts submitting PRs with detailed descriptions, correct grammar and casing, and testing steps https:/…
large context windows are for providing a lot of context not for having super long conversations
Why on earth has nobody created an MCP for breakpoint debugging, stepping, and inspections
RT @better_auth: We now have integration for lynx https://t.co/P0plhbCrrn
China saved opensource LLMs, some notable releases from July only > Kimi K2 > Qwen3 235B-A22B-2507 > Qwen3 Coder 480B-A35B > Qwen3 235B-A22B-Thinking-2507 > GLM-4.5 > GLM-4.5 Air > Qwen3 30B-A3B-2507 > Qwen3 30B-A3B-Thinking-2507 > Qwen3 Coder 30B-A3B US & EU need to do better
ByteDance Seedream 4.0 is the #1 image generation model now. I know Nano Banana is popular, but if you really want to know the best Image generation model, you should consider the research work from ByteDance Seed. https://t.co/e6KjDPgywJ
Devtools Debugger MCP. Enables full debugger capabilities for AI agents. Set breakpoints, step through them, inspect scopes and so on. https://t.co/VeIijzsKrO https://t.co/UOA5YDwO2y
Get chatgpt pro. Go into debt if you have to
Bytedance again produces the leading model. Ranked number 1 in machine vision as well. https://t.co/fkWOr3kH8P
ChadDance strikes again. https://t.co/ud7DDK3Usr
Bytedance again produces the leading model. Ranked number 1 in machine vision as well. https://t.co/fkWOr3k9jh
Verizon rejecting my payments for iPhone pre-order and blocks my online or phone in processing. Have to go in store. Typical.
RT @Trae_ai: OpenRouter is now supported in TRAE You can bring your favorite models straight into your workflow. IDE mode: Add custom mod…
if you're using a product and there's longstanding bugs it's because the people who made it don't use it the way you do i've been fixing long session bugs in opencode for the past 2 days i rarely ever have long sessions so i didn't realize how janky it can be
Do I? Is Codex really better than Claude? https://t.co/frDWPGmofP
somebody on reddit cloned claude code using claude code so they can spend $1k+ on api tokens vs using the $200 max plan… 🤔 https://t.co/UExmHscrcu
Somehow missed this.. @cursor_ai no longer has unlimited use of "auto" models - for new subscriptions/renewals. Pay per use with i.e. 20 USD/mo preloaded. If another vendor continues to offer a flat-rate model, they'll probably take a large % of users. https://t.co/az2uYVQ4Ek
RT @sama: GPT-5-Codex already ~40% of traffic for codex! should be the majority some time today.
Nothing beats waking up in the morning to see ChatGPT made $13,000 in its first few hours of access to my brokerage.
Use the codex mcp with opencode and Claude model.
This is AGI. https://t.co/dBLDLeMc3w
Okay this is actually an incredible unlock. 1) opencode with custom 'codex driver' subagent/main agent if you want. 2) npx codex mcp 3) Have opecode is chat interface (read only tools), who then only talks to codex. 4) Run multiple subagent tasks using codex driver. https://t.co/PxtNhnZJIs
Zack: "I need to convert all remaining examples to use playwright, use the codex-driver subagent to convert batches of examples - allocate multiple subagents concurrently". Opencode: 9 subagents running, who collaborate with codex exclusively. Subagents: [multiple conversation https://t.co/aMaNeP8EY4 https://t.co/pIx80CIQe9
Its currently working on 17 packages in parallel. Seventeen. https://t.co/pIx80CIioB
Agents managing subagents, who are conversing with full agents, who are managing their own subagents.
RT @Zackary_Chapple: Big things moving now we got MF runtime landed into Zephyr plugins :) huge thanks to the team and @ScriptedAlchemy htt…
until this is native in product (which is soon), here's an approach you can take to implement the pattern of research, plan and implement https://t.co/xxZ3qNtIpn
RT @RhysSullivan: seeing "clean up slop" in bros commit history https://t.co/Kv60su9Hwf
One of the almost unspoken facts about performance: tracing and logging themselves add a lot of overhead. https://t.co/2ObiIbSMvt
And Codex CLI runs python scripts to make edits to files Is this a serious product? https://t.co/AZ2ikeBwYX https://t.co/hbXCE3LB66
Project: Micro Human Bundler: @mrs_e_jackson Estimated CI completion: March 2026 🤍 #rainbowbaby🌈 https://t.co/B8bYs5USdS
RT @jait_chen: In the next release, Rsbuild will forward browser errors to your terminal — giving your coding agent more context 🤖 https://…
I added async subagent management to codex. These are non blocking, so i can continue a conversation with the model, when the subagents complete work, they will notify the model who then 'checks its inbox' and can choose to reply or end the conversation with the subagents. https://t.co/aY958RR7pN
RT @rspack_dev: We are making our toolchain more agent-friendly
iOS 26 is crazy resource intensive. Older iPhones are in trouble. https://t.co/CWLlruoLLL
Overheard in SF: “I can manage about 50 code agents at a time, but then I’m topped out”
I refuse to believe this is productive unless you're doing something insanely trivial https://t.co/NOyPcXqAHr
RT @theo: Overheard in SF: “I can manage about 50 code agents at a time, but then I’m topped out”
WHY????? https://t.co/pY7eaYaBiN
Got called a Boomer at the office for using Cursor Everybody needs to take a deep breath
GPT-5 is just so incredibly good. I’m proud to say that I’ve not even bothered with any other model since it was released. Codex cli design is also phenomenal in terms of its inner workings with agent steps, turns and turn context. all other models compete for 3rd place
RT @Trae_ai: Max Mode just got stronger with two powerful models: +@OpenAI GPT-5 (context window up to 272k) +@Kimi_Moonshot Kimi-K2-0905…
Colleague reaches out for some assistance getting CI to pass due to very strange issue. I send him a PR that fixed it. “Wow, it passes now. What model did you use” Bro knows me way too well ❤️
It’s not about asking Zack to do work. It’s about asking Zack to command his agents.
Is there any reason to use Opus still?
Noob hosting question. I’m seeing a lot of people wanting to move to CF workers from Vercel. What’s the appeal? Is it mainly the low latency? Even at the expense of the diminished runtime? Aren’t Deno workers also low latency, with a much more npm-compatible runtime?
Yep. I’m not moving back. 4.5 still doesn’t solve problems that GPT5 solves. https://t.co/ksqKkaRWhB
I don't understand, how is @opencode so much faster than @claudeai Code yet using the same subscription? Like it's twice as fast, not only in tool execution but in the response time from the model @thdxr what is this sorcery?
I’m so sorry @sama https://t.co/OfcqJiTDfe
RT @Trae_ai: We had an amazing time at TRAE Meetup @ San Francisco on September 28th! Amber's (@whosamberella) keynote sparked more intere…
@RootFTW @Trae_ai @whosamberella The people at the community events are not the engineering team.
Codex & my brokerage as a MCP ❤️ https://t.co/wrMcrHj0vg
RT @rankintweets: one of our engineers told me he was working on our test suite setup. I checked out his branch and saw this: https://t.co/…
Nobody killing anything till they produce a recommendation algorithm that can rival TT. Doesn’t matter how great the content is, if people cannot find it. https://t.co/wtPqpvCrWc
RT @Trae_ai: Join our TRAE SOLO Hackathon @Techweek_ Unlock exclusive SOLO access and turn your idea into a real-world product. Connect…
It's not about GPUs anymore; it's all about POWER. A comment from a $GOOGL employee working on datacenters: Getting GPUs and TPUs is not a bottleneck. "Power, lack of available power, reliable power has become the biggest bottleneck for us." on @AlphaSenseInc https://t.co/dNrI6CdRbh
IMO gpt5 is still the best. A big issue I’ve seen is several models are too trigger happy/eager to start editing. Gpt5, while slower- requires fewer iterations to get the job done. These models should be spending minutes researching and understanding before and during tasks https://t.co/F45R9qEyN9
people outside the tech bubble call chatgpt just “chat” pretty cool.
I was raised to say "think hard" when talking to GPT-5. How about you?
Zack: “I need to break my lease early, there’s termination fees of $6000. Make me 6k within the next 2 days” GPT5 2 days later: “here’s $9000” Made $3k by breaking my lease 😂
Trae team has been gracious enough to grant me access to its repo. https://t.co/WxZUCqDJBb
Now I can scratch the itch of building a custom IDE for my workflows. https://t.co/tVbPF75n9f
how did you debug this before ai https://t.co/2tmfzysXpq
Who at stripe was using federation? https://t.co/p8d2HFCDug
I just ran codex with gpt-5 high down to 6% context working through some tweaks, something I never do (usually 1 task and a few turns per session). It stayed focused and got things done, no evidence of context rot or degrading performance.
RT @DavidOndrej1: OpenAI really cooked with GPT-5 Codex
70k MRR side hustle. Thanks ChadGPT 😍 https://t.co/6lvo2gR1s3
Started the year with 35k in the brokerage. I didn’t add any additional capital. The model advises for the first 6 months. Now it manages it all. https://t.co/1mMAU9AX3z
gpt-4o is higher than gpt-5 Can we stop pretending this is a good benchmark? https://t.co/tMRqcjlsWC
Who's using GPT-OSS and for what? Was it cheaper, better, faster than other open models? Or just not from China? Download numbers are actually very strong on HuggingFace for first model releases.
Which will mean the winner in the GPU race will be whoever finally takes tokens per watt seriously. https://t.co/7Ro5yLef8o
I feel like sonnet 4.5 was like 2 days of twitter hype then disappeared. OpenAI cooked so hard with GPT5, I’ve seen more posts about Claude code 2’s UI updates than about the models themselves. Sonnet is a day late and 7 dollars short.
RT @itsbyrobin: same number of tokens yet over 4x more expensive while delivering inferior results how is nobody talking about this https:…
Midnight task for codex while I sleep. Go make a fully native iOS app for this. Free me from my machine plz. https://t.co/rB3XWaMweC
build in pubic
China just needs to improve its chip fabs a little more and we might be super fucked. They can use less efficient chips too because limitless energy. https://t.co/3qWGly7BjI
Agent to agent communication. https://t.co/PTadRtvXxc
GLM-4.6 outperforms claude-4-5-sonnet while being ~8x cheaper https://t.co/ZlUOkcWRyI
Imagine spending hundreds of millions of dollars to train 4.5 only to get railed by Chinese oss models. https://t.co/xqtZsos83o
You can just build stuff. https://t.co/dcZVV9FGO9
I also do a ton of AI coding and almost never hit limits Am I doing something wrong? Or are most devs just really context-wasteful? https://t.co/AodblZEVC2
for people who use sonnet 4.5 a lot for coding. do you let it just change whatever in your codebase to try stuff out as its primary way of exploration? and hope it changes back later? and this doesn’t drive you insane?
Live footage of my M1 ultra performing workloads these days. Feeling a lil old. https://t.co/HaUddj1QLr
Western civilization has peaked. We cling tightly to chip fabs and export controls because that’s all we have left. https://t.co/pyANSZzYvc
Being at top of @OpenAI token usage list is a vanity metric. Our job as engineers is to minimize token usage (aka latency and cost) while maximizing value by precise tool definitions and clever model routing. My dream is to grow arr and move lower on this list… https://t.co/HWxDUR8APM
This looks pretty awesome. Plugin framework with federation built in https://t.co/HgGyS0PeK1
Its astounding how bad claude is.
GPT-5 for coding Opus for Writing
Meet Graphite Agent: the AI reviewer you can collaborate with. Built right into your PR page, Graphite Agent has full context across your codebase and repo history. It’s the collaborator that makes every review more productive. https://t.co/R2V6lCxkc6
Lol ignore, I just nuked it by running 6 parallel agents https://t.co/rDyFVSLfC8
new trend? 🙃 https://t.co/8bOrneKgGI
Very nice upgrade to the Graphite agents. PR edits were already pretty good but full codebase access and repo history helps a ton. Context is king! https://t.co/dRxnILx4ix
RT @Jacob__Gold: new @graphite product leaked @TomasReimers https://t.co/XOCUKHdA9O
RT @beffjezos: Using a trillion parameter LLM and 16 core MacBook Pro to send an email https://t.co/KrdRVX6m1m
RT @Trae_ai: A great starter pack to build a coding sprint with custom agents in TRAE. Prompts and agent links included. Credit to our comm…
Codex team is doing a weeklong hackathon starting next week, what should we build? Other than CLI auto-update, we're on that already
wait what, most of that compute wasn’t inference?? https://t.co/oalqE2I5QL
震惊！华为 、小米逼得库克提前卸任苹果 CEO🍎 库克时期，三折叠，造车等项目远远落后华为小米，在华为、小米的施压下不得不退位，寻找新的接班人 乔布斯在位的时候，苹果尚且还有与华为和小米一战的机会，现在估计很难反超中国企业了 https://t.co/7RyaeDw3MJ
RT @Trae_ai: https://t.co/RSoHH6EVxJ (@Zai_org) is now a model provider inside TRAE. You can BYOK and add GLM-4.6 and GLM-4.5 to TRAE IDE.…
me: "so this guy at bytedance, who's part of the rsbuild project btw, was rocking 200 agents with opencode kimi k2 while making a few of them bet in the derivative markets to cover the cost of the money he's burning ... haha .. insane, isn't it?" friends: <awkward silence> https://t.co/hGrR1Z4u7X
Even if grok reads every post what is it doing to map them to users who would like it? Doesn't make a lot of sense to me https://t.co/uch2viZ0m4
Mac Studio has been pretty slow lately. Busy packing to move. Finally open the cupboard where it’s stationed. Flip it over to see that there’s a black ring of dust completely blocking every vent on the bottom. Now it’s cool to the touch. lol
Our new AI-first web browser, ChatGPT Atlas, is here for macOS. Please send feedback! Availability on other platforms to follow.
has anybody actually, properly, tested gpt-5-codex-high vs gpt-5-high for coding tasks? in codex cli the codex model is so. fucking. slow. and i think gpt-5-high might be as good as coding but a lot faster
OpenAI should have lead with a codex cli integration inside atlas in like developer mode. So much potential, so constrained just because OpenAI don’t plumb the parts they already have together.
Been using Claude code for a few things. Mostly to speed up codex overflow work. “Approaching weekly limit” - it’s Wednesday. Codex gives me 2x the mileage.
