Vram Altman

Cluely summarized in 90 seconds. @im_roy_lee @cluely anything missing?

Make something people want.

every single time

@jasonpiepmeier @dylan522p it's called negging and it works

@dylan522p I'm talking about Dario's Gulf comments. Reads like "at first we didn't want your money Gulf States, but then we realized we need you after all even though we are lumping you in with 'Bad Persons' and China" Agree w round omega oversubscribed tho. I respect ur opinion highly🙏🏻

Anthropic has been a series of ideological decisions later defeated by business realities

@willccbb reminds me of arxiv.org/abs/2409.13697 which seems underhyped

cant stop thinking about this one insanely elegant, seems insanely powerful

@OpenAI manus competitor?

ChatGPT agent is ready to introduce itself. openai.com/live

Bro bought Twitter but can't decide if he wants to turn it into openai, tiktok or pornhub

Roy Lee will be taught in B-schools in 20 years.

@frantzfries instead of?

not sure how you see a chart like this then just continue living your life

@solitaryintp @ArfurRock sir, it's called meme

@VramAltman @ArfurRock Lazy and unoriginal

We're drunk on our stupidity, again. Bribing founders with homes? Free secondaries as % of investment? Celebrating founders who leave the team out to dry? Counting GMV, bookings, distribution contracts, and equity raised as "ARR"? Manufacturing positive GM by burying COGS? Taking 10x+ more in secondaries than ARR? Lying about illicit content on platform? Fake performance data to arrange meetings with S****, S**, for clout? Call girls on corporate cards? All while investors cheer from the sideline and beg to give more. FTX 2.0 is coming. p(0.9) <24 months. I won't stop the music, just a reminder — it's so obvious. You're not serious people. We have the Mandate of Heaven. History is our promise. Be better. Enjoy the party in the meantime. I'll be waiting on the other side.

It’s time.

@willccbb sadly, not really maintained anymore though

@willccbb yep, github.com/openai/simple-…

are there any standard LLM evaluations that have a friendly programming interface and clean documentation for using them with API clients? i do not want to run a custom cli tool i want to uv add and import and pass an OpenAI client

@calvinfo @OpenAI please don't tell me you go to meta

As they say, some personal news– I just left @OpenAI after launching Codex. Extremely grateful to everyone there who I got the chance to work with and learn from. Still figuring out what's next, but there's a lot left to build out there.

@miramurati congrats; very excited

Thinking Machines Lab exists to empower humanity through advancing collaborative general intelligence. We're building multimodal AI that works with how you naturally interact with the world - through conversation, through sight, through the messy way we collaborate. We're excited that in the next couple months we’ll be able to share our first product, which will include a significant open source component and be useful for researchers and startups developing custom models. Soon, we’ll also share our best science to help the research community better understand frontier AI systems. To accelerate our progress, we’re happy to confirm that we’ve raised $2B led by a16z with participation from NVIDIA, Accel, ServiceNow, CISCO, AMD, Jane Street and more who share our mission. We’re always looking for extraordinary talent that learns by doing, turning research into useful things. We believe AI should serve as an extension of individual agency and, in the spirit of freedom, be distributed as widely and equitably as possible.  We hope this vision resonates with those who share our commitment to advancing the field. If so, join us. thinkingmachines.paperform.co

@mntruell @patrickc you look like brothers

A conversation with @patrickc on old programming languages, software at industrial scale, and AI's effect on economics/biology/Patrick's daily life. 00:15 - Why Patrick wrote his first startup in Smalltalk 03:35 - LISP chatbots 06:09 - Good ideas from esoteric programming languages 09:12 - Brett Victor and Dynamicland 16:37 - Programming human organizations 20:28 - A codebase's "Big Bang" moment and MongoDB 25:49 - Rewriting Stripe 32:00 - How do you, Patrick Collison, use AI? 38:25 - Changes to GDP/TFP 41:56 - Programming human biology 46:10 - Unexpected beneficiaries of AI

@jxmnop what do you mean; aren't all those recent self-learning papers variants of this?

pretty crazy that DeepMind got self-play working w AlphaZero in *2017* yet basically no one has been able to make it work since

@haridigresses @andrewaaren @jeffwsurf you're right that GMs are not comparable as of now, but inference costs are dropping 3-5x every few months which is accounted for in those multiples

I'm completely aware of the benchmarks but I'll respectfully disagree for one reason and one reason only: ARR multiples are useful as a heuristic when comparing SaaS (all 70-90% margin). Otherwise it's apples and oranges. The moment you are comparing 20% GM and 80% GM, you have to normalize, and the only way to do that is to compare gross profit multiples.

Debrief / wrapup on Windsurf-Cognition: Deal terms & dynamics - Amazing that the team is getting taken care of (at least acceleration / participation in the deal terms) –– incredible job by @jeffwsurf and fellow execs - Guessing Founders Fund played a notable role here. They are investors in both Windsurf and Cognition. - I'd guess acquisition price was in the range of $200-400M. Typically, a business in the AI application layer with $80M+ in ARR, 2xing year on year, with 20-30% gross margin (I'm guessing) would be worth $250-500M. But the dynamics here would haircut that by 50%+ => $100-200M. Add in $100M of cash and the fair total consideration paid would have been $200-300M. - That works out to 5-8% of Cognition (last valued at $4B). If so, totally worth it IMHO. - Very curious what that means -- must have been entirely in Cognition stock. So... ~10M shares still remaining on the Windsurf cap table = $200M => a better price / share than the Google acquisition?! Admittedly this is in Cognition stock not Google stock, so it's far more speculative and should carry a premium. Post-merger dynamics - Product is a near-perfect fit for Cognition / Devin, it's a brilliant move by @ScottWu46 - This ~4xes Cognition's team (from 80 to 300, I'm assuming there will be some attrition). Culture will swing dramatically –– a lot of the incoming folks are GTM, while Cognition is mostly product / engineering. Again, complimentary, but a thoughtful integration is going to be critical to avoid organ rejection. Final notes: The fact that the remaining team immediately sold to a competitor speaks volumes. Google would not have wanted to make Cognition (or any other viable AI coding agent player) stronger. This hints that the remaining Windsurf team didn't want to "keep operating this business standalone" but rather wanted to go big –– and found their own way. What a fascinating story. Overall, the space just got a lot more interesting (and competitive), and it's awesome to hear the employees were looked after, even if it's not with an immediate, liquid exit.

@shaneguML @EMostaque talk to any western enterprise/corporate and you will learn that (sadly) they won't touch any Chinese model with a ten-foot pole. if those startups in question can greenwash it through fine-tuning, maybe they have a chance

@EMostaque Yes

RIP to the unicorn AI startups that have zero products, zero foundation models, and were just going to depend on big labs releasing open-source models for free to model merge. I know one or two.

@corbtt what do you mean by > productivity stats basically haven't changed.

it would be hilarious to time-travel back to 2019 and show claude code to a group of engineers and tell them that in 2025 everyone can use this. but then ALSO tell them that productivity stats basically haven't changed. they'd totally crash out.