# ![@michael_chomsky Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1298023616253005826.png) @michael_chomsky Michael

Michael posts on X about ai, if you, vercel, claude code the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

### Engagements: [------] [#](/creator/twitter::1298023616253005826/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1298023616253005826/c:line/m:interactions.svg)

- [--] Week [---------] +228%
- [--] Month [---------] +608%
- [--] Months [---------] +112,549%
- [--] Year [---------] +3,110%

### Mentions: [--] [#](/creator/twitter::1298023616253005826/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1298023616253005826/c:line/m:posts_active.svg)

- [--] Month [--] -53%
- [--] Months [---] +1,295%
- [--] Year [---] +1,644%

### Followers: [------] [#](/creator/twitter::1298023616253005826/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1298023616253005826/c:line/m:followers.svg)

- [--] Week [------] +65%
- [--] Month [------] +134%
- [--] Months [------] +871%
- [--] Year [------] +1,084%

### CreatorRank: [-------] [#](/creator/twitter::1298023616253005826/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1298023616253005826/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  [finance](/list/finance)  [social networks](/list/social-networks)  [stocks](/list/stocks)  [countries](/list/countries)  [celebrities](/list/celebrities)  [vc firms](/list/vc-firms)  [musicians](/list/musicians)  [fashion brands](/list/fashion-brands)  [products](/list/products) 

**Social topic influence**
[ai](/topic/ai), [if you](/topic/if-you), [vercel](/topic/vercel), [claude code](/topic/claude-code) #1660, [twitter](/topic/twitter), [openclaw](/topic/openclaw) #369, [code](/topic/code), [company](/topic/company) #1753, [cloudflare](/topic/cloudflare), [llm](/topic/llm)

**Top assets mentioned**
[Cloudflare, Inc. (NET)](/topic/cloudflare) [Alphabet Inc Class A (GOOGL)](/topic/$googl)
### Top Social Posts
Top posts by engagements in the last [--] hours

"@swyx An AI engineer is anyone who claims they are an AI engineer. Usually this occurs once a uni student places at a hackathon or two. Since this is America the onus is on you to prove they are not in fact an AI engineer but rather a junior dev with langchain installed globally"  
[X Link](https://x.com/michael_chomsky/status/1720390084875424200)  2023-11-03T10:38Z [---] followers, [--] engagements


"Every long-term project I've left behind lacked Typescript. Managing sizable projects without typing isn't just challenging; it's taxing on your well-being. Occasionally I might opt for 'any' but inevitably adding types later reveals hidden errors. Use types. Sleep soundly"  
[X Link](https://x.com/michael_chomsky/status/1728593927031451727)  2023-11-26T01:58Z [---] followers, [---] engagements


"Placing a bounty of a diamond pickaxe to the first person to recreate the Voyager paper in Typescript"  
[X Link](https://x.com/michael_chomsky/status/1729424524876337414)  2023-11-28T08:58Z [---] followers, [--] engagements


"@sincethestudy @natfriedman I agree. Experimenting with this myself but with a $10 limit"  
[X Link](https://x.com/michael_chomsky/status/1785147085547028931)  2024-04-30T03:20Z [---] followers, [---] engagements


"@HanchungLee @jobergum @vespaengine It's usually best to show retrieved results first in some manner before generating. Otherwise the search feels slow. If course there are exceptions"  
[X Link](https://x.com/michael_chomsky/status/1814029622784430159)  2024-07-18T20:08Z [---] followers, [--] engagements


"@DhravyaShah @OpenAI @perplexity_ai @AnthropicAI @cursor_ai @vercel @v0 I pay for all of these + grok. Am I cooked"  
[X Link](https://x.com/michael_chomsky/status/1826062415240528159)  2024-08-21T01:03Z [---] followers, [----] engagements


"@JosephKChoi http://smry.ai http://smry.ai"  
[X Link](https://x.com/michael_chomsky/status/1827826598118306099)  2024-08-25T21:53Z [---] followers, [---] engagements


"@kristianf_ @honojs @vercel @OpenAI @CloudflareDev Or use http://smry.ai http://smry.ai"  
[X Link](https://x.com/michael_chomsky/status/1828190183071850960)  2024-08-26T21:58Z [---] followers, [---] engagements


"@DamiDina Mostly. The prompt helps a lot and it magically renders before code is done generating"  
[X Link](https://x.com/michael_chomsky/status/1829005061474959444)  2024-08-29T03:56Z [---] followers, [--] engagements


"@pashmerepat @levelsio I've done this with works pretty well https://replicate.com/philz1337x/clarity-upscaler https://replicate.com/philz1337x/clarity-upscaler"  
[X Link](https://x.com/michael_chomsky/status/1829636096994787705)  2024-08-30T21:43Z [---] followers, [--] engagements


"Reflection 70B just dropped beating GPT-4o and Claude Sonnet on key benchmarks while being much smaller. This is huge but the real story is how it happened: Matt Shumer isn't your typical AI researcher. He's a prompt engineer who's been in the trenches since early GPT-3 days building for millions of users. That experience led to a key insight: LLMs suck at admitting mistakes when reflecting. They'll make errors just to have something to correct. The solution Fine-tuning for genuine self-correction without new errors. So simple it's brilliant and the big labs missed it. The result is a model"  
[X Link](https://x.com/anyuser/status/1832147061338489304)  2024-09-06T20:01Z 10.6K followers, 93.3K engagements


"Last week I went viral for tweeting about Reflection 70B saying I was massively excited the training strategy (purportedly) used to train it. Of course that turned out to be a fiasco. But now OpenAI has released o1 using an extremely similar strategy But o1 is only the beginning. If you do not care about inference time/cost you can make LLMs massively better. Benchmark gains become easy: just either use a bunch of different LLMs and then score/summarize with a smarter model or just add even more latency/chain of thought. The point is you can scale inference compute vertically (like o1) or"  
[X Link](https://x.com/michael_chomsky/status/1834688077224452345)  2024-09-13T20:18Z [---] followers, [---] engagements


"Any VCs want to sponsor a Developer Advocate dinner in SF I'm bringing together devrels from OpenAI and my favorite AI companies"  
[X Link](https://x.com/michael_chomsky/status/1835424832667992444)  2024-09-15T21:05Z [---] followers, [---] engagements


"Postulate: P1: There is no such thing as value investing in venture capital. P2: The steep power law distribution of returns means you should invest in the best startups at whatever the price happens to be. Proof: [--]. Let Startup A and Startup B be two identical startups differing only in valuation: Startup A is valued at $10 million. Startup B is valued at $100 million. [--]. According to P1 and P2 the price is irrelevant and one should invest in Startup B regardless of its higher valuation because power law returns dominate. [--]. Both startups are identical in potential outcomes so their future"  
[X Link](https://x.com/michael_chomsky/status/1839837408143045044)  2024-09-28T01:19Z [---] followers, [---] engagements


"@skeptrune Are you telling me that Sentry is a clickhouse wrapper"  
[X Link](https://x.com/michael_chomsky/status/1860833525647462801)  2024-11-24T23:50Z [---] followers, [--] engagements


"- Be @shadcn - Make a modular component lib - Lib goes viral - LLMs learn to use it - Shadcn is now the default component lib for LLMs - Majority of code will be written by LLMs in coming years especially for new websites - Shadcn is spared"  
[X Link](https://x.com/anyuser/status/1868362207593160754)  2024-12-15T18:27Z 10.6K followers, 165.8K engagements


"Just found out OpenAI has a predictive text feature that makes editing code files (eg Cursor) 5x faster. Made a lil demo with bolt dot new to show how much faster it is:"  
[X Link](https://x.com/michael_chomsky/status/1869070990468374803)  2024-12-17T17:23Z [---] followers, [---] engagements


"If typescript didn't exist I would have quit webdev"  
[X Link](https://x.com/michael_chomsky/status/1870728468151046308)  2024-12-22T07:09Z [---] followers, [---] engagements


"twitterwrapped. exa .ai is brutal its not my fault every post about shadcn goes viral 😭😭😭 typescript stuff is accurate though"  
[X Link](https://x.com/michael_chomsky/status/1872386793414566075)  2024-12-26T20:59Z [---] followers, [---] engagements


"@voidwalker_com @JungleSilicon Whoa cool how'd you make the backgrounds"  
[X Link](https://x.com/michael_chomsky/status/1873238811309949365)  2024-12-29T05:25Z [---] followers, [--] engagements


"o1 pro just solved a task that would take a SENIOR engineer [--] days. I was dumbfounded I didn't expect it to solve it at all but thought why not and got working code"  
[X Link](https://x.com/michael_chomsky/status/1879332732758831604)  2025-01-15T01:00Z [---] followers, [---] engagements


"Use Vercel. Just be careful with API routes once you scale and go client side when you can. SSR is an optimization and doesn't always make sense especially when LLMs are involved. Use SSR for dashboards when you have fast db requests. If your requests take longer than 700ms probably don't do SSR and ask yourself why If you are using slow proxies def don't do SSR. Learned that the hard way. Don't use API routes either. Vercel free tier is goated if you use it right and I think Vercel is a benevolent company"  
[X Link](https://x.com/michael_chomsky/status/1925364034041413674)  2025-05-22T01:32Z [----] followers, [---] engagements