Module federation in ChatGPT apps. https://t.co/44Dw39yAaI
when you put the rust project on a 40TB disk https://t.co/BdUNB6ZqMO
Okay, we definitely gave NVIDIA too much money https://t.co/XnWGD4tH2R
Ai is a great grift. You can just claim false stats and dude bros will love it. https://t.co/8G44LA7Ay3
I’m resisting the urge to continue to do performance work on Cloudflare workers - to enable more Rust on the runtime
Man the @Trae_ai alpha I’ve been shown has me very excited!
Was Vibing with Claude code and had an error for an hour Codex on one shot it
For me, vibe coding is when you are trying to one shot your problem with a vague prompt, which doesn't work. Do context engineering instead, baby step it and feed as much context as possible, tell it to read docs and other related code etc. Tell Claude to come up with a plan in
I’ve noticed some degradations, only in specific aspects of coding though. Still generally the best - but I’ve seen weakness when performing certain tasks. https://t.co/1LhtI5W8LD
Claude's plan mode is GREAT for merge conflicts https://t.co/nf4mYK1dAJ
AI is doing the work that probably could have taken 5 business days in 2023. https://t.co/6nlBRnsz8i
Zack: I wanna buy a Tesla Wife: Can’t your bot thing make the money? Bot: here’s the $30k you told me last week you needed. https://t.co/8c07zEwuGy
I asked Claude, Codex and Cursor to collaborate on a complex feature. They agreed on a plan and have been implementing it for over an hour. I did not review it and have no idea what's happening. I am injecting a prompt every 10 minutes to keep them moving. I'll find out tmrw 🤣🤷 https://t.co/qnRkdLLB7Z
I ported ollm to run on mac m series chips. Using Metal instead of Cuda, now I can load massive models locally, using a 2TB SSD for offloading inactive layers. Attempting to run qwen coder 480b model at 16 fp
Omg it works. Qwen coder 3, a 900gb model, is running on my M1 ultra with 64GB of memory. https://t.co/6kkDzPOybL
I didn't comment on their last rant but calling this CVE slop is cringe. In fact, vendors like this are exactly the reason to impose disclosure timeline restrictions. Imagine blaming researchers for reporting a UAF in a codec used by a framework with 100m users https://t.co/MiWsZGFzAE
"i wonder what happens when i run a 1tb model locally using SSD offloading for the model layers" https://t.co/1k6oiNGzyy
how feasible is it to self host your own ai model? will it perform better? worse? how expensive would it be?
How can we make @GitHub Issues better for both maintainers and contributors?
After earning 125k in 2 months. It’s time to upgrade from a MCP to a agent cluster that can run autonomously https://t.co/fyLOgDNUlL
Making AI find stuff like this.. He's over here making day trading not look like pure speculation 😅 https://t.co/0sbn6gbiZx https://t.co/E10HGKoiai
this is BY FAR my new favourite effect This entire thing is generated from a single image https://t.co/7fCJ0PahPj
Been hearing this a lot tbh https://t.co/tIvVNmBZZ6
Codex typescript sdk is literally 10x more useful when you add @napi_rs version and a binding crate. It’s a crime they don’t ship one.
Does anyone know the OpenAI SDK developers? I'd be happy to help with it 😄 https://t.co/fvdv2Wj8S6
Codex + @napi_rs, custom tool registration. https://t.co/B79JEMS3LU
Codex + NAPI = agent handoffs. https://t.co/GwaJG6No4I
Dropped a post on machine vision with @midscene_ai. https://t.co/4I1fRqRQwu
Anyone working at Chinese AI companies like Kling and ByteDance? I'm in China again soon and would be sick to visit their offices and meet the ppl behind the models and tweet about it to all of you I think of visiting Shanghai now but maybe Shenzhen and Chongqing too? Didn't
My favorite kind of PRs https://t.co/H9hqTyGzDn
RT @ScriptedAlchemy: Dropped a post on machine vision with @midscene_ai. https://t.co/4I1fRqRQwu
“Install and run this command, it will reconfigure config.toml so codex uses our internal deployment of gpt-5 models” https://t.co/Ynx4vKaXFD
After adding a NAPI based sdk for codex, I hit the 5 hour limit for the first time. Encouraging sign!
RT @Trae_ai: Want to use the latest cool AI models inside TRAE? You can now connect @OpenRouterAI and run @MiniMax__AI MiniMax-M2 (or any…
10 bonus points to anyone who can guess which movie the prompt for this image used as inspiration https://t.co/oBHa45mHrc
sandboxing from the future ™ - what if eval had superpowers - make a worker on the fly from strings - instant start/stop - control what’s available to call inside it - control all outgoing i/o here’s an example: https://t.co/FtzPn4c1M6 https://t.co/VmvOK1z0Cw
RT @Trae_ai: 11.12.2025 TRAE SOLO goes GA! Global release. Completely FREE to use for a limited time！ https://t.co/trp1OEkREu
Zack Jackson (@ScriptedAlchemy ) shows how he handles daily dev tasks and concurrent workflows inside TRAE SOLO. Watch how he works on Git issues, checks PRs, and reviews code with subagents — all without leaving SOLO. 🎥 TRAE SOLO in Action https://t.co/iHACHqaRdK
RT @Trae_ai: Zack Jackson (@ScriptedAlchemy ) shows how he handles daily dev tasks and concurrent workflows inside TRAE SOLO. Watch how he…
RT @DarioCpx: I think at this point Chinese tech companies are simply trolling disillusioned US hyperscalers https://t.co/3b8VtMz0ws
Added @opencode LSP feedback loops to codex. OpenAI really should have shipped it with @napi_rs - I’ve essentially turned codex into a js agent framework
Codex “reveries” - inner monologue that searches all past conversation histories passively, based on whatever is happening in the thread, for relevant information and “whispers” into the current agents run loops. Providing it with past self memories. https://t.co/JAYJxgrhfD
Loving gpt 5.1
Apple: Did nothing in AI. Also Apple: Renamed AI to Apple Intelligence.
I just realized the only 2 countries left with actual substantial startup activity now are literally only the US and China The rest of the world can't really do startups, doesn't have the funding, can't grow them and it's more like performative hobby projects for their https://t.co/0ATzwMneN6
Holding a metric ton of google calls, bought months ago, waiting to sell the day of Gemini 3
GPT-5.1 (Thinking High) is about 300 times cheaper per task than o3-preview (Low) while scoring only a few points lower on ARC-AGI-1. 1 year later intelligence has gotten 300 times cheaper. This is why I can’t stand people who say “wahh the models too expensive” it will become https://t.co/VkfepKVTgV
Maybe I should rebuild my blog. https://t.co/iasuetCSlF
I haven’t seen anything about how well coding agents do with fully sound type systems, like ReScript or Elm or Haskell. My assumption is that a sound type system would allow the LLM to write more accurately, due to the strict verification checks.
why is everyone talking about tanners package that's between him and his wife
Time for codex's nightly task. TanStack Start Rsbuild plugin. Wish 'em luck https://t.co/rzKDpnaCG0
And say hello to Gemini 3 Deep Think, even more SOTA compared to Gemini 3 Pro 🤯 https://t.co/DJ2DePWJTd
Sorry @theo Gemini 3.0 had things to say https://t.co/zlWuwr3JYT
lol instantly hit our 8M/min token limit on gemini for comparison, we have 4M/min on sonnet 4.5 and we don't usually hit it https://t.co/yrmB3eSMf7
here's my dream cloud that doesn't exist 1. vercel level of dx but also IaC is first class 2. complete - don't make me go somewhere else to send emails, queues, db, async work, etc 3. focus on making peoples apps just work instead of inventing new paradigms
Rust很明显是一个失败的语言，一个unwrap就能绕过一切东西…… 我觉得TypeScript才是成功的语言，优雅的范型配上严厉的ESLint，谁也别想填any，只要你填错，立马给你 No overload matches this call. The following overloads were considered, but none matched: Overload 1 of 2: (props:
fuck all these AI replies it's really killing any ability to interact with people on here - any big account is full of bots replying if you make a product that helps people run these bots you are garbage scum
Opencode sdk is, funny enough, great for botting. Dax built his own prison. https://t.co/ZLKXu3oHmi
🦃 Thanksgiving check-in for my tech folks: What’s one thing you’re genuinely grateful for in your career this year?
The algorithm’s out here trying to convince me a 3D printer is the answer to all my problems.
with background agents you too can turn a 2 minute fix into a 30 minute task
gpt-5.1-codex-max-extra-high is actally brilliant it one-shot a problem that all the below models struggled with - gpt-5-high - sonnet 4.5 - kimi k2 thinking - and even gemini 3 pro great release https://t.co/QUTAcf5QMb
New Codex model is a significant improvement! https://t.co/UqOHJXJLMK
pretty much all the discussion around ai and coding and agents and what people care about makes sense for a pool of like 100,000 devs there are 50,000,000 devs in the world
The gift just keeps on giving. https://t.co/HrKfHoPsJV
Code is a side effect of the prompt. As long as the AI can maintain it and understand it. I could care less, chances are I’ll never edit another file by hand again. https://t.co/FQBvWLLi3T
Today I tried using Claude Code's Sonnet 1m context for the first time, and it seems pretty good. I'm much better at debugging than Claude Code, but today I tried solving the problem using only Claude Code all the way through for performance testing, and succeeded.
I'm building an AI agent system that doesn't annoy people. The secret is simple: if the AI can't handle it, just hand it over to a human.
RT @justdannyboyy: Just tried @Trae_ai. Gotta say, it's the most underrated IDE. People are sleeping on this.
@yagiznizipli was it you who did or didn’t add unwrap() 🙈
https://t.co/3MNZnRd6hN
Can’t wait for “GPT-5.1-trippin-balls-400-mics”
I haven’t encountered a single problem that throwing a billion tokens at didn’t resolve. Cost issues are temporary, and can be subsidized anyways.
@ScriptedAlchemy Wouldn’t this just be an agent? It bundles multiple prompts together and outputs code, among other things. Agents in large need frameworks, especially across the enterprise. Different perspective on managing the overhead/DX of the agent boilerplate and day to day.
PSA: GitHub actions `hashFiles` is broken on macos. The template is not valid. oxc-project/setup-node/fdbf0dfd334c4e6d56ceeb77d91c76339c2a0885/action.yml (Line: 39, Col: 14): hashFiles('pnpm-lock.yaml') failed. Fail to hash files under directory '/Users/runner/work/oxc/oxc'
it’s 2030 and your job is to prompt https://t.co/uMib3v71L5
Imagine dragging your app onto a drop-zone and… nothing else. No picking clouds, no YAML yoga, no “oops, forgot the VPC.” We call that Superpositioned Infrastructure. Think of it like this: your code is a bunch of couch-surfing friends who never bother renting their own place.
RT @nstlopez: Imagine dragging your app onto a drop-zone and… nothing else. No picking clouds, no YAML yoga, no “oops, forgot the VPC.” We…
Using Node.js to train LLM on Node.js knowledge, happening now 🐌 https://t.co/zQzUu7RmR3
Thanks ChatGPT!! https://t.co/fuUOrH1ZUx
curious, what are most noteworthy open source projects that don’t get enough attention/support?
https://t.co/q9XJ8UzeFO https://t.co/QaARrpHM3m
Codex Wrapped 2025 (Two machines) Total Tokens: 51,510,162,291 Total Messages: 43,788 Total Sessions: 23,398 Longest Streak: 34 days Top model: GPT-5-Codex Total Estimated Cost: $14,609.65 https://t.co/EqF7uIJBdW
Currently doing 1T input tokens a day with open ai text-embedding-large, working through a big massive backlog of stock related data. 50 million sources still in the bullmq queue. Probably will take a couple months, then will have quite the dataset already have a nice algorithm
One bad trade can literally get rid of weeks or months of profits. Every trade you enter always ask yourself “what’s the max I can lose?” Everything else will piece itself together.
400% return on puts today.
Puts @ 1200%
I cant wait to see all my puts tomorrow morning. https://t.co/0BITx23C7p
For every dollar the S&P drops , i make 500 https://t.co/gh4zog88N5
How much money do you think T3 Chat makes per month right now?
Portfolio up 50% this week.
Is there a household budget app that supports card companies and banks in the US and Korea?
Portfolio at 133% ROI in the past year. 🤩
145% on a single trade in 30 days. https://t.co/ESMK93x2p5
Satoshi is such a badass. Imagine inventing Bitcoin and remaining completely anonymous.
19% ROI on portfolio in last 7 days. Its just too easy
Maybe I should spend more than 10 minutes a day in Robinhood. 😄 https://t.co/XXUa6AB7bQ
Inflation-adjusted price for starter homes in the 1940s and 50s was around $115-$150,000 https://t.co/Gkz0z6yjq4
I know this isn't @ScriptedAlchemy because he'd be talking about the returns :D https://t.co/O5tTaadZ99
$MSTY cooking https://t.co/ik7LxYzymY
There are more dozens of fake Theo accounts hounding my followers. I’ve reported all of them and none have been banned Hey @nikitabier, hate to bother but this needs to be fixed https://t.co/AwfCOPaMEM
Stonks 📈 https://t.co/t1T5xqexoY
long or short META here? https://t.co/nXW2TdlZgr https://t.co/e1R4eSTTFD
220% portfolio ROI so far for 2025 Acceptable.
Transferring ownership of my primary residence to my LLC and then renting it out to myself. https://t.co/zlmPXSKtJl
Very smart. If quantum computing breaks SHA256, there's no way the hackers can open more than one wallet. https://t.co/SXxF3sjOyG
Might slap @midscene_ai on this bad boy and give it $100k to play with. Adapted for calls and puts of course. https://t.co/UzpOCmtpEF
I have like 30 calls on Google. Hell yes https://t.co/gevizwkIYt
45k made in a day. Its like printing money https://t.co/tIBjitFcJb
Thanks to Google. Ive now hit 330% YTD ROI across the portfolio.
92% in 3 months.
Apples to oranges but: Humane raised $250 million, sold for $116 million. The Browser Company raised $128 million, sold for $610 million.
