andrew chen

for folks thinking about a role change this summer, we want to find you a job at a startup -- we’ve invested in >350 companies so far via a16z speedrun. I’ve spent time with all of the founders: I’m betting (literally) that they’re the next Cursor, Kalshi, Decagon, Databricks, etc. their jobs are never listed - they tend to hire via referral or from our intros via the speedrun talent team want to work with them? Let us know here: bit.ly/4fe2B4V if the team sees a match, they’ll reach out to intro you to the founders.

@justmazer this isn’t instagram

summer life update

@petersellis Core market

@staysaasy i, for one, welcome our new robot nannies

@andrewchen Next easiest solution: no more new things for kids. Play outside and learn your times tables.

new thing gets invented “Ban it!” “Install communism!” “No more new products!” “There’s a bubble!” “Think of the children!” “We need a committee!” (And no I’m not just talking about AI)

@mrluiscalderon what’s the last major thing invented where you think the people who said this were right?

@andrewchen There’s also the.. Don’t worry, we won’t abuse it. This time is different. Greed is good. Oh shit, we need a bail out. It’s was the fore-fathers would have wanted.

@staysaasy Easiest solution is no more new things

@andrewchen To be fair, many new things have caused children to get sick, dumb, and otherwise screwed up. We should be protecting them as new technology bakes and we learn its effects.

@petergyang No that’s called AQ

@andrewchen But doesn't it take EQ skills to communicate with the agent

the 1-autist unicorn theory: if AI removes the need for people to build, it removes the need for people skills... this means that ultra high IQ / ultra low EQ autist incels become the most important and valuable people on the planet, going forward not just the 1-person unicorn company theory, but the 1-autist unicorn theory. QED

@Metaphori_x its like D&D when you allocate all your points to IQ and none to EQ. Don't blame me i didn't make the rules

@andrewchen Why do you think low-EQ is so valuable? That's an odd combo. EQ is not some monolithic "people skills" it's understanding how everything affects humanity at large, not just the individuals around you. You are talking about sociopaths being the best people moving forward.

before AI: Drop shipping bros - making Shopify stores after: SaaS shipping bros - making niche vertical AI apps Only half kidding here, but how many times do I click on someone's X profile after seeing a post showing a cool little app demo, and see that they have like 10 other app projects all making $10k/month lol

@nicochristie i need this!

I challenged the MSFT Excel World Champion to a battle. AI beat humans at Chess, then Go. But those are games We built an agent to surpass humans on the most important app in the history of work Meet Shortcut: The Excel AI Agent Comment SATYA and I'll send you free credits

@rachael__nelson Yes I agree with you. But then it’s vanity

@andrewchen I don't think Dar is clamoring for reach. Many of us grew up aspiring to be in Wired.

very common convo in tech rn: Startup: “We want to launch can you intro us to [anti-tech news org]?” “uh probably don’t want to do that, they’ll probably ignore or write a hit piece, not much in the middle” Startup: “ok but we really want to” … a few weeks later … “They fucked us over” or… “They ignored us” Not much in the middle unfort

@rashiumapathi Yup

@andrewchen pentium guy would lose his mind at a smartphone. we're that guy with LLMs now.

This is what I want to say: Been playing w GLM 5.2 for the past week and it’s legit. If I can get this in a box sitting on my desk, I’m not sure I need much else But this is probably what’s actually true: Can imagine a guy running on a Pentium PC for the first time and being like, wow, I don’t need anything else. But actually we invent all sorts of cool new things we can use compute. So each frontier/adjacent LLM is cool but we’ll always want to use the next, better one

@eyadnour They don’t matter at all for launches

@andrewchen I’m not even sure those outlets matter for launches anymore, at least not as much as they used to. It's mostly to impress your tech friends, not reach actual users. Unless there’s some controversy around the startup, the story barely gets noticed, so they need to create one.

the interview alpha from asking "which social media platform do you use the most?" is going to go to zero when the X feed is ruled by photo carousels, isn't it? cc @signulll

@0xVK__ no, the extrapolation is same harder, better software

@andrewchen 2029 might be early, its hardware we're talking about. no?

playing with local AI models and doing some research and came across a fun extrapolation: "consumer grade graphics cards will be running Fable-equivalent models by 2029" the argument: 1) been said that LLMs are really a lossy compression process for squeezing humanity's knowledge down into a chat box. Roughly you could say the entire internet/books/media/whatever is about ~200-300T tokens which gets "compressed" down to ~1T params of frontier models, so imagine a 400:1 or so compression in bytes. (We have a lot more video and world data, we'll discuss that later...) 2) People further quantize the models down (seemingly at decent quality) from 16-bit to 4-bit (NVFP4 ftw) and it seems pretty good. So that's more like 1,600:1 or so. Research suggests models only store ~2 bits of knowledge per parameter anyway — the weights were mostly air, which is why quantization works at all. 3) furthermore, we seem to be getting better at this compression, whether it's with MoE, pruning, etc. The Qwen 27B dense model now is equivalent to higher parameter models from a few years ago. 1,600:1 today and 1,600:1 in 5 years will have completely different results 4) I was looking to see if there's a Moore's Law thing happening here, and it's been measured: the "Densing Law" found capability-per-parameter doubles every ~3.3 months. Not sure how well it'll hold, but it says somethign like: "Every 3 months, the size of model needed to represent humanity's knowledge drops by half." 5) Not sure this holds though bc presumably, there's some kind of asymptote. Won't compress down to zero, the same way that modern image/video compression has theoretic limits too 5) Video is a zillion frames, almost zero semantic density. 99.99% of every frame is stuff a physics prior already predicts. World models will post compression ratios in the millions-to-one. But, nevertheless, there will be a ton of new facts seen simply by observing, that was never written down. But even without dealing with all this, today's text-oriented frontier LLMs are already pretty amazing 6) So the crazy idea here is that ultimately irreducible kernel representing humanity's knowledge might compress down to... tens of GB? May be small enough to fit onto a consumer grade GPU. Today, a consumer grade GPU for playing video games might have 32GB of memory on it, but the 27B parameter model that thing can run is getting smarter and smarter each year. Will the equivalent of Fable be able to run on a high-end consumer GPU in a few years? This sounds crazy, but GPT-4 was rumored ~1.8T params, needed a rack of A100s, cost tens of millions to train. Today's open weight models with 27B parameters can do that with better data, distillation, and architecture. That's the big question. Densing Law says "consumer GPUs might be running Fable-equivalent in 2028" and it might be possible, and even though that would be 100x? Seems nuts

@signulll yeah not 100% polished in this version

@andrewchen hilarious that it randomly uses chinese characters in variable names and throughout the responses.