"i think the quality of most software built with vibe coding is probably waaaaay worse than most people think it is. like maybe it functions but it probably does many things way worse than necessary. something like coderabbit obviously helps but not enough very junior developers are vibe coding complex apps and i PROMISE they are dumpster fires in lots of ways. literally faster for me to code from scratch than fix opuss mistakes 30% of the time. i just did a big migration with opus and low key regret a lot of the things i did. ended up ok because of good instincts/understanding of my problems"  
[X Link](https://x.com/michael_chomsky/status/1929335127190224903)  2025-06-02T00:31Z [---] followers, [---] engagements


"i need an excuse to make a search index with 1B items. what dat should I search needs to be valuable"  
[X Link](https://x.com/michael_chomsky/status/1931160898254889034)  2025-06-07T01:26Z [---] followers, [---] engagements


"@defi_dua Lets see 30M on Linkedin next 🔥"  
[X Link](https://x.com/michael_chomsky/status/1933330205365711037)  2025-06-13T01:06Z [---] followers, [--] engagements


"I always wondered why OpenAI bought Windsurf for 3B. Turns out everybody knows the obvious reason. They bought it for the training data. Windsurf gets proposed and accepted code completions absolutely invaluable training data thats difficult to get without having an IDE. Its also a place to try out new coding and code completions models. Now it makes sense why Anthropic is no longer serving Windsurf. They have the best coding model but Windsurf would quickly be able to take those gains and funnel them to OpenAIs new models. Anthropics main advantage isnt that they are good at marketing to"  
[X Link](https://x.com/michael_chomsky/status/1933624259659862253)  2025-06-13T20:35Z [----] followers, [----] engagements


"Introducing brainstormcool 😎 I'm so tired of getting one response when I chat with an LLM. What about 500+ Everyone's using LLMs wrong for ideation. They're not magic answer machines - they're exploring possibility space. But they can't read your mind. The only way to find what you actually want is to generate hundreds of options and let your brain pick the winner. LMArena raised $100M to show you [--] options at a time for free. We show you 500+ responses because that's how search really works. You know the right answer when you see it not when you describe it. This only works because"  
[X Link](https://x.com/michael_chomsky/status/1935819548093268329)  2025-06-19T21:58Z [----] followers, [---] engagements


"my launch [--] hours ago failed so i'll try again in lowercase letters. try an AI tool that generates hundreds of responses at once and then ranks them. ok that's it thanks http://brainstorm.cool/123 http://brainstorm.cool http://brainstorm.cool/123 http://brainstorm.cool"  
[X Link](https://x.com/michael_chomsky/status/1935852938297328110)  2025-06-20T00:11Z [----] followers, [---] engagements


"gemini [---] flash: 6k pages per dollar for pdf markdown llama [---] 11b vision on @inference_net : 16k pages per dollar llama wins on price but gemini might be better quality"  
[X Link](https://x.com/michael_chomsky/status/1936640816749117578)  2025-06-22T04:22Z [----] followers, [---] engagements


"-be me -build something with lovable -looks good with relatively little prompting -move to cursor once UI is somewhat cooked -it's vite + react + a lot of tool decisions that are completely unnecessary for my simple proj -shoot that's not my stack but ok -needs an api -lovable can't do APIs without supabase BS -switching to app router is too much work -ok fine make server -now I can't deploy to vercel and need to deploy on hetzner (too much work I have zero users) -next time will start in cursor I think"  
[X Link](https://x.com/michael_chomsky/status/1936730774881083564)  2025-06-22T10:19Z [----] followers, [---] engagements


"I have developed a the perfect stack for AI projects. I call it NUVubbtVHVPS. Next.js + Upstash (for weird route handler limit workarounds/rate limiting) + Vercel until a big bill hits then migrate Vite + Hono + VPS/Railway. The reasoning is that building in Next.js and migrating to Vite is literally easier for me than building in Vite originally for small projects. Initial velocity is just so fast on Next.js for me and LLMs and I both understand all the edge cases super well. @joshtriedcoding is JStack ready yet bc this acronym is too long"  
[X Link](https://x.com/michael_chomsky/status/1938354800187642169)  2025-06-26T21:52Z [----] followers, [---] engagements


"@justinstorre lmao congrats but the y axis got me"  
[X Link](https://x.com/michael_chomsky/status/1938372038668025923)  2025-06-26T23:01Z [----] followers, [--] engagements


"@gustofied this can work with vector dbs too lol"  
[X Link](https://x.com/michael_chomsky/status/1939070478725062784)  2025-06-28T21:16Z [----] followers, [--] engagements


"windsurf famously has unmatched gtm especially with big co's and banks. probably one of the biggest reasons devin bought them looks like devin might be a 10B company in [--] years at this rate This was supposed to be our big announcement today. Excited to be working with @Citi https://t.co/UAELQEO4fM This was supposed to be our big announcement today. Excited to be working with @Citi https://t.co/UAELQEO4fM"  
[X Link](https://x.com/michael_chomsky/status/1944885264838746341)  2025-07-14T22:22Z [----] followers, [----] engagements


"Introducing LookalikeCeleb: an open source app that tells you what celebrities you look like For example this pug looks like Sylvester Stallone. Personally I can see the resemblance. LLM inference powered by @inference_net"  
[X Link](https://x.com/michael_chomsky/status/1947456652250386573)  2025-07-22T00:40Z [----] followers, [---] engagements


"@inference_net Check it out here: https://www.lookalikeceleb.com https://www.lookalikeceleb.com"  
[X Link](https://x.com/michael_chomsky/status/1947457225427194074)  2025-07-22T00:42Z [----] followers, [--] engagements


"@b_dammylolah @tan_stack if there is tiktok for code this would be on it"  
[X Link](https://x.com/michael_chomsky/status/1949296787635970282)  2025-07-27T02:32Z [----] followers, [---] engagements


"Built an open source calculator that shows the true energy cost + carbon footprint of running LLMs. Couldn't find one anywhere so I made it myself"  
[X Link](https://x.com/michael_chomsky/status/1950269621761609877)  2025-07-29T18:58Z [----] followers, [---] engagements


"deployed to http://energy.inference.net http://energy.inference.net"  
[X Link](https://x.com/michael_chomsky/status/1950269671934201873)  2025-07-29T18:58Z [----] followers, [--] engagements


"tried an AI IDE today that actually works and the experience was magical. better than everything else by a factor of [--]. not cursor windsurf or anything you know. [---] users. dont dm me for an invite im gonna gatekeep this one"  
[X Link](https://x.com/michael_chomsky/status/1951158904064450889)  2025-08-01T05:51Z [----] followers, [---] engagements


"TIL bun run - file.js pipes the file through stdin and Bun auto-parses it as TS/JSX great for when your files are js files but have typescript in them for some reason lmao"  
[X Link](https://x.com/michael_chomsky/status/1953552153144177061)  2025-08-07T20:21Z [----] followers, [--] engagements


"@digitalconner https://inference.net/blog/what-s-left-is-distillation https://inference.net/blog/what-s-left-is-distillation"  
[X Link](https://x.com/michael_chomsky/status/1953963101172511100)  2025-08-08T23:34Z [----] followers, [--] engagements


"@zaidmukaddam i need to be in this slack channel asap lmao"  
[X Link](https://x.com/michael_chomsky/status/1957871471222747368)  2025-08-19T18:25Z [----] followers, [---] engagements


"@signtheverse @inference_net bro i wrote this myself. this is human slop"  
[X Link](https://x.com/michael_chomsky/status/1957872988852027521)  2025-08-19T18:31Z [----] followers, [--] engagements


"@signtheverse @inference_net lmaooo"  
[X Link](https://x.com/michael_chomsky/status/1957873681411211319)  2025-08-19T18:33Z [----] followers, [--] engagements


"New evidence from a research paper (with a hilarious title) that I did not read says that threatening or offering to tip a model does NOT improve performance. I have told an LLM that my job depended on a task being done well dozens of times and this suggests it could be a waste. Or not. I didn't read the paper"  
[X Link](https://x.com/michael_chomsky/status/1957892761027530983)  2025-08-19T19:49Z [----] followers, [---] engagements


"@inference_net ok so this model accurately identified this dog that kinda looks like a cat as a puppy which gemma 27B could not do no matter how many times i tried"  
[X Link](https://x.com/michael_chomsky/status/1958279254950776998)  2025-08-20T21:25Z [----] followers, [---] engagements


"@atbeme @inference_net yeah big labs nerf models or run quantized versions all the time without telling users. kind of unethical imo"  
[X Link](https://x.com/michael_chomsky/status/1958325417112424484)  2025-08-21T00:28Z [----] followers, [--] engagements