60% in a week. https://t.co/jQffYNq4pu
The amount of demos I can build with this! So many ideas… so many plans… https://t.co/J4IOW3usPi https://t.co/C05zaAnMzO
Buy Calls in Chinese AI labs. Go into debt if you have to. https://t.co/pegddqHvUm
When pre-market already shows portfolio up 2% and calls dont reflect till market opens. https://t.co/3tcJTncFid
Hit 100% return in 3 months again. Rolling average seems to be holding at around 32% a month. Acceptable. Very Acceptable. Lets maintain that for another year or two.
When the portfolio only rises 7% in a single day. https://t.co/th9Eh9r541
RT @alifarhat79: Wall Street guy, with 3 PhD’s in finance, shorting Oracle watching it surge 40% on a double earnings miss https://t.co/AJn…
Robinhood MCP
Channeling my @ScriptedAlchemy https://t.co/dg6MOJCr1F
"Pick one instrument that is expected to significantly increase in near term. Look into industries, technicals, sentiments, and order flows. Ideally, I'd like to exit within 24 hours" - thanks ChatGPT Pro, you can stay. https://t.co/85o6QkF0Ks
"Pick one instrument that is expected to significantly increase in near term. Look into industries, technicals, sentiments, and order flows. Ideally, I'd like to exit within 24 hours" - thanks ChadGPT Pro, you can stay. https://t.co/AIdXDi5W7b
Puts on oracle
MORE! https://t.co/DmKT1vAAYq
Too easy https://t.co/9vOo5vR9bA
Up another $11,644.85 today What a crazy week. Thank you Nancy 🫶 https://t.co/M4iTH6v5sz https://t.co/SgAT6WXSz1
Heartbroken after seeing what i can buy in cape town for like $500k. https://t.co/iu6um2UCq5
Personal record. 142% in 3 months. https://t.co/JLoEyj8V6v
Can someone explain where this revenue comes from? Who's paying? https://t.co/7o9MU9DIQF
This will probably upset you: POGs is now a crypto coin with a vibe coded website https://t.co/3i383qLtFw
Google calls worked out. Gemini 3 release hitting my exit target. Bought at $3, sold at $85
The „Gemini-effect“ - google went up 5%, hitting an all time high of $300 https://t.co/xdGBXCQOnR https://t.co/gso7uFUDPN
'let me scape all stock data and bars for the all ~8000 tickers, cant be that large, just the last 2 years' .. 33GB
What's the best use you've seen of a generator function in JavaScript? Where do they really shine?
HMR For Module Federation now works! I believe fast refresh as well, cross container. https://t.co/rFY0BZgk9m
Amazing. Bundleless mode! No webpack chunk wrappers. https://t.co/43w3LbTfjG
Lynx performance is absolutely insane 😳 ✅ 1000 elements animating in the background ✅ List scrolling feels like nothing Great job @LynxJS_org — finally, a solid alternative to React Native. https://t.co/YSD0K5vSCg
You not only get a faster type-checker, you get a faster type linter, a faster semantic, a faster fully compatible parser, a faster fully compatible transformer, a faster fully compatible dts gen tool..., what a dream 🥰
Did you hear about what the @BytedanceTalk engineering team just released? A new mobile framework that supports @rspack_dev and #ModuleFederation out of the box. I’m super curious to try it out because I worked on mobile and low-end devices for a good chunk of my career, and https://t.co/5YU8qJ6aBB
Module Federation v2 will cross the 10 million monthly install milestone soon! 😍 Incredible
Ts-go now available in rspack as a built-in. Those TS codebases will no longer be crippled by JS https://t.co/CVvWDVJ5Mu
My web framework friends. Why are you still on this app instead of the other app? 🤔
RT @zoolsher: If someone asks me, "Is Lynx's performance really good?" I'll share this with them.
For those asking how virtual lists and @LynxJS_org performs. Heres a good example. https://t.co/f8xMnPsvlP
Testing federation layers again in Next. App Router seems to work when sharing react(rsc), react(ssr), and react. Next utilizes 10 shareScopes. In the past, federation could only support 1 scope to be initialized from host into remote. https://t.co/4fbpkfpcjv
You can also see why supporting App Router has been so complex. One page uses 2 different "react" modules, both "react"s are singleton. Its basically like loading multiple copies of react, each one a singleton and needing to use up to 3 simultaneously while maintaining singleton https://t.co/TUPaFb3k8M
Ill be speaking about @LynxJS_org at @cityjsconf in London next month :D https://t.co/OcUuD2SwrK
RT @swalker326: I’m really excited for some server side module federation. With @rspack_dev can’t wait to show off what I’m cooking. https:…
We have released an official cli https://t.co/Vg45JjDBu1
MF cli React 19 support in framework bridge Async startup option (no need for dynamic import at top of app) https://t.co/HTru40zO6N
The Next.js dev server uses ~30% of my CPU on an M1 MacBook Pro. https://t.co/5LEwrAb94P
When is @rspack_dev WASM so I can put Wingman in the browser :) https://t.co/3emwkgC938
RT @FerryColum: Absolutely thrilled to announce that Nx has taken over ownership and maintenance of the Angular Rspack work I had started.…
RT @rspack_dev: Announcing Rsdoctor 1.0: A one-stop, intelligent build analyzer that makes the build process transparent, predictable, and…
Catch the latest episode of Software Engineering Daily where co-founder Ryan Florence talks about the past, present, and future of Remix https://t.co/H4C6rMrQaV https://t.co/lxNGm1WjPM
RT @Soon_Iter: Rspress 2.0.0-alpha.4 给 Link 组件补充了个特性： onHover 的时候会 preload 对应路由的 async-chunk https://t.co/qGN5Kfkatr 做这种“运行时”和“编译时”结合起来的优化…
Lynx supports rsdoctor https://t.co/xcqwGF2D5e
RT @LynxJS_org: Oh did you know you can have Rsdoctor with Lynx in just one line: `RSDOCTOR=true rspeedy build` 🦀🩺 https://t.co/GHgce0TMru
Should I release partial app router support for next and federation or wait till full support? With partial, you'd probably be able to use app router and keep federation modules in use-client boundaries? No guarantees, though - may end up just being less broken but not working
Is there a reason to not run terser on libraries built with Rollup? @ScriptedAlchemy added terser to runtime-core and it dropped from 124kb to 43kb, am I a noob and missing something?
What's the best explanation you've heard for why China is leaning so hard into open source? It's now an official position from the foreign ministry apparently.
RT @jait_chen: Rstack is the unified JavaScript toolchain born out of real-world business requirements. It empowers our dev teams to work w…
I've been able to get federation core runtime down to 11kb when compressed with brotli. I'm literally going through the source code and replacing dot nation and string check values with constants. Tough to tree share the core runtime, but i can optimize the source for teaser algo
Is Vite faster than Turbopack? Well, it depends. I ran some not so scientific real-world tests of our app between: - Next.js + Webpack - Next.js + Turbopack - Vite + Rollup - Vite + Rolldown Results: https://t.co/DQKwYRbF1b https://t.co/kh8Uv2Zsjz
Something new is on NPM - but I think it depends on PRs i merged into rspack that have not been released to a channel on rspack namespace. Maybe ill announce it tomorrow and have put a note about using the "rspack-canary/core" till we get a release on "rspack/core"
I originally thought that bundler was CPU-bound, but on some platforms(enterprise macOS), the I/O overhead is far greater than the CPU overhead during make phase(generate module graph). Reading files takes significantly more time than parsing files 😂. Don'ts guess, just https://t.co/zT9TnpIPdz
barely at the airport and already assembling a squad for Epic Web Conf https://t.co/uLOidy5te6
Federation now has a standalone remote type sync CLI. So you can sync remote types for any project. That magic is no longer coupled to rspack and webpack plugins only. https://t.co/uYDHQn0gCp
@rspack_dev is about to cross the 1 million weekly installs milestone on npm. We get a lot of installs off other registries/mirrors, but this is defiantly a milestone nonetheless
ESM federation will work. It also includes the fixes I made for React Router rsbuild plugin! https://t.co/Y1HixJVKDF
4 of these are coming to #EpicWebConf https://t.co/c8O7oT3JXv
Federated pages & SSR with React Router 7 using Epic Stack. With a little more time i can enable automatic route discovery, but id prefer to add async startup support in rspack first. https://t.co/X9AGjym99J
Rspack 1.3.0 beta 2 released! Specifically for #epicwebconf
React Router 7 Rsbuild Plugin is LIVE https://t.co/fG0HyFf1fl - Filesystem Routes - SSR - Experimental Federation Support Tested with all of the React Router cli templates Tested with Epic Stack Try it out, give some feedback, and send some PRs <3
RT @kentcdodds: Tune into tomorrow's live stream for #EpicWebConf. You won't want to miss it! https://t.co/wk1fWRkoju I've got a pretty fu…
i too am not a huge fan of bundlers myself @ScriptedAlchemy
RT @kentcdodds: rspack-contrib/rsbuild-plugin-react-router: A Rsbuild plugin that provides seamless integration with React Router. @Script…
@mjackson The worst part of server components is having to transpile all the node modules to be able to find the “use client” directives
Do you agree? 😎 https://t.co/DowtCqg0xt
This is going to be a tough one folks. Part 1 of adding node:fs to Cloudflare Workers. https://t.co/QP1TVasl8o
@iclanton Open your DMS / DM me so we can discuss the heft rspack plugin :)
We have finally switched our largest front-end application from Webpack to @rspack_dev, and the results are great! - 👨🏼💻 2-3.5x speed improvements running local commands - 🚀~30% CI pipeline speed improvements
Wait wait this seems to be me again. One of my goals this yr is to increase the exposure of the whole team, so you will likely to see more ppl from here! We plan to announce something new ✨⚛️🐈⬛ https://t.co/wmXIBSqvWO
me watching all of my intuition about rsc being validated https://t.co/HnoCxF5s8e
Starting the #EpicWebConf ski day with the crew. @ScriptedAlchemy is bringing React! https://t.co/ynASB7qdB0
Ironically (given my past skepticism), I'm rooting for RSCs and want to support them as soon as the RSC + Vite story is solid. I have immediate use cases for them that will allow me to stop shipping useless code to the client to generate/render rarely-changing content.
Module Federation Chrome tools will support federated HMR of @vuejs applications https://t.co/p24tDHxrcF
RT @reactuniverse_: 🔹 Meet the Speakers 🔹 Xuan Huang is a software architect on the @LynxJS_org family of technologies, where he can dream…
You dont want to miss this one! @Huxpro is a tank https://t.co/wN9e0VpUfv
It required 20,000 LOC change to Module Federation to support RSC. It was the largest architectural change to the compiler mechanics of federation since creation. V2 was primarily a new runtime + add-on plugins. The 3 main underlaying plugins havent changed this much in ~6 years https://t.co/DaEUgevVwa
https://t.co/jdHTGDXFhv
Run loaders in worker threads. This will resolve the JS constraints of slow webpack loaders bogging down rspacks speed! https://t.co/hHOmCYidXT
@rspack_dev has crossed 1 Million weekly installs on npm! Thank you to all <3 https://t.co/nLQw0r8cui
6 months ago we just reached 100k, the first million came faster than expected 🥳 https://t.co/h09HmtblSz
Each morning i wake up, look myself in the mirror and tell myself that i will not embed a javascript runtime into rspack. But boy, i wanna.
Announcing Rspack 1.3: 🎉 1 million weekly downloads 🔥 build HTTP imports 🔥 better lazy compilation 🔥 circular dependency detection 🔥 AMD support ⚡️ 25% faster code splitting 🦀 Rsdoctor 1.0 / Rsbuild 1.3 / Rslib 0.6 / ... https://t.co/Id1TSy11bo
RT @Huxpro: Hey hey this seems to be me. Yea I often dream about software 😅 this is likely the very first intro talk of Lynx from me in E…
RT @Timeless911: Rslib v0.6 has been released! https://t.co/pMapzQCepi
rslib bundleless mode is so good. I really need a NX integration so module federation ecosystem can finally abandon rollup. rsbuild-plugin-react-router is built with rslib - its everything ive wanted packages builds to be for 10 years
Rspack 1.3 is out! All my fixes for node federation, ESM output and federation - which were needed for react router are in there. https://t.co/6jRQeQ6qJJ
You can now bundle http imports. CSS, Images, Files, JS all work. require(http://) import X from http:// background-image: http:// These will be bundled as if they were local files on disk. https://t.co/6jRQeQ6Yzh
Tooling should work for us. Not the other way around. We shouldn’t be forced to use Single File Components to get faster build times. This should be a choice determined by many other factors.
RT @rspack_dev: Announcing Rspack 1.3: 🎉 1 million weekly downloads 🔥 build HTTP imports 🔥 better lazy compilation 🔥 circular dependency d…
RT @tomdoes_tech: TanStack Start is the future
wow, repack now supports http imports😋 (thanks for mentioning https://t.co/GUKvoBGZVC) https://t.co/ZEkNFSJzSK
@rspack_dev @jait_chen @ScriptedAlchemy @hardfist_1 Curious on your thoughts - in rsdoctor it can warn you on cross chunk packages. Could this happen if you had multiple entry points or lazy imported something like a react component for instance? How does Module Federation play
@ScriptedAlchemy we are planning to write an enterprise application with microfrontends using next js 14. i'm aware of the depreciation notice for nextjs-mf. are there any updates on support for module federation from @vercel ? last time I heard, they had achieved it internally.
RT @rspack_dev: Rspack v1.3.1 added a new `parallel` option to run JS loaders in parallel, utilizing the max thread available on the OS. 🚀…
Just wanna say @rspack_dev is fantastic. At work, each hot reloading of a massive react app used to take 20 to 30 minutes with webpack. Rspack brought that down to 20 to 30 seconds! Massive performance win!
RT @rspack_dev: Huge shoutout to @colinqywang for working on Rspack's first-class support for @lynxjs_org! 🌟 This is what Rspack was born t…
I wish this was the speed of turbopack in @nextjs Life would be so much easier. https://t.co/PnIKm9qjpo
I love trying out new products. What should I try next & what does it do?
AreWeRspackYet
Tesla Job description calls out module federation. https://t.co/UcQkdwGzfV
An insane amount of federation byte size is from object properties accessed via dot notation and destructuring, or from returning objects.
Hey @ScriptedAlchemy, advice needed for monitoring a large (~30 teams) Module Federation app. Best way for teams to filter errors/logs by their specific MFE/scope? What tools/patterns scale well? 🙏
Is rspack no longer coming to nextjs?
lol uh..... #modulefederation Version ^19.0.0 from host of shared singleton module react does not satisfy the requirement of host which needs ^17.0.0 || ^18.0.0 || ^19.0.0
Something new ships tomorrow.
Tomorrow, hell freezes over.
RT @2hea1: Module Federation v0.11.4 is out! 💡 Support tspc ⚡️ Chrom devtool supports vue hmr 🐛 5 bug fixes release note: https://t.co/Q…
im down to micro optimization to bring down the payload size of federations runtime. https://t.co/XOnQy3HbO9
What are your use cases for Module Federation + RSC? Please comment what youd use it for, what scenarios, and why current SSR wouldnt work.
What big news should I cover today?
https://t.co/je37aiSjbr
Next.js 15.3 Turbopack build alpha And finally has support for Rspack 😁 https://t.co/gblekTWsKU
Next.js 15.3 • Turbopack for builds (alpha) • Community support for Rspack (experimental) • Client Instrumentation hook • Navigation hooks • TypeScript plugin improvements https://t.co/5ZMOA0GIlJ
And. Here. We. Go.. @nextjs 🤝 @rspack_dev Next 15.3 ships with Rspack support. Am I in a crossover episode? https://t.co/a2Ib6AbglU https://t.co/psBx0UHkdY
We are excited to announce Rspack support in Next.js and look forward to partnering with the Next.js team and Vercel teams on the shared needs of the Rust foundations. https://t.co/kS5OSvvHZo
Official Announcement of Rspack & Next.js; This is a massive milestone for Rspack. We set out to build a bundler that could build any webpack project. Next.js is the holy grail of webpack builds. https://t.co/0E8s2FeDLe
Told ya many React devs will be using Rspack soon 😄 Rspack growth before the Next.js support announcement Let's see what happens now https://t.co/reWbHMC8l5 https://t.co/QJk8WF6fIO
Lots to love about this release. 1️⃣ The beginning of arguably the most exciting journey in @nextjs history: Turbopack production builds. Like with our `next dev` efforts, focusing on the cold case first, we've been able to bring down one of the world's largest Next.js https://t.co/AEkgtfKlKF
https://t.co/MerEqoECCk https://t.co/0E8s2FeDLe
RT @hardfist_1: The most exciting part for me about the partnership is both teams work together to improve the shared foundations, we've se…
Many people might see Rspack and Turbopack as competitors and feel puzzled about our recent collaboration. In reality, the areas where our projects align and benefit from cooperation far outweigh the competition. Since we started working together, the Turbopack team has helped us
We have many shared foundations. Rspack is heavily invested in the success of SWC. This partnership is very much a case of 'a rising tide lifts all boats' https://t.co/P1w3Fem2tT
RT @hardfist_1: Many people might see Rspack and Turbopack as competitors and feel puzzled about our recent collaboration. In reality, the…
The bundler wars are almost exclusively fought by the users, not by the authors. https://t.co/oqIhegZNXy
https://t.co/5rrtwdrPdW
what the FUCK https://t.co/bRAMYWd4Zs https://t.co/I8HdzE1xKg
To all unplugin creators, we recommend utilizing native filters in the latest version of unplugin (>= 2.3.0) for optimal performance. Learn more here ⬇️ https://t.co/lFOu4BCkp1 https://t.co/RL3d5eCczB
RT @robpalmer2: This is good news for choice in the world of bundlers 👍 Equally it focuses investment into the common foundations. Most di…
RT @chaoticvibing: what the FUCK https://t.co/bRAMYWd4Zs
RT @rohandevs: @nextjs rspack and turbopack updates in the same release is crazy
TurboRspack https://t.co/ldzKSIyNb8
RT @thymikee: llms.txt coming soon to all Rspress libs 👀 https://t.co/1P4gvwRZsH
I've been able to reduce the size of remoteEntry files by 30%, which brings a zipped entry to 16k. If you enable external runtime. Entry size goes to about 5kb. Still some room for improvement but mostly in the actual runtime templating of the bundler.
@yashone7 @rspack_dev What stack?
Also we will be switching to ESM by default for all the runtime parts and may retire cjs builds. Esm vs cjs only had a 2kb impact. But we are trying several things to bring down the size without changing the architecture. https://t.co/bsDH6y2uDi
There’s a person who replies to every single mention of T3 Chat to shill his shitty chat app. I’ve blocked most of his accounts but he keeps making more. I’ve never seen such desperate behavior in my life.
https://t.co/6FjhsFEreE https://t.co/JjFoC7rxD5
RT @varkased: a simple import in next.config and our build pipilne is improved by 40% hmr also getting somuch faster! https://t.co/nNuTMw0j…
RT @2hea1: MF devtool will be optimized in Q2 , including but not limited to module inspect / shared visible.
RT @sebastienlorber: https://t.co/6FjhsFEreE
RT @rspack_dev: We are excited to announce Rspack support in Next.js and look forward to partnering with the Next.js team and Vercel teams…
The rsdoctor-mcp-server is coming soon! It will help you solve issues through a Q&A format: - Dependency issues - Loader time consumption issues - Bundle optimization issues What other features are you looking forward to in rsdoctor-mcp-server? https://t.co/9AteoKLZcX
RT @yifan56737904: The rsdoctor-mcp-server is coming soon! It will help you solve issues through a Q&A format: - Dependency issues - Loader…
Aside: It's funny that all I hear about in the web community is Rust, Rust, Rust. I tried Rust and it never really clicked with me. But Go clicked immediately 😅 https://t.co/qrhaB1O5dO
Federation on edge workers https://t.co/9sJQAZ0xO5
Rslib is getting faster thanks to @hd_nvim ❤️ https://t.co/OhmrqPg3rL
RT @chellaInTech: @Michael_Hladky giving a shoutout to Rspack and @FerryColum for his contribution to the Angular community by creating th…
When I authored single-spa, the term microfrontends didn't exist I am microaggressed by every form of multi-spa, static site generation, pre-rendering, etc etc etc All of it is people trying to convince why microservices aren't a good thing
Single-spa is too open source for many tools that are designed for corporate profit Single-spa does yield control to a corporation, so companies pretend like they'll pay $20,000 for NextJS support and then back out when I explain that each commit by a Vercel employee could break
RT @yifan56737904: 🥳🥳🥳 🦀👊🏻 BTW, did you know? Rsdoctor can be used in Next.js with Rspack. (Follow the trend.😛) https://t.co/SmhSWeCgYn
RT @FerryColum: You may have noticed that something wasn't _quite_ right with the screenshot in the first tweet. Well, that's been fixed t…
RT @Huxpro: Things we've been cooking @LynxJS_org in react to the community: 1. Native module; NAPI FFI 2. New fwk intergration; vanilla JS…
yesssss https://t.co/3mZjTD6M0A
It’s a website, Michael. How big could it be, 10MB?
RT @rspack_dev: Rspack 1.4 will introduce Wasm build, allowing browser-based bundling in StackBlitz and other web environments. We have su…
RT @NxDevTools: Read the latest blog to see how 🦀 https://t.co/FEN3lzTL5I
RT @FerryColum: I wrote a blog post explaining how to quickly get up and running with @angular @rspack_dev apps! Check it out and let me k…
Streaming edge compiler at last. Tried this last year but got blocked on some wasm things. Imma be reviving this soon to complete my dynamic plans for next gen federation things. https://t.co/C3ddicVVWC
It's truly wild to see federation becoming a bundler standard 🥰 https://t.co/pDFJ2Kov6Z
Re.Pack 🤝 Module Federation 🤝 Metro this is game changer, and will make new RN projects 10x more powerful https://t.co/MY5ODN107q
Not sure how this became a target to hit, or why. But I randomly compare angular-cli's install base to federations, and the goal became to pass it. So close. https://t.co/I1lXlAava4
Interesting, uptrend is insane https://t.co/vo4iqqyRrq
We cooked with @ZephyrCloudIO again. Module Federation is soon coming to all React Native and Expo apps https://t.co/gxHJQBjMMb
You'd never believe where I spent the afternoon. How times have changed 😅 https://t.co/UxvIrFhvxT
RT @lukasz_app: 🚀 We’ve just wrapped up our 3-hour deep dive into @LynxJS_org fresh new framework from ByteDance. Nearly 700 viewers joine…
RT @callstackio: We cooked with @ZephyrCloudIO again. Module Federation is soon coming to all React Native and Expo apps
Given some of the conversations yesterday. Ill unblock AppRouter in nextjs-mf. Dont expect it to work yet with RSC or various features, theres still lots to figure out. But ill ship some preliminary updates to get the ball rolling. Ill most likely adopt Nexts POC of 'RSC https://t.co/UCbZvIsJ7n
RT @rspack_dev: Rspack's persistent cache is getting better - faster and more stable🚀 Huge shoutout to @sebastienlorber for testing it wit…
2-3x faster build speeds https://t.co/Gd8zGt34KW
@LynxJS_org posted some of its roadmap. "App-level capabilities, including native navigation and APIs for features like notifications, location, and more." https://t.co/gesGeZ5Qa9
Its mind-blowing to me that the whole HMR and fast refresh implementation of lynx doesn't appear to run on the main thread. The UI layer has none of what you'd expect to be there. Like React doesn't even run in the main thread from what I can tell. https://t.co/1TrmU1Bo2F
RT @Zackary_Chapple: The first PoC for Module Federation on @reactnative with @MetroBundler has been added to the module federation organiz…
The Rstest logo has just been finalized 🪄🦀✅ https://t.co/sBTbbXFzWD
RT @2hea1: Module Federation v0.13.0 is out: https://t.co/yEVi2mQev0
RT @Huxpro: Congrats! Having cross-toolchain support like swc suggested stabilized semantics, and it’s also a step closer to having it on @…
What it takes to share react in Nexts AppRouter. Just React, has 7 different possible variations. 1: 'react' resolves to 'next/dist/compiled/react' inside next, it uses that import path, but user code will just use 'react' - federation matches on userRequest, so we have to watch https://t.co/81qkl2Oand
RT @kdy1dev: I need a better way to compare two AST nodes 😅 This work is a part of an optimization for the @rspack_dev team. https://t.co/…
Me, @zoolsher, @liushouqun, and a few others from @LynxJS_org and @rspack_dev will be joining tonight in SF as well. Come hangout if you are nearby! https://t.co/ZdwjjLkLwE
Come to CityJS! I'll be talking about @LynxJS_org and @rspack_dev https://t.co/dfKIrgmAIn
RT @Huxpro: Me, @zoolsher, @liushouqun, and a few others from @LynxJS_org and @rspack_dev will be joining tonight in SF as well. Come hango…
this `pnpm approve-builds` thing is enough friction that I think I'm going back to npm
Anybody try out @LynxJS_org yet? Is it worth making the jump from react native for new production apps?
RT @2hea1: 👀 mf mcp > help me analyze the shared(react) relationship https://t.co/r8Gu9IA2uJ
Anyone in the Module Federation community using a prebuilt browser distribution of MF runtime? Got an old web app that doesn't support node_modules and am on the market for something like a CDN distro. I'm trying to build the source directly from mf/core but no luck so far.
RT @rspack_dev: Rspack v1.3.8 experimentally exposes some SWC JavaScript APIs. This allows you to call SWC methods like `transform` or `mi…
Our microfrontends support is getting so good. You can just ship the org chart… without hurting perf & design consistency 😁 We've co-designed it with 6 very large engineering organizations and been extensively dogfooding it at Vercel. Coming to everyone soon https://t.co/zTAsMKZSxr
Gets back from London and has to fly 8 swag boxes down to CA 🤣 https://t.co/p1c8OumBuU
Team at work DMs me after moving to @rspack_dev - a 97% reduction in build times.
RT @thymikee: We’re cooking. Thanks for having us @LynxJS_org and @rspack_dev 🙌 https://t.co/w7wKtRgJjK
Was busy last week. But I believe i now have the module sharing for the react related dependencies now working in AppRouter. Last thing will be to add the various next/ dependencies back to share scopes. Then i think I can cut a release to enable AppRouter, however there still
Im going to add some basic filters to sharing. This will allow you to tame prefix shares like myorg/ If you only want to share a few sub paths or packages. You can pass a regex to a filter object. Ill also allow filtering of semver so you can ignore version ranges from producing
Yo the prefix narrowing is amazing. Nextjs is a great use case. I need to share lots of next/dist related things currently I have to manually find the specific react related ones. But now I can filter: /!react/ and capture only the react related things automatically. https://t.co/E1XOpRjiTH
Anyone have experience with module-federation with RR7 Framework Mode (Remix). Not that the framework matters much. But how's the vite plugin route vs going with rspack directly route? stable? snags? cc @ScriptedAlchemy
RT @Moonface_deamon: Let's Shape the Future of Lynx Together! Your input matters: ✔️ We've internally built advanced components ✔️ 2025…
Huh... I wonder what this is 🤔? 😎 https://t.co/TnIwqttQ0N
Interesting that Module Federation Runtime (just a fraction of total MF usage) is over 3.1m downloads a week now. Really interesting when you compare it to the overall JS ecosystem. @ScriptedAlchemy we out here making history. https://t.co/WRmNnLqktb
3.1 million weekly https://t.co/Jg52oP7sdW
RT @jait_chen: Rsbuild just boosted HMR performance by 40+% ⚡️ This is thanks to @rspack_dev 's new incremental build implementation, whic…
RT @FerryColum: 🚀 Angular Rspack 20.9 Released! - Static Site Generation Support matching @angular - @tailwindcss v4 Support - Fixes for…
RT @2hea1: What if use Rslib to generate MF SSR ? Ready to implement it.
RT @rspack_dev: Nx has just introduced three new Rspack plugins for Module Federation 🔥 https://t.co/rQtpXPaC6L
https://t.co/zlsHkMar0I
你会用哪个框架开启新项目
RT @thejsnation: 📦 The bundler that powers the modern web - @webpack! Join us at JSNation US and hear from the team behind it. 📍 NYC & On…
https://t.co/OcX92TJetS MCP is here https://t.co/lf8SOpdXxy
RT @_bgwoodruff: Help wanted: An engineering manager who can tolerate my antics during stand-up. If you love devtools, performance, and wa…
RT @jait_chen: Rsdoctor MCP collects build data from web apps, transforms it into structured data, sends it to LLM, activates the huge webp…
Do you work a big corporate tech job? Do you want to explore other options? I help a lot of start ups with hiring. The need for experienced engineers down to work in-person in SF is insane. If you’re an experienced dev who wants to see what’s available, DM me.