"@0xdrej @inference_net lmaooo a 1B model would likely think so"  
[X Link](https://x.com/michael_chomsky/status/1958330228063011068)  2025-08-21T00:48Z [----] followers, [--] engagements


"I didn't expect this while I was building it but a tool where you can paste an image already in your clipboard and get a caption is insanely useful. http://cliptagger.inference.net http://cliptagger.inference.net"  
[X Link](https://x.com/michael_chomsky/status/1958346305245479054)  2025-08-21T01:51Z [----] followers, [---] engagements


"technical blogs have such high leverage. you mean you can write about the terrible headaches you had engineering something and then some engineer will read it empathize with you deeply and feel gratitude for not having to deal with said headaches and then advocate for your product for the rest of their life"  
[X Link](https://x.com/michael_chomsky/status/1958693776572260586)  2025-08-22T00:52Z [----] followers, [---] engagements


"@skeptrune @thekeystatic convince my team to migrate please rn rn rn"  
[X Link](https://x.com/michael_chomsky/status/1958757812957585822)  2025-08-22T05:07Z [----] followers, [--] engagements


"@skeptrune @thekeystatic i want a web ui as well tho"  
[X Link](https://x.com/michael_chomsky/status/1958758044277543418)  2025-08-22T05:08Z [----] followers, [--] engagements


"Reducto has the best merch out of any startup in SF. @inference_net will be taking notes and catching up limited Reducto merch drop: boxy-fit shirts 220gsm 100% shrink-free cotton. made and garment dyed in the US and screenprinted here in sf. https://t.co/SRqzHC4sJL limited Reducto merch drop: boxy-fit shirts 220gsm 100% shrink-free cotton. made and garment dyed in the US and screenprinted here in sf. https://t.co/SRqzHC4sJL"  
[X Link](https://x.com/michael_chomsky/status/1958769403753111964)  2025-08-22T05:53Z [----] followers, [----] engagements


"yes. pretty sure they charge per seat. which means they just need to sign the right enterprises. Windsurf was able to close a lot of large enterprises (massive orgs like banks with thousands or tens of thousands of developers) if Cline can do the same their success is guaranteed. BYOK is just something that makes their product better an implementation detail"  
[X Link](https://x.com/michael_chomsky/status/1959455339763618166)  2025-08-24T03:18Z [----] followers, [--] engagements


"I found a Craigslist furniture post that's clearly a coded drug ad located right in the Outer Richmond. The item is a Dogon Granary Ladder Blues Sculpture. Dogon ladders are real West African objects used to reach granaries and rooftops; collectors buy them for dcor. But here's where it gets interesting this post is obviously for drugs like coke meth and opioids. If you look closer there are lots of signs in the language: One could oil the wood but we will coke option to the new owners. The coke option here is literal--you can literally buy coke. Frame hums crystal tiny imperfections yet"  
[X Link](https://x.com/michael_chomsky/status/1960766588933824918)  2025-08-27T18:09Z [----] followers, [---] engagements


"@hoopcutter bro. the language was interesting and reddit has already roasted me dw 😭"  
[X Link](https://x.com/michael_chomsky/status/1960969291982889187)  2025-08-28T07:34Z [----] followers, [--] engagements


"exactly zero people saw this coming. Cloudflare is killing it in the AI space. -Agent auth -LLM inference at the edge with their custom serving engine faster than vLLM -website/pdf to markdown Today we're announcing an unlikely partnership. We believe that agents need reliable responsible web access. That's why we're partnering with Cloudflare in support of Web Bot Auth and Signed Agents a new standard to allow good bots to authenticate themselves. Details 👇 https://t.co/i4iiKX37qw Today we're announcing an unlikely partnership. We believe that agents need reliable responsible web access."  
[X Link](https://x.com/michael_chomsky/status/1961101868639326442)  2025-08-28T16:21Z [----] followers, [---] engagements


"apparently if you arent building a saas with nano banana rn youre ngmi and will die alone"  
[X Link](https://x.com/michael_chomsky/status/1961295667835646074)  2025-08-29T05:11Z [----] followers, [---] engagements


"@reggitales looks like someone needs nano banana"  
[X Link](https://x.com/michael_chomsky/status/1961296049626386933)  2025-08-29T05:13Z [----] followers, [---] engagements


"@ammaar @artificialguybr @OfficialLoganK i found my api key in under a minute. this was a world record for me"  
[X Link](https://x.com/michael_chomsky/status/1961595214206046517)  2025-08-30T01:01Z [----] followers, [--] engagements


"Our next Commander and Chief should undoubtably be @TheRock and our Director of the CIA should be John Cena. Dwayne is contractually physically unable to lose a war and with a completely invisible CIA head every covert operation would be a success. All threats to our country would be eliminated within the first year and we would be able to safely reduce our military budget by $700B"  
[X Link](https://x.com/michael_chomsky/status/1961622455413608523)  2025-08-30T02:50Z [----] followers, [---] engagements


"I'm thinking about starting a personal blog. is there a platform that handles using JSX/markdown as my content acts like a super fast CMS and also handles emails"  
[X Link](https://x.com/michael_chomsky/status/1962610878668513312)  2025-09-01T20:17Z [----] followers, [---] engagements


"@skeptrune I'm so confused. Astro is a javascript framework not a CMS"  
[X Link](https://x.com/michael_chomsky/status/1962611806519861369)  2025-09-01T20:21Z [----] followers, [--] engagements


"so sorry about the most recent xai leak. i told Codex to center a div and it ran this instead 🥀 git clone git@github.com:elonmusk/xai.git cd xai git remote set-url origin git@github.com:elonmusk/xai-share-openai.git git push origin --all git push origin --tags"  
[X Link](https://x.com/michael_chomsky/status/1962687948748673437)  2025-09-02T01:24Z [----] followers, [---] engagements


"@nikunj I love this Will be sending you swag http://Inference.net http://Inference.net"  
[X Link](https://x.com/michael_chomsky/status/1962691926555271593)  2025-09-02T01:39Z [----] followers, [--] engagements


"hire @LueStudio to design you landing page go into debt if you have to (not sponsored. I wish I had a b2b saas so I could hire them)"  
[X Link](https://x.com/michael_chomsky/status/1962721717148905594)  2025-09-02T03:38Z [----] followers, [---] engagements


"Apparently big labs would pay millions for a dataset of images of UI converted to pixel perfect tailwind"  
[X Link](https://x.com/michael_chomsky/status/1963114897694249461)  2025-09-03T05:40Z [----] followers, [---] engagements


"The list of cos currently hiring devrels is insane: @ExaAILabs (just raised at 800M) @AnthropicAI (pays up to 400k + equity) @Meta (up to 380k + equity) @nebiusai (up to 345k + equity) @p0 @Snapchat @LiquidAI_ (working on cool SLMs) @infisical @Railway @Apple @e2b @TigerDatabase @Sourcegraph (Amp) @pydantic @mintlify @ClerkDev @mozilla @SignozHQ @DescriptApp And many more Introducing https://t.co/9z48Labvtc. I get a ton of inbound from companies trying to hire a devrel but it's hard to find the rare candidate with both a passion for educating developers and the skill to reach them. The"  
[X Link](https://x.com/michael_chomsky/status/1963348139995533546)  2025-09-03T21:07Z [----] followers, [---] engagements


"I'm super bullish on Exa. Not only did I work there briefly in mid-2023 (lore drop lol) but I use their search engine for 20% of my web searches. They don't really feature it bc it's so b2b but it's a valuable part of my workflow. It's here please try it: Just play with it and it will change your understanding of web search. I use Websets weekly. My hobby is creating GTM lead lists and Websets is currently the best tool in the world for that. Although Websets is currently used mostly for GTM it's a tool that 30-90% of humans on earth would use regularly if taught to use properly and made"  
[X Link](https://x.com/anyuser/status/1963392720116093396)  2025-09-04T00:04Z 10.6K followers, 86.9K engagements


"I used to prototype websites in figma v0 bolt lovable etc. Now I genuinely think the best workflow is drawing an outline in @excalidraw I nail down every piece of copy and high level element then pass to a designer or go straight to vibe coding. Copy provides clarity. Everything else is just details. In [--] years I may be using Figma but my mind works so much more clearly in Excalidraw and Im able to iterate much faster"  
[X Link](https://x.com/michael_chomsky/status/1963492436220780704)  2025-09-04T06:40Z [----] followers, [---] engagements


"Founders need to spend less time building and more time on Linkedin"  
[X Link](https://x.com/michael_chomsky/status/1964084470342107282)  2025-09-05T21:53Z [----] followers, [---] engagements


"every day my respect for a16z grows lock in inspo https://t.co/ocQ3HjSfYO lock in inspo https://t.co/ocQ3HjSfYO"  
[X Link](https://x.com/michael_chomsky/status/1965322026782278023)  2025-09-09T07:50Z [----] followers, [---] engagements


"'The LLM Data Company' is making RL environments for labs because of course it is. I predict it gets bought by Atlassian by EOY"  
[X Link](https://x.com/michael_chomsky/status/1965467236337910195)  2025-09-09T17:28Z [----] followers, [---] engagements