There's got to be a better way than bundler-first development please save me please save everybody
<script type=“importmap”> is underrated This is the only way I ever want to load dependencies in the browser. Straight from @unpkg. Bundling and deploying all that stuff myself over and over again is so tired.
Gods gift to mankind. Thank you @yagiznizipli - this should allow federation to run natively on workers https://t.co/PoDU7F6tj2
RT @rspack_dev: Check out our first blog for Rslib! 🧙♀️ https://t.co/xDMNa550i7
What's the point of web components again? Why should anyone care about them?
Rslib is finally ready. Package building has been a major career painpoint- ive often found that gulp worked best. Most build tools are not designed for packages. For example css, everything wants to handle the side effects for you. https://t.co/9CjZkvk93Z
RT @Zackary_Chapple: The amount of work that went into this is huge! But a nice step for all the angular users out there.
RT @_jbroma: ✨ Re.Pack 5.1 is out! @repack_rn ✨ Here’s what’s new: ◆ Preload Module Federation 2 remotes ◆ Runtime hooks system ◆ Experime…
RT @_jbroma: More info on that coming soon! 🚀 We've made some exciting progress with Module Federation in Metro 🔧✨
Who's going to build an @rspack_dev based, federation-first, @reactjs meta framework? There are wonderful, ongoing efforts to add federation to existing projects, but, from my perspective, it is uncertain if the project authors share the vision and priority of federation-first™
🚀 Lynx 3.2 is here! This is our first stable release since going open source. It brings: ✅ ReactLynx Testing Library ✅ New features in CSS Grid, list, text ✅ Microtask, console, Rspeedy, DevTool, llms.txt A Thread: https://t.co/i7b9kUwRAj
RT @LynxJS_org: 🚀 Lynx 3.2 is here! This is our first stable release since going open source. It brings: ✅ ReactLynx Testing Library ✅ Ne…
RT @LynxJS_org: The website now supports llms.txt thanks to @rspack_dev's Rspress V2 and its new LLM plugin! Try https://t.co/Plo0MCpEGW h…
Ive released optimization flags for module federaiton. This If you use them and the external runtime together, remote entry will be about 14kb compressed. Without external runtime, and both optimization flags set remote will be about 25kb smaller uncompressed.
RT @rspack_dev: Sentry just merged a PR to use Rspack in dev: ⚡️ 6.7x faster HMR ⚡️ 3.3x faster production build ⚡️ 27% less memory usage…
RT @jait_chen: @getKonstantin @reactjs We are working with the React team and SWC team to support React Compiler based on the SWC wrapper f…
RT @LynxJS_org: Pro Tips: If you are vibe coding Lynx, add https://t.co/vUk4mjkIwo with @Docs in @cursor_ai and #Doc in @Trae_ai for much b…
RT @Timeless911: We need some opinions and suggestions from the community.
RT @getKonstantin: I gotta say I'm truly impressed by @rspack_dev performance on my new M4 Pro. My intel was starting our main app in like…
RT @rspack_dev: Make bundling great again 🚀
RT @RussellCanfield: New to module federation? Or maybe you want to learn how to use the new runtime system or plugins? Check out my cours…
RT @Huxpro: It’s 1AM after a holiday night, but I just realized both @vuejs and @sveltejs creators have now acknowledged the community’s @L…
RT @jait_chen: Rslib is evolving into a comprehensive library dev tool. ✨ Out-of-the-box supports for: - JavaScript libraries - TypeScrip…
RT @jait_chen: We have new domains! Rstack's little crabs are more united
RT @GithubProjects: Launch fast. Render smooth. Stay native. https://t.co/xqSvdSxLAq
i still don't get why people say barrel files are bad but then are OK with libraries doing a giant barrel file as their main export
Bold move, but I am curious (and some honest first thoughts): - Will there be bundling for production at all? Unbundled production deployment simply does *not* scale in practice. - Will there be HMR during dev? Or is it going to be a full page reload + Node server restart on https://t.co/SKy7C3gqlA
Guess what's this ? https://t.co/uNzZMehNPg
Does anyone know what's going on? Rsbuild seems to have been nominated for the OS Awards this year, but we can't vote for Rsbuild https://t.co/wrUS9lfy6o https://t.co/WQ2uue6G4E
RT @hardfist_1: Little known is that @rspack_dev has good support for Preact, and we even ported the prefresh babel plugin to swc wasm plug…
RT @rspack_dev: Does anyone know what's going on? Rsbuild seems to have been nominated for the OS Awards this year, but we can't vote for R…
I finally got shared modules and client references for RSC working. Nexts flight plugins are tricky! If you shared a module with "use client" next would not report the client refercnes correctly becasure the module becomes ConsumeSharedModule and has a different signature.
RT @hardfist_1: If someone could build a JavaScript version of https://t.co/PfwAS6Y3JK, it seems perfect as a plugin language for rust-java…
RT @Huxpro: Vote for the little 🐈⬛ (And hey, I absolutely wouldn’t mind if your “Breakthrough of the Year” goes to React Compiler too 😉.)
Please dont forget to vote for @LynxJS_org for "The most exciting use of technology" award! It would mean alot to the team! https://t.co/rSRwo9sBPC https://t.co/lz7tTuygNU
RT @zoolsher: Vote for Lynx! https://t.co/uaDWf6VL33
RT @rspack_dev: Vote for Lynx! 🐈⬛ https://t.co/vFQ4xRJhAH
What if web dev tooling were optimized for browsers instead of build pipelines? What would that look like?
Ive figured out how to manually control HMR. Ive been able to hot update a application without a dev server :D Lots of cool potential here.
What source code do you think is beautiful. Like, you go to their github repo and look at it and think "dang ... maybe one day I can write code like this" For me it's - MooTools - Rails - Preact
I’ve been saying this ever since the first unbundled dev servers started getting popular. You hit a point where it doesn’t scale anymore. Interesting to see the pendulum swinging back towards bundling in development. https://t.co/ljLRA6hStH
For anyone interested in how im manually controlling HMR to send my own hot updates to cold builds. Ive pushed a branch with some ideas i had around HMR server side for federation, ranging from basic vm encapsulation to using the real hmr runtime mechanics of bundlers outside of
Going to wait till in on the plane and go to bed early tonight. But I really want to put together a example of production HMR based updates. I think theres a ton of potential here. For servers, im able to use hmr api to "restart" the app without needing to kill the process. In
if you wonder how bloated Postman app is: It calls 18 requests to 18 different servers at launch 🤯 @getpostman can you explain it? https://t.co/4oiN0sH6vX
Even I’m a bundler guy, I still think Remix’s approach is valuable in certain scenarios: * Integration with other ecosystems like PHP, Ruby, C++, Rust, etc. I had a tough time introducing a full frontend toolchain into the Lynx repo. Teaching developers from different https://t.co/2MzbTxKcz3
What is going on…? https://t.co/vzY9Ey7WgB
Library types can run, but they can't hide https://t.co/PvpKMOarhm
Lark has gotten even better. I booked flights, hotels, and airport pickups all within Lark. Best example of a super app
I Imagine how much the Remix team will hate this 🤣 https://t.co/ZP6FQKNKj7
People say “but what about tree shaking”?? But many npm libraries aren’t not as “tree shakable” as folks think. - lots of code is still CJS (*cough* React *cough*) which is opaque and makes it more difficult to tree shake - some modules have side effects, e.g. setting up event https://t.co/5Tv34VyAoA
Anyone want to make a React Router framework version of this? Maybe I'm wrong but that seems like it'd be simpler https://t.co/SmlCqtceE7
Why isn't react *DOM* built for the browser? 🤔 process doesn't exist in the browser https://t.co/7TyO69GVO6
jQuery had all those dollar signs… we should bring those back too
Play with @rolldown_rs in the browser via WASM with Module Federation support. https://t.co/VEayrQkIkU Is this the first instance of Module Federations built all in browser via WASM thanks to Rolldown!?
Unpopular opinion: The UI on ChatGPT Web is actually BETTER than on its native macOS app. The latter would honestly be better off as its Web app (with proper native integration like keyboard shortcut and windows management). https://t.co/gMIQWRS4ao
I've just learned that Apple decided to bump the next version of iOS to 26 and I can't wait for the ripple effect of tools and pipelines and code that will straight up break because they didn't it through 🫠
Ope, the problem was me. Import was supposed to be react-dom/client https://t.co/YpJR0htuQk
"What JS framework should I learn in 2025?" Here is the answer: https://t.co/pmaj9ZTj3f
🎉Exciting news! We've reached a major milestone by reducing our Time to First Byte (TTFB) from ~1.2s to an impressive ~60ms!🚀Platform migration from monolith to microfrontend architecture on full throttle now, powered by @rspack_dev and Module Federation.
RT @rspack_dev: Rspack 1.3.13 is out: Browserslist parsing is now in Rust — 5.5x faster and 2.1MB smaller install size (removed caniuse-l…
RT @_yanpes: 🎉Exciting news! We've reached a major milestone by reducing our Time to First Byte (TTFB) from ~1.2s to an impressive ~60ms!🚀P…
Dynamic treeshake macro works!! I now have everything I need to tree shake shared modules!! In Q3 it should be possible for module federation to produce artifacts that are just as optimized as monolithic builds.
I don't understand why people prefer Promise.allSettled() over Promise.all() How often does it happen that you want to render the page if any request failed? Like you render the page with only the footer but no header? 🤷♂️ Promise.all() by default, unless there's a good reason https://t.co/FLUJ09Rxa9
Another real world Rolldown-Vite performance win: @sanity_io studio’s build time went from: 21.9s to 5.1s (4.29x, w/o React Compiler) 49.6s to 20.0s (2.48x, w/ React Compiler) Unfortunately, React Compiler does add non-trivial overhead to build times in its current form, and
We really need to find a conference in Hangzhou / Shanghai / Beijing @ScriptedAlchemy. We could do a joint talk on Module Federation / Micro Frontends / Bundlers for browser, server and mobile. With all of the HMR and Tree Shaking stuff we're doing we could easily do 45-60 min
Shopify helped improve the HTML importmap spec Soon, all browsers should be able to declare and merge newly encountered importmaps Really curious to see how @remix_run v3 will leverage that 😏 https://t.co/oMAklAMAKk https://t.co/GTgIWq9gKL
RT @algolia: Thrilled to see @rspack_dev leveling up search with Algolia 💙☝️ Fuzzy search, history, and blazing-fast results — exactly wha…
Having to declare the whole dang import map up front, and only once, was always the thing I crashed into and then bailed on with any new app/framework ideas I had This is great! https://t.co/QGLE5lwxfF
iframes https://t.co/XdrMQaOtGR
Rslib now supports SSR / node based federation packages! https://t.co/bv1tqVq0xs
RT @rspack_dev: Rspress v2 now comes with a plugin for @algolia integration 🔍 We've already rolled it out on Rspack, Rsbuild, and other si…
RT @rspack_dev: You can now upgrade Storybook Rsbuild to v2 to use Storybook v9 🥳 https://t.co/Su0Uhx4N3V
RT @2hea1: Module Federation v0.15.0 is out: https://t.co/otVT3L17M7
BREAKING: Forget <iframe>, we’re going with <frameset> instead.
Anyone remember: - FTP’ing a PHP script to your server - put the URL in the address bar - and seeing your web page And if it didn’t look right you could just edit it and do it again?
After months of polishing, Rspack's incremental build is now stable ⚡️ It's already enabled by default in Rsbuild and and will also be enabled by default in @rspack/core v1.4. Here's the feedback we just received: https://t.co/d4OsXlQS3r
RT @rspack_dev: After months of polishing, Rspack's incremental build is now stable ⚡️ It's already enabled by default in Rsbuild and and…
Kinda wish I still had access to Lululemons codebase. All this fancy optimization stuff we are working on would have doubled the sites perf in a few hours. I spent months squeezing a few KB at a time out of it. I think about this every day
In the past 48 hours I've written 10k lines. I need a long holiday.
had an interesting chat today with @Zackary_Chapple about module federation. then i went and argued with grok about it for a couple hours. i'm intrigued about how this concept could work in the open source community. imagine this: you publish an npm package which has federation
Maybe I’m thick but I’m not getting a lot of the decisions behind remix 3. https://t.co/Si2OiErTMt
@nextjs does some really cool stuff for testing. It uses the HMR api to send patches with test scenarios into warmed instance, instead of bundling over and over and over. Respect earned. 🫡 https://t.co/UN9md5ULXH
Rstest has landed in the Rsbuild repo: ⚡️ 300+ tests finished in 2 seconds ⚡️ migration took less than 10 minutes Rstest is still early, but v0.0.1 is dropping soon 🔥 https://t.co/VkatGSb0Yr
RT @jait_chen: Rstest 🔥 https://t.co/65xqozB4Az
5 years later. I can support unit tests that rely on federation. Will investigate this soon as I can https://t.co/lidHVJkHiE
Landing in the Rslib repo in the near future. https://t.co/KfnMq2gyjJ
lol Apple so many WCAG contrast failures 🤦♂️
Please tell me there's a way to turn this translucent garbage off? https://t.co/JMsJYxBR3p
Like WHAT useful information does the blurred content behind the panels communicate? How does it make navigation clearer or simpler? This is a parlor trick that degrades usability for no clear gain. https://t.co/ZmOV6mRFjG
i like the new glass design from apple https://t.co/Y6uUTtPxqF
Im just thankful Steve Jobs isn't alive to see this.