"Literally [--] companies are lining up trying to hire this kind of person. If this is you dm me and I'll introduce you to a founder. There is insane demand for people who can understand and explain technology in a compelling way. There is insane demand for people who can understand and explain technology in a compelling way"  
[X Link](https://x.com/anyuser/status/1966598440248578295)  2025-09-12T20:23Z 10.6K followers, 689K engagements


"Ok Mintlify heres the request. No judgement AT ALL if you dont implement this. Here we go: Every single code snippet in the @inference_net documentation needs to be tested. Right now we have a script that extracts snippets into individual files We can then easily run code snippets in individual files or just run all python files for example. We have another command to update the snippets inside the docs once changes are made in individual files. Ideally this would be handled by mintlify. Id like a run button next to code snippets when running in dev that runs the code snippet and shows me the"  
[X Link](https://x.com/michael_chomsky/status/1966684034186281024)  2025-09-13T02:03Z [----] followers, [----] engagements


"please send me some of your best writing videos or anything else youre proud of if you sent me a dm. It will save me a great amount of time filtering through hundreds of dms. I only introduce founders to the best but devrel talent comes in all forms Bonus points if there is a specific company on you want to work for http://devreljob.com http://devreljob.com"  
[X Link](https://x.com/anyuser/status/1966956005063094589)  2025-09-13T20:03Z 10.6K followers, [----] engagements


"Im willing to pay $300 for software that goes through my twitter dms deepresearches and identifies everyone and places them into a CRM"  
[X Link](https://x.com/anyuser/status/1967285835835027660)  2025-09-14T17:54Z 10.6K followers, 217.8K engagements


"@kristianvtr there's specific models for this now like https://inference.net/models/osmosis-structure-0.6b https://inference.net/models/osmosis-structure-0.6b"  
[X Link](https://x.com/michael_chomsky/status/1967324756174372923)  2025-09-14T20:29Z [----] followers, [----] engagements


"I couldn't find a single API on the internet that just takes a twitter user's handle and enriches their information (at least with a Linkedin profile/email/relevant websites.) Someone is going to solve this problem and have people absolutely begging to pay them. Im willing to pay $300 for software that goes through my twitter dms deepresearches and identifies everyone and places them into a CRM. Im willing to pay $300 for software that goes through my twitter dms deepresearches and identifies everyone and places them into a CRM"  
[X Link](https://x.com/anyuser/status/1967353317241299256)  2025-09-14T22:22Z 10.6K followers, 108.3K engagements


"@kwogga_ idk about all that"  
[X Link](https://x.com/michael_chomsky/status/1967421276554953103)  2025-09-15T02:52Z [----] followers, [---] engagements


"@RhysSullivan Rhys how do you balance the benefits Twitter gets you with the very obvious downsides I get a lot from this app but overall more bad than good"  
[X Link](https://x.com/michael_chomsky/status/1967453222794346911)  2025-09-15T04:59Z [----] followers, [---] engagements


"@RhysSullivan Gotcha yeah hopefully we save our prefrontal cortex and stay sane"  
[X Link](https://x.com/michael_chomsky/status/1967454463838597453)  2025-09-15T05:04Z [----] followers, [--] engagements


"@HanchungLee idk if that's the main issue. there's really not much info you can get from a twitter handle without some sophisticated web agents"  
[X Link](https://x.com/michael_chomsky/status/1967479251487347051)  2025-09-15T06:43Z [----] followers, [---] engagements


"I like all my own posts and will continue to do so. Any attention I lose goes to someone else and as a utilitarian I find this invariably unproblematic. Sometimes I bookmark too but only if I find my own post worthy of my very exclusive palace of bookmarks I never read. This one isn't"  
[X Link](https://x.com/anyuser/status/1967480822631747840)  2025-09-15T06:49Z 10.6K followers, [----] engagements


"Posting on X feels like hopelessly throwing a fishing line into the void and then suddenly catching an [---] lb. tuna"  
[X Link](https://x.com/anyuser/status/1968011249133961557)  2025-09-16T17:57Z 10.6K followers, [----] engagements


"People forget more value was created by YC rejects than successful YC applicants If youre a YCombinator reject email me your application and lets talk yc@launch.co If youre a YCombinator reject email me your application and lets talk yc@launch.co"  
[X Link](https://x.com/anyuser/status/1968149185120932244)  2025-09-17T03:05Z 10.6K followers, 303.3K engagements


"If you're a successful founder and were rejected by YC either initially or in all attempts please DM me. I'm attempting to accurately estimate the value of all companies with founders rejected from YC. Here's what I have so far: People forget more value was created by YC rejects than successful YC applicants People forget more value was created by YC rejects than successful YC applicants"  
[X Link](https://x.com/anyuser/status/1968195008638615668)  2025-09-17T06:07Z 10.6K followers, 61.1K engagements


"@pirchavez @mattturck dude this really moves the needle on my calculations"  
[X Link](https://x.com/michael_chomsky/status/1968235305330467258)  2025-09-17T08:47Z [----] followers, [---] engagements


"A good devrel has their pick of any post-PMF developer facing company in the valley. The real challenge is getting in before the series B to get that generational bag. Same for any good engineer that can write. If youre an engineer with a few great blog posts companies will start aggressively fighting over you. Its wonderful to watch really. theres never been an easier time to join a top tier startup. all you have to do is become obsessed at one discipline (eng design product etc.) for literally a couple months and become world class at it. the bar is so low cause ppl cant stick to a thing"  
[X Link](https://x.com/michael_chomsky/status/1968526791968178594)  2025-09-18T04:05Z [----] followers, [----] engagements


"My favorite ML account rn and its not close. Bro picked up 4k followers in like [--] days writing some of the most banger content I have ever seen"  
[X Link](https://x.com/michael_chomsky/status/1968901108006301853)  2025-09-19T04:52Z [----] followers, [----] engagements


"People who are good at this would absolutely CRUSH in pre-sales. Vibe coded monstrosities on short notice is the new pre-sales meta. If you are a goat at vibe-coding polished things quickly DM me and Ill try to find you a pre-sales job (not working with founders yet for this but I expect presales-vibery to be an entire job category soon.) Marketing idea: 💡 Create a daily/ weekly live "vibe coding show" that puts spotlight on AI and vibe coding tools launched into the market but centered around showing what's possible with the tools vs yapping lofty things Focus more on live demos - less on"  
[X Link](https://x.com/anyuser/status/1969188345671795106)  2025-09-19T23:54Z 10.6K followers, [----] engagements


"@AlmostMedia Its so tough. Im working with companies to find strong devrels and even a single placement takes so much time and effort"  
[X Link](https://x.com/michael_chomsky/status/1969479955571753032)  2025-09-20T19:13Z [----] followers, [---] engagements


"I don't think most people understand that creating a good simple todo app is a herculean feat. Show me one minimalistic todo app. You can't. Because it doesn't exist. I'll find accessibility errors over-complication a lack of basic features that I need unnecessary animations etc. Show me that you've actually built a good minimalistic todo app and I'll recommend you the next time someone asks me if I know a good UX engineer. (this might be projecting because I'm trying and failing to build a todo app I'm happy with for myself.)"  
[X Link](https://x.com/anyuser/status/1969800539081736546)  2025-09-21T16:27Z 10.6K followers, [----] engagements


"@pocketbytes i feel that. Im terrible at budgeting and would pay for an app thats beautiful makes me better at it"  
[X Link](https://x.com/michael_chomsky/status/1969808598676259105)  2025-09-21T16:59Z [----] followers, [--] engagements


"Vercel needs way more credit for this imo. DX like this doesn't grow on trees"  
[X Link](https://x.com/anyuser/status/1969822684688376151)  2025-09-21T17:55Z 10.6K followers, 123.3K engagements


"its [----] you have an AI app idea you go to v0 it builds you an app in Next.js + shadcn using a fine-tuned Next.js LLM AI gateway + AI SDK supported out of the box your app is cooked change your theme with one prompt want vercel analytics add with one click bot protection done. one click deploy to Vercel you need a domain name vercel domains integrates with your deployment you can buy with one click its [--] weeks later and youre at 20k MRR Vercel sends you an email: based on your analytics and AI logs retention is dropping slightly you open up Cursor for the first time and clone the repo you"  
[X Link](https://x.com/anyuser/status/1969947315390812181)  2025-09-22T02:10Z 10.6K followers, 356.6K engagements


"This is unironically Guillermos master plan for Vercel. The first stage was vertical integration across the stack: created Next.js Vercel to deploy apps easily Turborepo acquired to make dev experience better and to own monorepo builds/caching Fluid compute and heavy investment into serverless infra Vercel Firewall Then horizontal integration across the JS ecosystem: NuxtLabs acquisition Hiring Rich Harris (svelte) Astro sponsorship/partnership Then horizontal integration across the new AI development stack: SplitBee acquisition-Vercel Analytics hires ShadCN (guy behind ShadCN component lib"  
[X Link](https://x.com/anyuser/status/1970188623401161120)  2025-09-22T18:09Z 10.6K followers, 86.5K engagements