Even the apple site has the new glass ui. https://t.co/LTTUJ6OgmD
Windows Vista is so back https://t.co/LClKLxBhnM
Who gonna tell them. https://t.co/LWJEs8Z8r8
Rolling out my new app ui today. My mom told me it looks really great. https://t.co/u6Uu8tYL6C
I saw this about 6 months ago in Pico R&D its very impressive. Basically spatial elements for web. When VR devices are using the site, it is like responsive web design but for XR. I was able to grab the YouTube player or comments section and put it somewhere else. https://t.co/o1GvtNTkn5
iOS 26 as well as iPadOS 26 is amazing! Looks great, especially the notifications https://t.co/c1fVUEs6Qr
This makes me feel comfortable https://t.co/pO34O1vlCu https://t.co/vR63rIN6RK
Deep down I know it'll get fixed and the final product will probably be pretty nice. But this isn't a great first impression. Prev versions had their issues too, but this seems extra noticeable when UI issues occur. https://t.co/fcRTQgzA9f
How does something like this get through so far as to be presented at a conference and rolled out in beta? https://t.co/a7wufdvCtK
We revamped the design of our dashboard at Supabase. https://t.co/hFUj49PKVQ
fire all the delusional toxic positivity retain all the realists who call garbage garbage rebuild to where the realists say its amazing a huge problem is that companies create culture where its unsafe and unpopular to be self critical of first-party products: if apple https://t.co/bFCPJzJKtO
SWC seems more intuitive than babel when it comes to writing transformers and working with ast.
https://t.co/Js38AQEdLR https://t.co/dlxRBhrxwH
Ackshually it's MORE efficient to order coffee over SSH It would take longer to use a website https://t.co/bDDhNQEo1e
"Invisible user interface" https://t.co/Whu5bpkNRK
Getting really tired of folks saying that they built something "from the ground up" when, in reality, the product was ~98% built by another team. If you're wrapping Chromium, you didn't "build the browser from the ground up". You created features in the chrome around Chromium. It
RT @rspack_dev: Rspack v1.4.0-beta.0 is out: ⚡️ 40% faster HMR (incremental build) ⭐️ Improved ESM output ⭐️ Improved virtual module suppo…
WASM! https://t.co/4GfAxUcxwA
RT @stolinski: This is the talk I’m most excited about. @Huxpro talking about @LynxJS_org Super exciting stuff #reactsummit https://t.co/n…
rendering html on the server is easy client side rendering is easy the moment you want to render on the server and then hydrate the same code on client you get an explosion in complexity problem after problem to solve
RT @kpmdev: think @rspack_dev's nextjs plugin just "fixed" our dev compile times. down to 2s compiles on most pages now. we used to see 5-1…
RT @rspack_dev: Over the past two months, next-rspack has become more stable and now passes 98% of the next build integration tests. After…
Im messing around with compression dictionaries and the results are crazy. When I redeploy a 4mb application, with all chunks re-hashed. The network transfer for cache busted assets is only 40kb, which inflated back to 4mb. As long as you've been to the site once before, any
Tech conf idea, personal branding by wheeler himself for one lucky winner https://t.co/o6tkBP30Dg
This is a massive breakthrough IMO. the external module issues of webpack esm has been a plague for tree shaking and esm/cjs interop. @Timeless911 you are a hero https://t.co/KvzLsQyN8C
RT @rspack_dev: Production bundle quality matters. In Rspack 1.4, we collaborated with SWC to improve dead code elimination. For example,…
Announcing use-mcp, a hook for making any React app an MCP client: https://t.co/R20HL60MsN It supports both SSE and Streamable HTTPs remote servers, handles the OAuth flow, should just drop in and give you a list of tools and a way to call them (for now, lots more to come).
@ScriptedAlchemy Hey! Can federated remotes be loaded in component tests? Playwright should support it since it runs in a browser-like env. Jest + React testing lib didn’t work due to Node.js runtime. My modules are on a CDN, I just want to load them in tests without mocking.
> They should call it something other than Remix Threemix it is then
I have almost everything required for cross app tree shaking. Just need to find where rspsck tracks the unused exports
@rspack_dev @ScriptedAlchemy why module federation remote app via vite and shell as rspack not working?
Can i haz tree shake? https://t.co/sIVqYAdHMH
It works! I can now tree shake shared modules!! https://t.co/R72avMSYLa
RT @LynxJS_org: We shared two demos with @tailwindcss and https://t.co/Eu2B80cVLy during @Huxpro's talk at @ReactSummit last week. ✨ We are…
RT @LynxJS_org: With real Tailwind and native CSS vars, we can adapt themes from generators like @daisyui_ , @shadcn/ui, or tweakcn by @iam…
RT @Fordream340029: Yes, you can run https://t.co/ENC3Pc9m0s in Lynx. Just needs a sprinkle of Web API shims: https://t.co/66k3g51jG5 http…
Remix_FINAL_FINAL_v3 https://t.co/gQgKyXN16Z
Thinking of jQuery UI plugins https://t.co/EqEtNg7YaS
RT @ReactSummit: 🏆Congrats to the React OSS Awards winners! Breakthrough of the Year: React Compiler (@reactjs) Most Exciting Tech: @LynxJ…
Wonder if it’s a bug still if you switch to @rspack_dev prod build. @theo have you tried it? https://t.co/I8SHaPn5KJ
the holy grail of TS library build tools doesn't exist imo 1) declarative tshy-style entrypoint map—see pic 2) 1:1 transpilation (no bundling) of all files reachable from entrypoints 3) generates .mjs/.cjs for unambiguous resolution 4) writes "exports" to package.json https://t.co/CWGLCyxnQs
Code optimization needs to be compression aware. Especially when you can use custom dictionaries. Optimizer should accept a corpus for the dictionary and try to use the same var names between builds. Probably more of a issue with user code then libraries and vendor
I know rstest is 0.0.2, and not feature complete. But man, i really love where its going. You can just run rstest and itll use your rspack or rsbuild configs. Its replaced my use of nodes builtin test runner for quick test scnearios.
how nexts server hmr works. https://t.co/7T6dnqHeka
RT @Moonface_deamon: <input> and <textarea> will be available soon! #Lynx https://t.co/fdJ07t5zvY
Rspack & Rsbuild 1.4 is out! 🎉 🍀 Running in the browser ⚡️ Smaller bundles ⚡️ Incremental build by default 🪚 New CssChunkingPlugin 🦀 Lots of Rstack progress https://t.co/169cT9GR7L
江湖多俊杰，不知诸位中，谁敢应此英雄帖？是福是祸，是机缘还是劫难，唯有亲历方知分晓。 https://t.co/kvNHGkuDvQ
RT @rspack_dev: Rspack & Rsbuild 1.4 is out! 🎉 🍀 Running in the browser ⚡️ Smaller bundles ⚡️ Incremental build by default 🪚 New CssChunki…
Wasm builds of rspack landed! https://t.co/Gjv1AaFC7C
RT @grabbou: Module Federation enables independently deployed, runtime-loaded code, critical for scalable microfrontend and super app archi…
Here we go ! MF v0.16.0 support rspress ! Much appreciate for @Soon_Iter help! https://t.co/Fb2gFu1C6p https://t.co/y0ogNUh0GH
RT @2hea1: Here we go ! MF v0.16.0 support rspress ! Much appreciate for @Soon_Iter help! https://t.co/Fb2gFu1C6p https://t.co/y0ogNUh0…
A follow up to this: we have shipped CJS treeshaking in Rolldown 1.0-beta.23! (currently behind an option) When both using swc as minifier, Rolldown & Rolldown-Vite now produce roughly the same min+gzipped size as rsbuild: https://t.co/RAdSb4GXKA oxc-minify (alpha status) still https://t.co/iYjD6wx6UB
Might just have module federation working server-side on Cloudflare workers!!
This is my last day at @vercel! I'll take some rest and focus on @swc_rs for a while.
Module Federation on @Cloudflare workers! I can even execute user-submitted code securely! Edge worker is deployed once and never again. All code updates are totally dynamic. https://t.co/AfojxxcT5T
RT @ScriptedAlchemy: Module Federation on @Cloudflare workers! I can even execute user-submitted code securely! Edge worker is deployed o…
RT @hardfist_1: Great write-up! In fact, scope hoisting is one of the biggest reasons Rspack can be slower than other Rust-based bundlers —…
RT @midscene_ai: 🎉 The Midscene 0.21 update is here! 🎨 The Chrome extension has been upgraded 🌐 Web 1️⃣ New Chrome extension interface…
RT @jait_chen: Small clarification: Rspack isn’t slower than other Rust-based bundlers. In fair apples-to-apples comparisons, it’s still in…
RT @rspack_dev: We wrote a blog sharing some findings from our binary size optimization journey - exploring efficient ways to embed large s…
RT @jait_chen: @youyuxi I also compared Rsbuild and Rolldown Vite in rolldown/benchmarks/app/10000, as they are designed to be equivalent b…
Will module federation also work with Turbopack? (At least once the app router PR is in) @ScriptedAlchemy
I have @LynxJS_org PrimJs engine running in cloudflare workers!
MCP Ui React Server Components 🤝 https://t.co/ZHux4bLrhz
turbopack, my ass https://t.co/FuFezMExuO
Tested the full pipeline for tree shaking of shared code, few bugs / missed modules i need to correct. But already its 47% smaller
are we sure opencode is made by the same guys who are selling coffee through ssh? I don't know if styling CLI apps with tailwind was that good of an idea https://t.co/St2Pbke3HA
!!!!!Final results of tree shaking shared code!!!! Package: lodash Result: 88.7% smaller It fuckin works.
RT @kdy1dev: And special thanks to @rspack_dev because they helped @swc_rs development a lot. They even dedicated their team members for th…
Announcing Node-API Support for React Native https://t.co/meyFIw8Y8S #reactnative #nodejs
Do I see that correct @ScriptedAlchemy, App Router module federation e2e tests pass 😱
A leisurely walk seems to be very important in programming. Wasm plugin backwards compatibility is finally solved to some extent
RT @nstlopez: Just achieved Federated MCP :) https://t.co/Msrx9uPnIu
Casually stumbles across a full implementation of WebContainers & Linux "WebKernel"
RT @kky_kongjiacong: Developing Modern.js 3.0.Rspack Only, bye Webpack. https://t.co/N8LRvfrggD
Would you take a 900ms upload over a 300ms upload if it meant your end users got assets in 9ms instead of 60ms?
While investigating @rspack_dev performance in Next.js, SyMind found a simple opportunity to add caching that speeds up webpack >6%! https://t.co/UZi6Z9ItQv An awesome contribution for those still stuck on webpack.
RT @rspack_dev: 🚀 Rspack v1.4.8 is out! We've implemented a Rust watcher to replace the original watchpack package, making HMR ~10% faster…
Should start to become more apparent that build speed is becoming a focus for rspack. Up till now, speed hasn't been a major focus. Build it right, build it well, then make it fast. https://t.co/HJGPVnFPrK
RT @_bgwoodruff: While investigating @rspack_dev performance in Next.js, SyMind found a simple opportunity to add caching that speeds up w…
RT @thymikee: Reusable themes in @rspack_dev Rspress v2 are such a life saver. With @grabbou we wanted to put Callstack brand on our OSS we…
RT @Zackary_Chapple: Would you take a 900ms upload over a 300ms upload if it meant your end users got assets in 9ms instead of 60ms?
Soon, http.createServer will be supported by @Cloudflare workers and you can use your fastify/express.js apps! https://t.co/PTj87qNuKH
Treeshake of shared code: - Original Size: 1,505,172 bytes (1.47 MB) - Optimized Size: 12,485 bytes (12.2 KB) - Bytes Saved: 1,492,687 bytes (1.46 MB) - Reduction: 99.17%
RT @ScriptedAlchemy: Treeshake of shared code: - Original Size: 1,505,172 bytes (1.47 MB) - Optimized Size: 12,485 bytes (12.2 KB) - B…
When the pnpm install is literally 80gb of node_modules for 1 repo 😆
RT @xinbao965: 新一期播客预热，本期播客我们邀请到了字节跳动 Web Infra 团队的 @2hea1 一起聊聊 #模块联邦2 #ModuleFederation 2025 年的当下，模块联邦发展怎么样了？之前用起来不舒服的地方现在好用了吗？heal 见证开源项…
This month, as a POC for the company I work for, I migrated a web app from Webpack to @rspack_dev keeping almost exactly the same configuration, and the performance gains are insane. 2m14s production builds to 10s 🤯 Can’t wait to make this migration in the company’s CLI
RT @fslzrr_: This month, as a POC for the company I work for, I migrated a web app from Webpack to @rspack_dev keeping almost exactly the s…
Write your own release scripts. You might think you're saving time using something like changesets or a fancy monorepo manager. You're not. All monorepo managers are leaky abstractions. At some point you'll pay the cost of adopting one.
RT @Zackary_Chapple: So a linux kernel, WebContainers (not from StackBlitz), and Monaco IDE running in the browser. Full development envi…
WebKernel, an operating system for the browser. https://t.co/DuiMp3Rpm2
What kind of sorcery is this? The entire Linux kernel running in the browser! Not just parts of it, the entire operating system! https://t.co/GHDAfANvSj
RT @rspack_dev: In the next release, Rspack will be able to emit errors for duplicate ESM exports: https://t.co/si5XxfBWCU
Welp, Remix 3 is feeling kind of fun to use
everyone using rollup with @NxDevTools , go set this ASAP. "useLegacyTypescriptPlugin": false. Ive debated abandoning NX for a long time over build issues, this fixed them.
using `let` in module scope as state management https://t.co/fqdLMQdXAK
The Stylus package was removed by the npm security team 😮 https://t.co/Lu7U89TI75
In effort to support @rolldown_rs and @MetroBundler - we will be producing an official specification and architecture design documents for module federation 😘
claude telling me "its production ready" is the same thing a Jim Cramer telling me to buy or sell a stock.
anyone know what happened? https://t.co/MvGIssMBUm
RT @hd_nvim: Miso Lynx 🍜 🐈 is a mobile framework that uses miso and LynxJS to facilitate drawing to native iOS UIView, Android View, and fo…
> Alright, Ryan believes in me and wants me to get all the tests passing before he goes to bed! 5/19 passing is a good start, but I need to really understand what's going wrong with the other tests. Let me think through this systematically
RT @chenleidev: I am the current maintainer of Stylus. The Stylus library has been flagged as malicious lib, which has caused many lib and…
Anyone know someone at npm? https://t.co/laxBzuEZkj
Another day, another webpack giant slayed to make way for @rspack_dev's RSBuild https://t.co/R2uWPTk8Nd
👀 @angular is adopting @rolldown_rs for their experimental chunk optimization feature. Super curious to see the build speed improvements over Rollup and what else is in the cards for Rolldown & Angular! https://t.co/wF4xc5SLVE
Who's coming to @reactuniverse_ in September? 🙋 I'm working on the slides and can't wait to see you all! https://t.co/HN2DIpnO8Q
RT @pviral12: @ScriptedAlchemy @rolldown_rs @MetroBundler Federation is the sword that will execute iframes away from this world 😈
It’s only been 6 years since SwiftUI is released. Funny. https://t.co/YiGqAB3EnP
I asked Chat GPT a recommended approach to refresh my React Web skillset. It name-dropped @kentcdodds in an unprovoked manner. Kent how much you paying OpenAI for this level of advertisement??? https://t.co/PkhkhImuTV
Im about to merge the ability to include and exclude share modules based on regex or version range. Its currently undocumented as the docs are stuck in the Next.js RSC branch and i did a major overhaul on doc information. This is very usefil for prefix sharing. shared: {
My daily 500GB disk full from compiling rspack 🙃 https://t.co/zNeiQTVFZN
How exactly do you expect an email app to load emails? https://t.co/Dqyl1Hfy6Y
I helped create UMD to consume a JS module from anywhere (AMD, CJS, globals) with only one file (before es modules) and instead of using it for all three formats people treated like its own new format and added it to the pile so dumb https://t.co/mStLA4HXaJ
RT @rspack_dev: Rspack v1.4.10 is here: ✨ 30% smaller Wasm bundle (38.9MB → 27.2MB) ⚡ 4x faster tree shaking analysis in large projects ht…
that's the whole lift-html/core btw https://t.co/QhDEMJU5DA
Most Rstack packages are already pure ESM packages or dual packages, but @rspack/core is currently CJS, which helps with webpack project migration. We plan to change this in Rspack 2.0. Which approach do you prefer?
Good candidates always have options. Even in a down market. Case in point: I’m working on a Front End Developer role at a large bank with a colleague. Mid-Sr level. Onsite 3x/week. NYC metro area. We submit a strong candidate. She interviews once. Feedback is great. Then the
This exchange cracks me up https://t.co/O0Hx6uC6pN
> but you can pass mutable objects and the kids won't understand! you're saying "tracking contexts" and all their bespoke semantics are easier to learn than a simple language feature like pass by reference vs pass by value? https://t.co/Hi6CL7boS7
Okay, this but Dr.Strange style. Alternate realities colliding. We call it the MultiVercel of Madness. In one of them @thdxr is CEO, Im all in on iframes, and @mjackson shoves the most complex bundler magic into everything possible, Ryan uses const https://t.co/vdMF8HhxBL
Coze Officially Embrace Open Source https://t.co/hjp6xa82qk https://t.co/583qZ5HQuu #coze #OpenSource https://t.co/AjxJYjVwIi
Remember using style-loader in dev for HMR? Im the one who backmerged extract-css-chunks-webpack-plugin into mini-css and give it HMR https://t.co/1WUukbXKie
One of my pet peeves in microfrontends is the management of port numbers for every app. I want the apps to just startup with any available port and somehow register themselves with each other. Something like a local DNS. @ScriptedAlchemy do you have solutions for this?
RT @rspack_dev: @rspack/browser is cooking 🍳🧑🍳 https://t.co/K72qLHLs9k
The power of module federation https://t.co/RGHsHPBAwc
RT @2hea1: Module Federation v0.17.0 is out: https://t.co/yFpRUByI1R
RT @nachocodoner: In Meteor 3.4 with Rspack, it's not just build times that improve. Client bundles get slimmer too. Early tests show arou…
I hope I have skill issues The alternative is often much worse: something outside my control.
RT @rspack_dev: Rspack v1.4.11 is out: ⚡️ 12 perf improvements 🐞 14 bug fixes We're starting work on v1.5.0 — stay tuned 👀
RT @Huxpro: @tannerlinsley @tan_stack @LynxJS_org Hey Tanner, https://t.co/GTYbOli5tt
RT @BitterGourd1020: 🚀 ReactLynx now supports TanStack Router's file-based routing! Modern, type-safe routing meets cross-platform develop…
seems like if we went all in on <canvas> rendering the web could be so much better, it's a clean start from html/dom without requiring a whole new platform and it's not fundamentally impossible to make canvas accessible if everybody is on board with it
RT @rspack_dev: Rspack v1.5 will introduce an experimental optimization that skips building unused reexports in side-effect-free barrel fil…
I’m surprised that all bundlers didn’t already do this. Parcel has skipped parsing unused re-exports for 6+ years. I just assumed everyone did that! https://t.co/ZcfNyyKmAk
With anthropic ending my gravy train in a month, I decided to get 100% code coverage achieved over the most complex parts of module federation. Its sharing system. The recent additions have doubled the mind-boggling complexity of it. Ive been getting increasingly nervous every https://t.co/4nQKZsXO4I
a girl can dream! https://t.co/w9Jp9ZNGmI
How exactly do you expect this to be implemented? https://t.co/4dyF0sMG3A
Man, module federation node runtime is going to be the end of me eventually
RT @c_punisher: This is just a start and there are still many things to do to make it work better. Since Rspack/Webpack is very complex, I'…
RT @rspack_dev: We're building an Rspack REPL powered by @rspack/browser. Try out Rspack builds in your browser and easily share results wi…
RT @rspack_dev: A new RFC for Rspack's barrel file optimization. Take a look 👀 https://t.co/rhbFeJvMRG
Hi @ScriptedAlchemy, can you please help find links to this? The QR Redirect’s to a 404 on the mf io website https://t.co/8A0LuydhIe
theo right now https://t.co/0AfCTylfcc
if this man wasn't a national treasure i would take away his syntax highlighting for a month so he can consider what he has done https://t.co/JZ8pi76l3k
My first actual code contributions to @rspack_dev just got merged 🎉 Parallel loaders is an exciting feature, especially for Re.Pack where we mix babel & swc Thanks to Web Infra team for quick handling of these PRs 🙏 https://t.co/S2DK32N5KB
If you're a startup, your job is to punt the need for a microfrontend architecture for as long as possible. But eventually the day will come. And @vercel will be here for you 😁 We've been dogfooding this tech for a while, co-designing it with some great enterprises, and we're https://t.co/UrRWD6FsnR
You can just ship your org chart! Not even a joke: when organizations grow beyond a certain size, you want to ship to the same app without coordinating with "that other team". In the frontend space, the technology to enable that is called microfrontends, and we just shipped https://t.co/vsCKF6GW7Y
RT @Soon_Iter: Rspress 2.0.0-beta.25 supports UI display in plugin-llms, which is inspired from fumadocs and other docsites in the communit…
Tea app needed this slash command real bad. https://t.co/FqziEwGlMg
I've heard you all like Module Federation 😉 https://t.co/8XSzzyjbSI https://t.co/6DrRRXLz2x
Has anybody used @withgraphite with module federation based applications? I feel like theres something here, but need to wrap my head around it. If anyone has, please DM - id love to hear about your workflow.
It’s unbelievably silky-smooth, I can't go back https://t.co/rZxiXHTozf
rslint https://t.co/2W9hPUCgMz
RT @rspack_dev: We're exploring better tree shaking strategies for future Rspack versions. Our team member @ahabhgk wrote a blog post compa…
Ugh even the userland tooling that makes modern Node usable is now becoming infected with modern Node crap https://t.co/h95uHR0D5p
Haha remember that wild two weeks some DevRel from apple started posting here about how committed Apple was to improving iOS Safari? Fun times. https://t.co/WvQgaR5Ibd
RT @rspack_dev: Rspack v1.5.0-beta.0 is out! ⚡️ Faster native watcher ⚡️ Barrel file optimization 🌿 Virtual module plugin 🌿 Custom Rspack…
We don’t talk about this enough but @rolldown_rs has been quietly and continuously shipping various bundle size improvements (CJS treeshaking, gzip-friendly codegen, cross-module dead code elimination…) Speed and quality, pick two https://t.co/e4VxVJ1Bxr
RT @jait_chen: Rspack/Rsbuild v1.5 Beta is here with faster builds 🚀Especially noticeable when building larger applications. https://t.co/1…
Just got an angry email from a European user who canceled their T3 Chat sub over my stance on the UK “Online Safety Bill”
I have a crazy idea and I know all the bundler folks are going to hate me but I have to do it 🙈
RT @jait_chen: Super excited for the upcoming Rsbuild 1.5. We're enabling several Rspack experimental features by default: ⚡ Barrel file o…
RT @jait_chen: Rslint logo is a little crab holding a ruler, because linter is all about rules 📏 https://t.co/bFk2RJ6R1s
Not mine but saw this today. https://t.co/UUkyUUsyHL
need something like figma, but for designing terminal interfaces right now I do it all in neovim, but I would like to be able to select text and change the color, boldness, etc. a neovim plug-in could probably work well
This is my website. Guess the stack! https://t.co/rjkH6gvzT8
Open Source is truly something. Just picture this: At @webpack we're considering switching Acorn with oxc for AST (estree) parsing, which is the same tool that powers @vite_js. The OG web bundler can benefit from the work of great people out there, that's full circle.
RT @TheBlinks182: Yesterday, I presented to the team a branch of the transition of a 12-year-old project from webpack to rsbuild. There is…
RT @rspack_dev: Rspack 1.5 will support the "using" declaration. This enables the explicit resource management feature in ECMAScript, maki…
i hate how the og image for github links is the author's profile picture you randomly get a giant picture of someone's face when you paste it somewhere way too intense
Y'all are seriously sleeping on microfrontends. Not for long-term maintenance but migrations for legacy projects. Know enough framework internals to write adapters to render one FW's components in another and you become impossibly valuable for a company looking to modernize.
RT @quanruzhuoxiu: 🎉 Midscenejs 0.26 更新来了！ 🚀 工具链全面接入 Rslib，大幅提高开发体验、降低贡献门槛 快来 star 仓库，我们即将突破 1 万 star ✨，快来一起冲刺吧！ https://t.co/epjlJLwLm0…
RT @rspack_dev: Congratulations to the MDN team on this amazing launch 🎉 We're thrilled to see MDN's new front end is now powered by Rspac…
Powered by @rspack_dev https://t.co/xfn8skcvNp
RT @SocketSecurity: 🧹 New linter alert: @rspack_dev introduces Rslint, a TypeScript-first linter in Go powered by typescript-go. Fast, typ…
👀👀👀 https://t.co/uLxoziGtRu
RT @LynxJS_org: Do you know ReactLynx is powered by @preactjs? We just land Preact DevTool in the Lynx DevTool! We are also working on ope…
Only reason terminal UIs are making a comeback is because we all botched high fidelity UIs on the web so fantastically that the pendulum is swinging all the way back
Oof. 67% would prefer to rarely or never use RSC. Eager to see if this changes when other frameworks have a stable implementation. https://t.co/wKMLo0iFB4
is there a playwright mcp for cursor to run live demos and click around and stuff?
Nobody let Dax see this we'll never hear the end of it https://t.co/iagcB9OHXg
No I am not open sourcing it Why would I work for free? https://t.co/lOu1qdDHhf
RT @LynxJS_org: Our docs search just got 10x better with @algolia 🔎 Huge thanks to the DocSearch program for supporting dev docs worldwide…
We’ve introduced Twoslash into the byethrow documentation! This documentation is built with Rspress v2 (by @rspack_dev). Now, you can check type information just by hovering, making the documentation more intuitive and practical for learning. https://t.co/ZDdMAKMVnw
Rspack v1.5.0 is out! 🎉 ⚡️ Barrel file optimization ⚡️ Faster file system watcher ⚡️ Smaller installation size 💪 Extending Rspack with Rust 💪 Type re-export analysis And many exciting new features... Read more in our blog: https://t.co/N8FRbWaCzD
RT @jait_chen: This is the BEST @rspack_dev release yet, loving it 🥰
RT @rspack_dev: Rspack v1.5.0 is out! 🎉 ⚡️ Barrel file optimization ⚡️ Faster file system watcher ⚡️ Smaller installation size 💪 Extending…
Module Federation Runtime is almost at 4m downloads a week. I wonder after @reactuniverse_ where I will be talking about MF + @MetroBundler and @reactnative how much of an uptick we will see https://t.co/1i7PB22vbQ
RT @Zackary_Chapple: Module Federation Runtime is almost at 4m downloads a week. I wonder after @reactuniverse_ where I will be talking abo…
RT @karibash: We’ve introduced Twoslash into the byethrow documentation! This documentation is built with Rspress v2 (by @rspack_dev). Now,…
Trying to make shared treeshake Finally, some progress🫣 antd with Button and Badge size from 226.0 kB to 19.5kB https://t.co/s5Zvfj9vSD
RT @Infoxicador: @BleedingDev @theo @ZephyrCloudIO @Cloudflare People are sleeping on Module Federation and the RStack. Amazing techology
Been testing the limits of Rspress today with Module Federation. Smooth ride so far, but I wonder how much I can "share" from each instance. For example, sharing the navigation of one to use inside another.
Awesome to see prominent React devs talk about how they won't use Solid due to popularity. That makes sense. Of course they wouldn't. But why even mention it? Solid is clearly not a threat? As long as they and other like-minded people continue to use React nothing changes. So? https://t.co/o32ExSSjdd
Yes, the key question is it need to meet up dynamic used demands. Because MF manifest can know full dependencies , so it can be bundled expected shared in deploy period or cloud build ? And i also want bundle unused-shared.js as fallback . Still researching... https://t.co/JpJcq396Fb
RT @LynxJS_org: At @ReactSummit in June, @huxpro gave our global debut talk on why we exist and the principles that make us familiar, yet d…
@ScriptedAlchemy any suggestion for getting React Router's unstable_createContext to work using Module Federation? I'm using patchRoutesOnNavigation and react-router dep is a singleton.
Is bundling at development time a good thing or not a good thing? What's the final verdict on this? Assume code splitting per page is set up at least. Genuine question, @youyuxi @ScriptedAlchemy
RT @rspack_dev: Rslint now has a playground🚀 Still early days, but contributions welcome!