"Once you realize: you only want to get rich to ship whatever you want wherever you want you'd probably have a job if you were rich anyway for the free Bay Club membership (it feels better when you don't pay) you can ship whatever you want now already startups are about getting generational wealth with others which is 10x better than indie-hacking in a basement the demons go away. Jack Dorsey is the best example of this"  
[X Link](https://x.com/michael_chomsky/status/1970197656203907173)  2025-09-22T18:45Z [----] followers, [----] engagements


"@nummanthinks They can def sell v0 preference and manual UI change data. Whenever a user changes a color or a border thats absolute gold"  
[X Link](https://x.com/michael_chomsky/status/1970200437295845795)  2025-09-22T18:56Z [----] followers, [----] engagements


"@alsibassi 100%. An IPO makes life simpler"  
[X Link](https://x.com/michael_chomsky/status/1970210976361255286)  2025-09-22T19:37Z [----] followers, [--] engagements


"Literally all the resources you need to be a cracked ML/LLM engineer (coming from someone learning ML rn) unsloth documentation vLLM documentation (goated) HF TEI docs (worth reading) GPU glossary I wish I started with these a month ago. Expect much better ML content from me now that I have found them. What am I missing"  
[X Link](https://x.com/anyuser/status/1970951623653384223)  2025-09-24T20:41Z 10.6K followers, 52.6K engagements


"at this point I need an app to track all the research papers I'm planning on reading"  
[X Link](https://x.com/michael_chomsky/status/1971160994409058553)  2025-09-25T10:32Z [----] followers, [---] engagements


"@dillon_mulroy @joshmanders oh cool interesting what you end up using. 300M domain names in the browser is kinda nuts"  
[X Link](https://x.com/michael_chomsky/status/1971401869449912345)  2025-09-26T02:30Z [----] followers, [--] engagements


"Be me Start a chat in Claude Spend [--] minutes carefully pasting all the context (usually from ChatGPT) Read the answer Answer is terrible Realize I was on Sonnet which is the default for some ungodly reason (why would I use Sonnet Ever) Switch to Opus This creates a new chat session because Anthropic decided I couldn't possibly want to continue my previous chat with a better model Spend [--] minutes carefully pasting all the context (usually from ChatGPT)"  
[X Link](https://x.com/michael_chomsky/status/1971579511637450869)  2025-09-26T14:16Z [----] followers, [----] engagements


"@mdanderson 100%"  
[X Link](https://x.com/michael_chomsky/status/1971739823523483876)  2025-09-27T00:53Z [----] followers, [--] engagements


"The future of LLM coding agents is tag-teaming. Sonnet is stuck Tag in grok-code. Grok-code is stuck Tag in gpt-5-high. Humans do this so why not LLMs"  
[X Link](https://x.com/michael_chomsky/status/1972732005269070116)  2025-09-29T18:35Z [----] followers, [----] engagements


"I'm deploying on Vercel. NO you should deploy on Cloudflare NO you should deploy on Hetzner NO you should deploy on a raspberry pi Guys [---] sonnet can deploy your app on a vape if you want it to. It's not a big deal"  
[X Link](https://x.com/anyuser/status/1972817735156670657)  2025-09-30T00:16Z 10.6K followers, 93.4K engagements


"this is what success looks like"  
[X Link](https://x.com/michael_chomsky/status/1973049698651930662)  2025-09-30T15:38Z [----] followers, [---] engagements


"Thanks to the resurgence of RL LLMs are finally able to reliably coordinate tools and reasoning to do high-precision retrieval. Companies like Happenstance Clado and Mintlify have already shifted to agentic search and it's only a matter of time until anything less feels broken to users. https://inference.net/blog/agentic-search https://inference.net/blog/agentic-search"  
[X Link](https://x.com/michael_chomsky/status/1973105635840630991)  2025-09-30T19:20Z [----] followers, [----] engagements


"I'm solving hosting. Stay tuned. I'm deploying on Vercel. NO you should deploy on Cloudflare NO you should deploy on Hetzner NO you should deploy on a raspberry pi Guys [---] sonnet can deploy your app on a vape if you want it to. It's not a big deal. I'm deploying on Vercel. NO you should deploy on Cloudflare NO you should deploy on Hetzner NO you should deploy on a raspberry pi Guys [---] sonnet can deploy your app on a vape if you want it to. It's not a big deal"  
[X Link](https://x.com/michael_chomsky/status/1973155394844315784)  2025-09-30T22:38Z [----] followers, [----] engagements


"I dont have Tinker or Sora access. Whats the point of being in sf if you cant be the first person to try all the cool stuff @thinkymachines I need that sweet sweet correct GRPO implementation"  
[X Link](https://x.com/michael_chomsky/status/1973534521724379159)  2025-10-01T23:44Z [----] followers, [---] engagements


"Ive never landed on a single website that uses Cloudflare Never met a single person who uses cloudflare Never met a single person who uses cloudflare"  
[X Link](https://x.com/michael_chomsky/status/1974285503987134956)  2025-10-04T01:28Z [----] followers, [---] engagements


"This would be a great AI research interview question. The try/except clauses LLMs generate are just reward hacking. The naive approach would be a small negative reward for try/except clauses. Whats the actual solution my goal in life is to join Anthropic delete all try/except clauses from Claudes training data and then quit. my goal in life is to join Anthropic delete all try/except clauses from Claudes training data and then quit"  
[X Link](https://x.com/michael_chomsky/status/1974590485022097651)  2025-10-04T21:40Z [----] followers, [---] engagements


"@jeffreyhuber Went over this in my most recent blog: https://inference.net/blog/agentic-search https://inference.net/blog/agentic-search"  
[X Link](https://x.com/michael_chomsky/status/1974671266125947089)  2025-10-05T03:01Z [----] followers, [---] engagements


"@jeffreyhuber yeah. I'm impressed with grep/ripgrep as a tool but we can do so much better. would be cool to rl an llm with ripgrep and embedding retrieval and compare that with just vector search on some code dataset and compare how it performs"  
[X Link](https://x.com/michael_chomsky/status/1974675277533098088)  2025-10-05T03:17Z [----] followers, [---] engagements


"Asked my gf if she liked Claude. She said 'that Doja Cat album only has six good songs.'"  
[X Link](https://x.com/michael_chomsky/status/1974941541434302569)  2025-10-05T20:55Z [----] followers, [---] engagements


"MY TWITTER HAS BEEN HACKED I DO NOT SUPPORT INDOR OR HAVE CRYPTOCURRENCY"  
[X Link](https://x.com/michael_chomsky/status/1975007260335284520)  2025-10-06T01:16Z [----] followers, [----] engagements


"@skeptrune n8n was just killed by OAI today. it's over I'm so sorry Nick"  
[X Link](https://x.com/michael_chomsky/status/1975343976204132413)  2025-10-06T23:34Z [----] followers, [---] engagements


"the male urge to scrape the entire internet"  
[X Link](https://x.com/michael_chomsky/status/1975407398526533804)  2025-10-07T03:46Z [----] followers, [---] engagements


"@ad0rsh such a problem. i have 11k bookmarks and no way to export them"  
[X Link](https://x.com/michael_chomsky/status/1975407977608905005)  2025-10-07T03:48Z [----] followers, [--] engagements


"@ad0rsh almost never. i cant search them effectively"  
[X Link](https://x.com/michael_chomsky/status/1975408276654501925)  2025-10-07T03:50Z [----] followers, [--] engagements


"Mintlify search UX suggestion: Sometimes I want the ai assistant answer but end up clicking enter automatically. I'd love another button other than 'enter' to take me to the ai assistant reply automatically"  
[X Link](https://x.com/michael_chomsky/status/1975738685527367880)  2025-10-08T01:43Z [----] followers, [----] engagements


"Extra points for whoever suggests the best UX for this"  
[X Link](https://x.com/michael_chomsky/status/1975744187397013996)  2025-10-08T02:04Z [----] followers, [---] engagements


"@grimcodes YES @skeptrune I seem to recall Dia handling this super well (when I still used Dia)"  
[X Link](https://x.com/michael_chomsky/status/1975778480051871915)  2025-10-08T04:21Z [----] followers, [--] engagements


"Charles is completely right. Specialized models are faster cheaper and often more accurate than foundation models on constrained tasks. But more importantly at a certain scale you're forced to move to small models. This is one of the reasons OpenAI's list of companies processing 1T tokens is so small: processing 1T tokens with frontier models is impossible unless you're a massive co with multi-million dollar budgets. But how do researchers regularly process the entire internet which is hundreds of trillions of tokens The answer is they distill small models. There's really no other way (unless"  
[X Link](https://x.com/michael_chomsky/status/1975977238962241945)  2025-10-08T17:31Z [----] followers, [----] engagements