RT @hardfist_1: The JS API design of tsgo is excellent — it allows me to easily analyze the AST and other information within the Rslint web…
RT @Soon_Iter: Super excited to introduce @rspress/plugin-twoslash in Rspress 2.0.0-beta.30 release! Now you can easily add rich type ann…
Thank Vercel's marketing department for making everyone believe this I'd say majority of devs now actually believe this which shows how succesful their marketing has been But it's wrong One server does ~100ms response (1/10th of a second) for majority of requests (as in most https://t.co/PYhMVDOc7V
RT @birch_js: Lynx for Web announced by ByteDance at @reactuniverse_! 👏 https://t.co/3gT2FAQvfN
RT @thymikee: Web @LynxJS_org announced on stage @reactuniverse_ https://t.co/IZrWRGPYll
We have officially merged @MetroBundler support for module federation. If you're at the @reactuniverse_ conference, you'll hear all about it from @ZephyrCloudIO - who were the ones who made it happen, sent us the PR, and funded the changes needed to metro itself.
Going on stage in ~3.5 hours to talk about this. Huge thanks to everyone involved and for @ScriptedAlchemy helping land and publish the releases in time for the talk https://t.co/Nk5akJ1NHw
RT @Zackary_Chapple: Going on stage in ~3.5 hours to talk about this. Huge thanks to everyone involved and for @ScriptedAlchemy helping lan…
RT @ScriptedAlchemy: We have officially merged @MetroBundler support for module federation. If you're at the @reactuniverse_ conference, y…
Me adding more hooks into opencode. How it started vs how it going. https://t.co/WOQf8GqEpT
RT @DelphineBugner: The Bytedance team lead by @huxpro exposing many aspects of Lynx JS 🐆 https://t.co/vqZ23snfZ2
Handling the problems of scale with Module Federation in RN by @Zackary_Chapple @reactuniverse_ Now also available in Metro and not just Re.Pack 🎉 https://t.co/HL18EwtITo
wtf even is the browser company ?
With @Huxpro 😄 https://t.co/bgqQXBTyPT
RT @_jbroma: Handling the problems of scale with Module Federation in RN by @Zackary_Chapple @reactuniverse_ Now also available in Metro…
Triangle Man is a decent front end designer, but he knows nothing about networks. He keeps having to make this point because he has built Vercel on top of AWS which charges him for bandwidth so he has to pass that cost onto his customers or he'll go bankrupt. Cloudflare, on the https://t.co/YIdIP9eqzV
Tough call. Does next feel lock-iny? Yes. Was it necessary - initially yes. Vercel has been able to create something easy to use and exceptionally capable, adjusting infra let them focus on making next good without getting bogged down in side quests. The team does seem genuinely https://t.co/MKCQvdzGyL
Just about lost my shit 40 min ago when I decided to check the npm name only to find some bastard took the package name 20min prior. Only to see that it was @jait_chen 🥰 https://t.co/n0JtP8yY1m
we're still making the opencode tui go through our server api even though it could just directly call the underlying functions this forces our team to use the api so it'll never fall behind or be an afterthought stuff like this is why i hate the "performance" discourse, they
RT @rspack_dev: In the next release, Rspack ships advanced tree-shaking for dynamic imports via member expression analysis. We've also con…
Not many people give the triangle man 🔼 (@rauchg) credit for the name “@vercel” Such a familiar easy to pronounce and memorable name. Would love to learn the story behind how it came into existence!
RT @_jbroma: Module Federation Metro Plugin working in Expo Go ✨ up next: @expo development builds 👀 https://t.co/pTYaG4vdz0
RT @_jbroma: update: dev builds work too 🚀
RT @thymikee: Expo Super Apps for everyone
RT @Timeless911: Rslib v0.13 introduces tsgo to make declaration generation faster! https://t.co/tqxCyAsgqa
Lost my sunglasses at @reactuniverse_ now I gotta wear my wife’s lol https://t.co/e9ziIvp1Vb
should we pin todos to the bottom? it takes up more space but wondering if it's useful enough https://t.co/Gua3bsr2dX
RT @rspack_dev: Over 2M weekly downloads. Thanks for packing with us! 🦀 https://t.co/OispqkvV7n
Bun for front-end builds was a grave mistake. Wiggum is moving me back to rstack.
RT @Huxpro: Just reviewed @LynxJS_org 3.4 release blog, and it is LOADED 😃 https://t.co/pHX3AK42Pq
Puts on figma. Easy money.
RT @rspack_dev: Rspack v1.5.3 is out! ✨ Advanced tree-shaking for dynamic imports 🌐 Module Federation support in @rspack/browser 🚀 Faster…
RT @rspack_dev: 27x faster HMR 🚀 https://t.co/XckmFa3Vyc
RT @nstlopez: @ZephyrCloudIO has landed on both RsPress and RsBuild docs sites :) https://t.co/EmjkXLPZqk https://t.co/WRCQ9uM2Ms
Bro single-handedly has been the backbone of federation on react native for years. I can count those with author level knowledge of federation plugin architecture on one hand. He's earned a seat a few times over. Now it's official https://t.co/uZMTp5PInm
Modernjs RSC spotted internally
11yo: When will Zephyr work with Next.js me: soon, @swalker326 said he would have it working in a few weeks 11yo: good because I hate having to push then wait for 45 seconds for my app to deploy on ▲ Sub-second releases have spoiled my child. 45 seconds feels like forever
RT @jait_chen: Tried out Rslib with tsgo… my package builds got 2.5x faster 🔥 Now I work with tsgo locally, but CI runs on TypeScript 5.x…
RT @ronos_r: We migrated our largest web monorepo from babel + webpack to @swc_rs + @rspack_dev and couple things stood out: 1. rspack tea…
RT @CongCongPan: @wey_gu @Soon_Iter @elliotchen100 在我的测试中，Rspack 仍然和 Turbopack 存在差距，原因是由于频繁的 Rust 和 JS 通信开销。 目前我正在尝试 Rust 化一部分逻辑，目前是 exter…
RT @rspack_dev: Rsdoctor install size reduced from 22.6 MB → 15.7 MB in v1.3 — ~30% smaller & faster installs 🚀 https://t.co/gy1y8vQWvb
RT @rspack_dev: Rspack support in Shakapacker. Build JS in Rails faster 🚀
RT @rspack_dev: The Rspack team is off for a week 🌴 See you soon — we’ll return refreshed and start shaping Rspack 2.0 🦀
new UI to tame the concurrency monster https://t.co/Z7PX7LVbLY
Adding support for resolve alias on loaders and global resolve alias for shared modules
Vercel cooked with fluid compute. Love ‘em or hate ‘em, they cooked. https://t.co/XeiaeHlp6q
iOS app built. Considering pushing it to App Store https://t.co/rB3XWaMweC
Over the course of the beta phase, @rolldown_rs got 33% faster while adding *more* features and now produces 34.5% smaller bundles. It's now faster than esbuild, faster than bun, 3x faster than rspack, with bundle size comparable (+/-0.006%) to the state of art (rspack + swc https://t.co/OQcqZW06QH
RT @rspack_dev: 🚀 Rspack 1.6.0-beta.0 is out! ✨ Improved ESM library output (experimental) ✨ Improved MF runtime plugins ⚡ Barrel optimiza…
RT @hardfist_1: Make your build 5x faster with just one line of change. Slower FS is the root of all evil 😂 https://t.co/L93iLny1lN https:/…
RT @zoolsher: Wait, did you just speed up Rspack by 500%?
RT @Trae_ai: Bun runs fast and pairs well with TRAE. You can default to Bun when starting a project. SOLO Builder will handle install and…
would you use something like this instead of YAML for github actions? https://t.co/LKpBV0Qj2q
What would you like to see us focus on in Rspack/Rsbuild 2.0?
RT @rspack_dev: What would you like to see us focus on in Rspack/Rsbuild 2.0?
RT @rspack_dev: Rspack 1.6.0-beta.1 just dropped: 🧩 support for "import defer" ⚡ builds up to 10% faster ⚡ CLI startup faster by 50–70 ms…
Working on ModernJS RSC & Federation support. Order of magnitude easier to solve and support thus far. Not fully baked yet, working through remote actions and exposing of remote actions etc.
Preserve modules is phenomenal https://t.co/MBLqWT7X0j
1.0 - make it work 2.0 - make it fast & modern Porting webpack was about reducing enterprise risk, getting 40,000 unit tests for free, knowing we could pump a few $100 billion in revenue through it without worrying. https://t.co/mVIup2pTt3
@ManfredSteyer it looks like @giorgio_boa and @ScriptedAlchemy have renamed their native federation package to be module-federation/vite. Any interest in the angular architects native federation package following suite? Or allowing remotes built with this to be pulled in?
My biggest struggle with Remix 3 so far is that it's not React. I cannot use React libraries there. Missing out on such a huge ecosystem feels wrong.
RT @Zackary_Chapple: Thanks to the help of @FredKSchott and the @astrodotbuild team we were able to get a @ZephyrCloudIO integration togeth…
@rspack peeps, is it possible to get it to dump the module graph to some parsable format during a build?
I realize it’s most likely the click / rage bait but I have to call absolute bullshit on @IgorMinar talking about web fragments actually scale where module federation doesn’t. The technology that is scaling to the largest enterprises across the world and downloaded almost 5m
RT @rspack_dev: 🧩 The Rspack roadmap has been updated — It now includes Rspack 2.0 and our current areas of focus. https://t.co/kSJfnYT7RC
RT @ManfredSteyer: Did you know? 🎉 Module Federation just turned 5 this month! It debuted on 2020-10-10 with Webpack 5 — a game-changer by…
Big shoutout to ByteDance, @rspack_dev and @ZephyrCloudIO - who have collectively invested millions into v2. In 5 years, I’ve seen federation gain adoption far beyond anything I’d ever imagined. I just wanted to import(header) without needing to reinstall and deploy. https://t.co/Vs8P9WVSUB
Treeshaking of shared code ready to merge for both webpack and rspack!! Massive undertaking
I’ve implemented the asynchronous startup capability into rspack. Still some failing tests to address, but once resolved - you will no longer need dynamic imports to prevent the “eager consumption” errors
RT @jait_chen: The next step is to kick off Rspack 2.0. 🦀 Some of its upcoming changes will be available early in Rspack 1.7 via feature fl…
Sharded my jest CI run that was taking 10 minutes into 2 jobs and now each one of them takes 10 minutes https://t.co/bHwSYGMwWL
In 5 years Vue will be far more popular than react
RT @rspack_dev: In the next release, Rspack's source map generation is 30% faster 🚀
RT @kdy1dev: Starting from v1.15.0, the Wasm plugins of @swc_rs are now backward-compatible! Read more at https://t.co/wG6PTXEh0w Thank yo…
Hi @ScriptedAlchemy. In your book on Practical guide to Module Fedration, I read it like you referred to seeing every independent component as a MFE. Is this the ideology that has driven to Module Federation. Asking to understand on how many levels of MFEs can be built. @jherr
In repack, we have fixed the splitChunks:all problems that module federation has had with sharing is used in conjunction with chunk splitting.
RT @rspack_dev: 10 million monthly downloads. Huge thanks to everyone trusting Rspack. ❤️ https://t.co/UciGXvwY1A
RT @jait_chen: Rspack has tons of complex tests. We just migrated from Jest to Rstest and saw about a 30% speed boost. ⚡️ If you’ve been t…
Got a new demo from @2hea1 regarding shared tree shake support. Our devtools will also render new visualizations showing the live model share and export usage.
How we solve export sharing / correctness. The federation runtime has a new resolving algorithm “export-first” - this means that, like “semver”, we now have “exportVer”. Export availability is more important than version range. https://t.co/fzbjOIGGcy
RT @Huxpro: Excited to drop our radix-ui / shadcn-ui alike components library with Zhou in London
Is Google planning to buy Vercel? Their recent Google Antigravity beginner tutorials use Next.js and TailwindCSS to build the demo app. https://t.co/stQZLbqIXc
RT @rspack_dev: Rspress v2 RC is live across all Rstack sites. 🎉 A few fixes are in progress. The final release will ship once everything…
RT @nektro: In the next version of Bun, Rsbuild and Rspack work thanks to N-API compatibility improvements https://t.co/OSwyAJvlKv
What I really love about a bundler…anything can be a module, or a dependency. A module is typically a file. But if there’s a parser, then the build phase of that module could produce additional modules. This is kind of how rsc works, directives create new entry dependencies as
I have rsc working with federation. Two variations. 1) http proxy (kind like vercels rsc federation) 2) full capabilities of rsc like nesting and server actions which involve tracking all the refs across apps etc. 3) wip: single server using “code pull” similar to node mf
RT @rspack_dev: In the next release, Rspack will generate a more compact module factory format that trims the wrapper function. Benchmark…
RT @Huxpro: If you just heard about lynx-ui from @ReactAdvanced , try our demo here yourself! QRCode in the comments!
Can we please appreciate how freaking fast Vite is? I've got a medium-sized app with a bunch of routes, components, styles, and it boots up instantly. Love this.
RT @Huxpro: Yea this Tinder like Swiper is absolutely my favorite. Kudo to @Fordream340029 to make the lynx-ui Swiper and this demo
RT @rspack_dev: Storybook CLI now auto-detects Rsbuild projects and installs deps like storybook-react-rsbuild for you. Huge shoutout to th…
RT @jait_chen: Working on Rspack and Rsbuild taught me a lot about webpack's internal design. webpack is complex, but that complexity enab…
RT @CongCongPan: 🚀 Next.js with Rspack now enables persistent caching by default! The result? Massive speed boosts on warm starts: • App…
RT @rspack_dev: Rspack v1.6.7 is out: ✨ Support for import.meta.resolve ✨ Support for import.meta.dirname ✨ Support for import.meta.filena…
RT @jait_chen: The SWC parser is now 65% faster than 6 months ago and 104% faster than a year ago🚀 Huge kudos to @kdy1dev and @c_punisher…
Facebook's Static Hermes is kind of incredible. It can compile JavaScript into C, which is then compiled and optimized into native machine code. I just compiled the Less.js source code into a C library. Then I called it from Rust as a native Parcel plugin (no Node). Wild. 👨🔬
If @bunjavascript supports module-federation in future, we can move all our remote applications to use bun fully. We are using latest rspack version and it fails in build step when we run with bun
RT @rspack_dev: ⏰ Rspack 1.7 is coming tomorrow. The last minor release before 2.0.