"Shitty search experiences cause me unspeakable pain. I'm looking for a YC startup called Webhound. I search for 'hound' Webhound a yc startup isn't in the top [--] results I search 'yc hound' on google and the first result is Webhound. I'm willing to submit a PR to fix this and make YC company search 100x better. This will undoubtably have a noticeable impact on business outcomes for YC companies. How do I get access to the YC codebase"  
[X Link](https://x.com/anyuser/status/1976374103708979553)  2025-10-09T19:48Z 10.6K followers, 36.2K engagements


"@mattsilv @inference_net You mean how to do it in mintlify https://www.mintlify.com/docs/ai/contextual-menu#menu-options https://www.mintlify.com/docs/ai/contextual-menu#menu-options"  
[X Link](https://x.com/michael_chomsky/status/1976862717677125937)  2025-10-11T04:09Z [----] followers, [--] engagements


"@NERDDISCO @mastra_ai @xai @OpenRouterAI @e2b @theo @Cloudflare I like this a lot actually"  
[X Link](https://x.com/michael_chomsky/status/1977080443569840535)  2025-10-11T18:34Z [----] followers, [---] engagements


"@mattpocockuk i prefer typescript to go but ok"  
[X Link](https://x.com/michael_chomsky/status/1977192421009375526)  2025-10-12T01:59Z [----] followers, [---] engagements


"youre not ugly you just need to download a korean app that tells you the exact color shirt to wear based on your skin and beard color"  
[X Link](https://x.com/michael_chomsky/status/1977224408180912303)  2025-10-12T04:06Z [----] followers, [---] engagements


"apparently this is what happened with browserbase they built some of the best browser infra in the world and one of the top use cases ended up being automating linkedin following/messaging for salespeople. so yes browserbase is a sales tool as much as it is a devtool. revenue is revenue. turns out 9/10 times when you build a general product the best use cases are in sales and marketing turns out 9/10 times when you build a general product the best use cases are in sales and marketing"  
[X Link](https://x.com/anyuser/status/1977493479673811023)  2025-10-12T21:56Z 10.6K followers, 128.8K engagements


"@ThePolderBaan @wangzjeff looks great only interested in sf tho rn"  
[X Link](https://x.com/michael_chomsky/status/1977494795682750908)  2025-10-12T22:01Z [----] followers, [---] engagements


"@vincent_koc @browserbasehq didn't they work directly with Cloudflare to get past this"  
[X Link](https://x.com/michael_chomsky/status/1977510989328511212)  2025-10-12T23:05Z [----] followers, [----] engagements


"@0xKyon nah blow your A instead"  
[X Link](https://x.com/michael_chomsky/status/1977529121409544256)  2025-10-13T00:17Z [----] followers, [--] engagements


"@aditabrm @reductoai @a16z huge W for reducto"  
[X Link](https://x.com/michael_chomsky/status/1978140432618918053)  2025-10-14T16:46Z [----] followers, [---] engagements


"whats the best app i can download on my phone to run bg agents must run the full environment. bonus points if it sends me a tunneled link to try the app and can run for 1h+. dont tell me to ssh into a VPS Im not an animal"  
[X Link](https://x.com/michael_chomsky/status/1978271515947770317)  2025-10-15T01:27Z [----] followers, [----] engagements


"@_WEEXIAO @Osmosis_AI this is so cool"  
[X Link](https://x.com/michael_chomsky/status/1978501821879529810)  2025-10-15T16:42Z [----] followers, [---] engagements


"We just trained a model for Cal AI 🍎 Here's the full story: Cal was scaling extremely quickly and their costs were ballooning but that wasn't even the main problem. OpenAI's API kept failing requests which meant users couldn't log their meals and Henry (Cal's CTO) was losing sleep. Henry and James (their AI engineer) had already tried everything. They evaluated different providers optimized their prompts and built retry logic. They knew the bottleneck wasn't something they could prompt engineer their way out of. So they came to us about training a custom model. The model we trained is 50%"  
[X Link](https://x.com/anyuser/status/1978566463654072563)  2025-10-15T20:59Z 10.6K followers, [----] engagements


"accidentally said 'product marketing' instead of devrel and they kicked me out of san francisco"  
[X Link](https://x.com/michael_chomsky/status/1978689698080891263)  2025-10-16T05:09Z [----] followers, [---] engagements


"@peytoncasper lmaoo my bad peyton"  
[X Link](https://x.com/michael_chomsky/status/1978860237147742562)  2025-10-16T16:27Z [----] followers, [--] engagements


"We trained Schematron to solve web extraction's biggest problem: cost. I've killed so many projects because I'd do the math and realize I couldn't justify the LLM expense. Enterprises hit the same wall at 100x scale. 1M pages costs $240 with Schematron-3B vs $5k for Gemini [---] Flash (and Schematron is more accurate). If you have a scraping project that's not viable because it's too expensive too slow or the quality isn't there with general LLMsDM me. I want to hear what you're building. Today we're releasing Schematron: 3B and 8B models that extract typed JSON from HTML with frontier-level"  
[X Link](https://x.com/anyuser/status/1978911982825947566)  2025-10-16T19:52Z 10.6K followers, [----] engagements


"@Chikor_Zi pages usually average 8-10k tokens. you still need to clean html a bit--completely raw html usually has a lot of irrelevant fluff that's not worth passing to the LLM"  
[X Link](https://x.com/michael_chomsky/status/1978948792108851439)  2025-10-16T22:18Z [----] followers, [--] engagements


"@Chikor_Zi yeah you'll likely find schematron 10x faster here"  
[X Link](https://x.com/michael_chomsky/status/1978966724633678075)  2025-10-16T23:30Z [----] followers, [--] engagements


"@walden_yan published a blog on this a few weeks ago: https://inference.net/blog/agentic-search https://inference.net/blog/agentic-search"  
[X Link](https://x.com/michael_chomsky/status/1979029681228673317)  2025-10-17T03:40Z [----] followers, [---] engagements


"I published a blog on this a few weeks ago. Very exciting to see a model like this so quickly but there are still many easy wins to be made https://inference.net/blog/agentic-search SWE-grep and SWE-grep-mini are creating a new Pareto frontier. The Fast Context agent can run [--] turns of agentic search in less than [--] seconds approaching the speed of an embedding search. https://t.co/GIHVXMbjCh https://inference.net/blog/agentic-search SWE-grep and SWE-grep-mini are creating a new Pareto frontier. The Fast Context agent can run [--] turns of agentic search in less than [--] seconds approaching the"  
[X Link](https://x.com/michael_chomsky/status/1979030392406523907)  2025-10-17T03:43Z [----] followers, [----] engagements


"Finally an underhyped AI product Weve expanded our index to include tweets YouTube countless pdfs and much more allowing you to search more of the web in expressive and unusual ways Heres search over twitter (2/7) https://t.co/gjHzX7mC47 Weve expanded our index to include tweets YouTube countless pdfs and much more allowing you to search more of the web in expressive and unusual ways Heres search over twitter (2/7) https://t.co/gjHzX7mC47"  
[X Link](https://x.com/michael_chomsky/status/1648588412251959297)  2023-04-19T07:24Z [----] followers, [---] engagements


"@waymodriver Dm me some of your best writing"  
[X Link](https://x.com/anyuser/status/1966777620697379254)  2025-09-13T08:15Z 10.6K followers, [----] engagements


"What was the most insane growth hack you've ever seen I think about @TheIshanGoswami's twitter roast app all the time"  
[X Link](https://x.com/michael_chomsky/status/1967791996371980550)  2025-09-16T03:25Z [----] followers, [---] engagements


"CTO IS ORDERING OFFICE SNACKS FROM COSTCO RN WHAT SHOULD I ASK FOR"  
[X Link](https://x.com/michael_chomsky/status/1976376701031678367)  2025-10-09T19:58Z [----] followers, [----] engagements


"this is a billion dollar idea right here. just wrap nick's workflow into an app sell it to late stage YC companies as the new 'must-have boost for engineer productivity.' here's the full game-plan that will turn you into a billionaire. the secret is raising 100M and buying billboards all over sf. even if your company fails you'll be respected as a lion that took risks. put out ads like: engineers paid 400k a year should be putting out PRs while on the toilet. be demure. be thoughtful. code at the bar--but on your phone. push to prod on friday with confidence. your employees have claude code"  
[X Link](https://x.com/anyuser/status/1977407161149088090)  2025-10-12T16:13Z 10.6K followers, [----] engagements


"what's the best tool for adversarial LLM tasks say I want an LLM to keep iterating on a tweet until a judge LLM decides it deserves a 9.5/10. I don't want to use DSPY I just want to write a prompt in some UI and get an email once it finally manages to do it or fails after [--] minutes"  
[X Link](https://x.com/michael_chomsky/status/1977899096586518569)  2025-10-14T00:47Z [----] followers, [----] engagements


"@the_mewc sure. but it's a very specific type. we used to call these 'LLM gates' and they're a known strategy to improve LLM outputs considerably"  
[X Link](https://x.com/michael_chomsky/status/1977903095079477314)  2025-10-14T01:03Z [----] followers, [--] engagements


"WE'VE RAISED $11.8M 🚀 It's been an absolutely incredible watching this brilliant team grow over the past [--] months. When I joined Inference was [--] people in a mostly empty office. Now it's almost double that and while some things have changed (Smash has been perma-banned at the office and the fridge is now religiously stocked with cherry-cola Celcius) the insane shipping velocity I joined for is still going strong. I came in at a precarious time: Inference had been building out a network for decentralized LLM inference for years but the team decided to go all-in on a completely new direction:"  
[X Link](https://x.com/anyuser/status/1978184443035816208)  2025-10-14T19:41Z 10.6K followers, 38K engagements


"the lions gf thinks that salesforce is just a bunch of dudes in a tower doing sales and the lion doesnt have the heart to tell her the truth"  
[X Link](https://x.com/michael_chomsky/status/1978283200053936548)  2025-10-15T02:14Z [----] followers, [---] engagements


"As someone who uses warp every day I dislike this ad. Yes warp codes for you. But it should be a terminal first and foremost. I use warp mostly as a terminal and this is a confusing message imo. Comet browser is still a browser despite using its base as a browser to also do something much more powerful. That said if it works it works"  
[X Link](https://x.com/michael_chomsky/status/1978945737917669722)  2025-10-16T22:06Z [----] followers, [---] engagements


"Yes and theyre only getting better. Prompting them well get you pretty far too (think GEPA). But training can beat frontier models at a price and latency that can go as far as opening up new use cases. The tools in developers toolchest is expanding. Also scaffolds can mean major albeit temporary performance wins. Think Cognition: they built sophisticated scaffolds and now they are big enough to train their own models and create even more of a moat"  
[X Link](https://x.com/michael_chomsky/status/1979417534593204259)  2025-10-18T05:21Z [----] followers, [--] engagements


"I see a future where YC and VCs start pouring money into lobbying to protect early employees and founders. Our engineer just got detained at the US border in SFO and hasnt responded for [--] hours What to do in this case Our engineer just got detained at the US border in SFO and hasnt responded for [--] hours What to do in this case"  
[X Link](https://x.com/michael_chomsky/status/1979421028993634759)  2025-10-18T05:35Z [----] followers, [----] engagements


"I asked chatgpt to find me a heavy case to help with excessive phone use. Technically it did find me a viable solution"  
[X Link](https://x.com/michael_chomsky/status/1979568653319217596)  2025-10-18T15:22Z [----] followers, [----] engagements


"IF YOU ARE THROWING A HALLOWEEN PARTY IN SF I HUMBLY ASK TO BE INVITED. I HAVE TOO MANY NICHE COSTUME IDEAS. here are some you can steal: dreamforce attendee vercel employee (all black w/ turtleneck) man with hammer in OAI datacenter in case AI becomes evil bun only in sf will you find someone dressed up as Claude for halloween https://t.co/FlOtaGe9p1 only in sf will you find someone dressed up as Claude for halloween https://t.co/FlOtaGe9p1"  
[X Link](https://x.com/michael_chomsky/status/1980048544213159974)  2025-10-19T23:08Z [----] followers, [----] engagements


"Id pay $2500 to see a g vs Zuck match live I had my first BJJ fight the other day. I see what the hype is all about now. All my boys are insta-enrolled. Brazil cooked. https://t.co/oFMzWe6gTu I had my first BJJ fight the other day. I see what the hype is all about now. All my boys are insta-enrolled. Brazil cooked. https://t.co/oFMzWe6gTu"  
[X Link](https://x.com/anyuser/status/1980054272671994254)  2025-10-19T23:31Z 10.6K followers, 18.9K engagements


"@skeptrune Effect still has no clear best practices that I have seen. Seems you still need to spend a lot of time on it to grok it. People that are super effect-pilled might disagree"  
[X Link](https://x.com/michael_chomsky/status/1980086129429991503)  2025-10-20T01:38Z [----] followers, [---] engagements


"@skeptrune Im not sure I understand. Lodash is mostly sync helpers. Effects main draw is type-safety for async code"  
[X Link](https://x.com/michael_chomsky/status/1980089189044769266)  2025-10-20T01:50Z [----] followers, [---] engagements


"@skeptrune @typescript @EffectTS_ what are the chances of this happening"  
[X Link](https://x.com/michael_chomsky/status/1980091006533140975)  2025-10-20T01:57Z [----] followers, [---] engagements


"Its a great start but the biggest problem for me is that I have no understanding of general patterns that emerge in real codebases using Effect. The ideal thing is a prompt that can teach me and an LLM how to use Effect well as the Effect docs just didnt do that last time I tried. something like a to an effect codebase http://CONTRIBUTE.md http://CONTRIBUTE.md"  
[X Link](https://x.com/michael_chomsky/status/1980092735362355460)  2025-10-20T02:04Z [----] followers, [---] engagements


"@skeptrune @andrewqu_ winner gets to fight me for ownership of http://smry.ai http://smry.ai"  
[X Link](https://x.com/michael_chomsky/status/1980097635135717398)  2025-10-20T02:23Z [----] followers, [---] engagements


"devreljobs is down. but so is AWS so"  
[X Link](https://x.com/michael_chomsky/status/1980185475483926550)  2025-10-20T08:13Z [----] followers, [---] engagements


"insane. couldnt be more bullish about daytona and e2b Weve broken records before at @daytonaio. But this one broke the chart. Over 1M sandboxes in a day. https://t.co/xlQtYsvfqA Weve broken records before at @daytonaio. But this one broke the chart. Over 1M sandboxes in a day. https://t.co/xlQtYsvfqA"  
[X Link](https://x.com/michael_chomsky/status/1980352475233415563)  2025-10-20T19:16Z [----] followers, [----] engagements


"@jxnlco we trained a model for this thats 17x cheaper than sonnet. try it: http://cliptagger.inference.net http://cliptagger.inference.net"  
[X Link](https://x.com/michael_chomsky/status/1980448040315154786)  2025-10-21T01:36Z [----] followers, [---] engagements


"This is moving in the right direction. Specialized agents to search your context others to make diffs (think morphllm) and frontier models to orchestrate. Absolutely love this. Cognition is about [--] months ahead of the curve. Introducing SWE-grep and SWE-grep-mini: Cognitions model family for fast agentic search at [----] TPS. Surface the right files to your coding agent 20x faster. Now rolling out gradually to Windsurf users via the Fast Context subagent or try it in our new playground https://t.co/2K2dzum9CX Introducing SWE-grep and SWE-grep-mini: Cognitions model family for fast agentic"  
[X Link](https://x.com/anyuser/status/1980803036948910290)  2025-10-22T01:07Z 10.6K followers, [----] engagements


"Update devrel jobs now has a compact card view. It's exactly the same but the cards are smaller. #shippedthisatthebar"  
[X Link](https://x.com/anyuser/status/1981028774960591121)  2025-10-22T16:04Z 10.6K followers, [----] engagements


"Hetzner blocked my accounts twice. Id make an account miss a payment or something and bam I cant log in to pay my balance. Burned my two primary emails doing this. Could never reach a person. Nobody cared about my business. Hetzner is literally 5x cheaper than some alternatives which all have their own issues. But the fact that I literally could not reach a human was the reason I gave up on them. Id rather pay 5x more somewhere else that actually responds. Even drug dealers have better support (I assume) In practice this really sucks. There are things I want to do that are only economically"  
[X Link](https://x.com/anyuser/status/1981418302040920276)  2025-10-23T17:51Z 10.6K followers, 110.9K engagements


"@NabilHunt Vercel is this for AWS"  
[X Link](https://x.com/michael_chomsky/status/1981621817354662207)  2025-10-24T07:20Z [----] followers, [----] engagements


"i heard yesterday that the AI SDK is grass-fed farm-raised hand-written trad-crafted high-perplexity primate-powered code lars is cooking but at what cost. We need an official name for not-vibe-coding. My top candidates so far: boomer coding chewgy coding trad coding Coding with capital C Suggestions are welcome. We need an official name for not-vibe-coding. My top candidates so far: boomer coding chewgy coding trad coding Coding with capital C Suggestions are welcome"  
[X Link](https://x.com/michael_chomsky/status/1981759590313377875)  2025-10-24T16:28Z [----] followers, [----] engagements


"The highest AI cost for most companies is still not OpenAI it's people. Optimizing LLM inference at scale is still a rare skill that's hard to hire for. That's why what we're doing at @inference_net is so cool--we basically save you 700k/yr for an ML guy and an AI infra guy"  
[X Link](https://x.com/michael_chomsky/status/1981785082324299810)  2025-10-24T18:09Z [----] followers, [---] engagements


"@mwfowlie @inference_net its not completely automated rn. we take the time to deeply understand client problems and find a solution"  
[X Link](https://x.com/michael_chomsky/status/1981940111878312094)  2025-10-25T04:25Z [----] followers, [--] engagements


"@skeptrune use image i heard yesterday that the AI SDK is grass-fed farm-raised hand-written trad-crafted high-perplexity primate-powered code lars is cooking but at what cost"  
[X Link](https://x.com/michael_chomsky/status/1981953243757281477)  2025-10-25T05:17Z [----] followers, [--] engagements


"use client use server use cache use workflow useguys chill under the hood its still use aws"  
[X Link](https://x.com/michael_chomsky/status/1981954897764557112)  2025-10-25T05:24Z [----] followers, [---] engagements


"if Im not answering dms its because Im desperately researching the difference between GRPO PPO Dr. PPO KTO CPO C3PO RLOO BCO ORPO XPO DDPO GKD PRM SFO HMO GMO MMO MMA and RPG. Bc if youre still on SFT youre NGMI. Taking careful notes and will share in time. Stay safe out there guys"  
[X Link](https://x.com/michael_chomsky/status/1983950276466766077)  2025-10-30T17:33Z [----] followers, [----] engagements


"@dominikdoesdev come out with me tomorrow night"  
[X Link](https://x.com/michael_chomsky/status/1984543096177377599)  2025-11-01T08:48Z [----] followers, [--] engagements


"I just used the Mintlify MCP and it's actually so good. I'm actually a bit surprised. I expected it to be just ok but it's a much better experience than I thought. I wouldn't be surprised if this was the most popular MCP on Cursor a year from now"  
[X Link](https://x.com/anyuser/status/1984728282416439665)  2025-11-01T21:04Z 10.6K followers, 51.1K engagements


"even crazier: OAI is likely paying for Ilya's legal fees Ilya founded SSI to do "a new and different kind of research" whatever that means OAI and Anthropic were strongly considering merging The merger didn't go through because of practical challenges given by the Anthropic side Mira Murati told Ilya that Greg Brockman was "essentially fired" from Stripe and that Sam was pushed out of YC for lying Ilya didn't expect OAI employees to care much when Sam was removed which turned out to be wrong The lawyers on both sides low-key sound like they hate each other From Ilyas deposition Ilya plotted"  
[X Link](https://x.com/anyuser/status/1985099252968898664)  2025-11-02T21:38Z 10.6K followers, 285.3K engagements


"@AsphaltCowb0y I couldn't even test on the free tier. I kept getting rate-limited for oauth which is insane"  
[X Link](https://x.com/michael_chomsky/status/1985225985881723252)  2025-11-03T06:02Z [----] followers, [--] engagements


"@AsphaltCowb0y sucks. I think the project I'm working on would work without it but it would require all kinds of gymnastics. better to just pay elon"  
[X Link](https://x.com/michael_chomsky/status/1985229511718584455)  2025-11-03T06:16Z [----] followers, [--] engagements


"@skeptrune how did you decide on this location Nick"  
[X Link](https://x.com/michael_chomsky/status/1985260999805706392)  2025-11-03T08:21Z [----] followers, [---] engagements


"Im in Germany and theres no wework where I am. My hotel room wifi is 4mbps/s. I havent seen a startup hoodie in days and impeccably dressed people on the street keep handing me money (Im wearing regular sf clothes) My products waitlist has grown to over [---] people but bun install never completes. Is this why europeans keep moving to sf Because I get it. Im so sorry you have to live like this. Id happily give up a place in YC for a European refugee in need"  
[X Link](https://x.com/anyuser/status/1986729381474058679)  2025-11-07T09:36Z 10.6K followers, 31.3K engagements


"Saw this post Skipped it way too long Saw @leerob call it a banger in the comments wait I may have to read this Saw its by the Val town founder Ok I have to read this top to bottom Read this top to bottom Agree completely Have something to add Vibe coding should strive to never increase cognitive debt but instead DECREASE IT by explaining things you dont know or referencing relevant resources. Cognitive debt is code you do not understand and have to maintain and add to. If you dont need to maintain or add to codejust vibe it np. But if youre in charge of a production app not understanding how"  
[X Link](https://x.com/michael_chomsky/status/1987232548330610808)  2025-11-08T18:55Z [----] followers, [----] engagements


"@stevekrouse Super grateful to Leerob for commenting banger so I read all that. Heres my take on where vibe coding is going: https://x.com/michael_chomsky/status/1987232548330610808 Saw this post Skipped it way too long Saw @leerob call it a banger in the comments wait I may have to read this Saw its by the Val town founder Ok I have to read this top to bottom Read this top to bottom Agree completely Have something to add Vibe coding should https://x.com/michael_chomsky/status/1987232548330610808 Saw this post Skipped it way too long Saw @leerob call it a banger in the comments wait I may"  
[X Link](https://x.com/michael_chomsky/status/1987232855341334941)  2025-11-08T18:56Z [----] followers, [---] engagements


"I just migrated from Next [--] to Next [--] and my app is like 10x faster. I have no idea why. Can anybody from @vercel help me figure it out"  
[X Link](https://x.com/anyuser/status/1987490564434960547)  2025-11-09T12:00Z 10.6K followers, 14K engagements


"The feeling when youre one banger away from generational wealth"  
[X Link](https://x.com/anyuser/status/1987596299974193265)  2025-11-09T19:01Z 10.6K followers, [----] engagements


"Flew to Europe to get an extra [--] hours on the YC application. Havent started it yetwill write it live on tpot spaces tomorrow. The cup is full of hot salty mineral water that flows from a reservoir under a palace. As I admire a hodgepodge of elegant buildings mansions and bohemian hotels to the gentle bubbling of a river I can think of only one thing: A certain orange room placed squarely in objectively the worst part of san francisco. Everyone who joins me tomorrow will be able to contribute to the application. If I make it back to sf wagmi to sf"  
[X Link](https://x.com/anyuser/status/1987684620884725865)  2025-11-10T00:51Z 10.6K followers, [----] engagements


"@BhavnickMinhas @skeptrune 200-300 feels super high unless this includes a cold start. same aws region queries should be super fast too. for reference my neon queries average 5ms but are way slower with a cold start. what was the final setup"  
[X Link](https://x.com/michael_chomsky/status/1987707136332497209)  2025-11-10T02:21Z [----] followers, [--] engagements


"@BhavnickMinhas @skeptrune ill be the first to read it. serverless dbs are so damn convenient but every single one has niche issues they dont advertise"  
[X Link](https://x.com/michael_chomsky/status/1987707929219440949)  2025-11-10T02:24Z [----] followers, [--] engagements


"Brands are so desperate to rent headers because they are directly associated with a person. Its like seeing a thumbs up from a creator. Completely different from regular X ads that have the worst CPM in the industry. Im seeing crazy CPMs for headerswell above $20. If youre an influencer that hasnt received an offer for your header its likely because of the friction brands experience sponsoring them. Theres no analytics its hard to make sure the creator isnt tweeting things that are against brand guidelines (creators also need help understanding the limits) and payment is HARD as every creator"  
[X Link](https://x.com/michael_chomsky/status/1987795179315569012)  2025-11-10T08:11Z [----] followers, [----] engagements


"To train an LLM with RL we give the LLM some tools and reward it for completing a task successfully. You can give it tools in lots of different formats but based on what I have seen by far the most pervasive is MCP. As big labs post-train frontier models on thousands of RL envs all standardized to use MCPs frontier models will naturally become super effective at using MCPs. This may even result in emergent MCP-related abilities like effectively using large sets of tools as long as they are provided in MCP format. But in general I assume in a year models will simply perform better if you use"  
[X Link](https://x.com/michael_chomsky/status/1987928972580073701)  2025-11-10T17:02Z [----] followers, [--] engagements


"@skeptrune @andrhlt maybe not but I can get you a Yerba sponsorship"  
[X Link](https://x.com/michael_chomsky/status/1988026385689178172)  2025-11-10T23:30Z [----] followers, [---] engagements


"@skeptrune @eliasstravik when I get to 5k MRR Ill email Monster for you"  
[X Link](https://x.com/michael_chomsky/status/1988029045817102573)  2025-11-10T23:40Z [----] followers, [--] engagements


"Chad X engineers push straight to prod. None of the Voodoo PM HCI UX QA bs. Just a couple engineers with an alias for pushing to main. They just like me fr"  
[X Link](https://x.com/michael_chomsky/status/1988384715691487514)  2025-11-11T23:13Z [----] followers, [---] engagements


"@stan_laisy this is peak UX idk what you mean"  
[X Link](https://x.com/michael_chomsky/status/1988409328559419515)  2025-11-12T00:51Z [----] followers, [--] engagements


"My dad who knows absolutely nothing about web development wanted to build an AI flight chat app. So I just spent the last two hours explaining HTML CSS Typescript React basics + useEffect hook server vs client browsers Next.js APIs route handlers file-based routing serverless functions Typescript data validation with zod RPCs (it came up) React Query (told them to watch out for AI generated useEffect hooks) LLMs streaming Vercel for deployment and AI SDK basics IN RUSSIAN. I attempted to use the Socratic method to go from zero to functional AI chat app on pen and paper. It was the most"  
[X Link](https://x.com/michael_chomsky/status/1988700983938154958)  2025-11-12T20:10Z [----] followers, [---] engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing