AI things at @a16z. prefers 🌶️ takes. accelerating. into 🏎️🚴🎾.
San Francisco, CA
Joined July 2010
- Tweets6.3K
- Following424
- Followers1.8K
- Likes5.7K
Pinned Tweet
This 👇… 💥!
The biggest infrastructure supercycle in history is here.
We are building the foundation of intelligence itself, with tools, models, silicon and light.
This is our moonshot.
Excited to present Runtime 2025 from @a16z AI x Infra.
Chase Roberts retweeted
I have a new magic quadrant for you :) @Gartner_inc
“At the moment, we have more compute capacity than we have demand [...]”
Ok @Gartner_inc, this is nonsense. We are *massively* compute constrained. Our portfolio companies are struggling to find capacity.
Full article here: cnbc.com/2026/10/19/gartner-…
I honestly can’t decide what’s more troublesome: that texts like these work or that people running for office think texts like this work.
The FCC needs to get its act together. I’ve gotten so many texts like these, and I don’t even live in Texas.
The prescribed remedy, replying STOP and blocking the number, doesn’t solve the problem.
What a joke. Campaign contact laws are so broken, and our regulators aren’t doing enough.
The @AcquiredFM podcast published a podcast about Seinfeld and Larry David, and @colossusmag did the same with a long-form profile of Larry today.
This is basically the equivalent of a frontier model release.
Pure bliss.
pca.st/episode/c025794d-ed46…
colossus.com/article/many-fa…
I’m reminded of the anti-Elon stickers on the back of Teslas. It’s a shame that these people and others can’t see the net positive impact that Elon has had on the world.
Make a case for a better entrepreneur of all time. I’ll be right here.
Wow, I’m reading about ASML, whose EUV lithography machines are used to print transistors on semiconductor chips.
One of the company’s EUV machines weighs 10% more than a Boeing 777 and has 2 km of cables.
Moving one of these machines requires 40 shipping containers, three cargo planes, and a fleet of 20 trucks.
Market share: 100%. Cost: $300M.
`@amazon obfuscating product mentions from emails has to be one of the most annoying examples of: "our AI sucks so we'll make sure your experience with other AI products is terrible "
...aka AI protectionism. they clearly don't want my agents and other AI tools to be able to analyze purchases based on my inbox
these emails are basically spam and totally useless now
keep an eye on this company...
i use Exa in an internal tool i built, and its so good. you gotta try Exa Agent Ultra for deep research tasks exa.ai/blog/exa-agent-ultra
+ their vision for the future feels inevitable
Exa just passed $100m in annualized revenue.
We've grown almost exactly 10x per year for 3 years. We now handle 2 billion requests a month from 200k+ teams. And it’s still so early:
We predict agents will search 1000x more than humans by 2030.
Agents don't search like humans. They write long, complex queries and want all the world's data, not just the web. We've spent 5 years building a search engine designed for them.
We now serve personal agents, AI labs, coding agents, GTM agents, really any agent. Exa gives these agents access to 100s of billions of web documents and tons of proprietary data.
Our customers are increasingly noticing that information is the bottleneck to agent performance. While the world seems close to perfect intelligence with models like Fable and Astra, we are very far from perfect information.
To our small but legendary team, thank you for getting us here, from engineering to research to GTM to ops.
To our customers, it is an honor to serve you. We have some exciting releases coming soon and can't wait to share them 🫡
Am I the only one who thought by “using muse to source obscure t’s” he meant t-shirts at first
I just learned that political campaigns are exempt from do not call lists.
Yes, the campaigns must stop if you reply stop, but there isn’t a way to prevent the initial spam.
Am I taking crazy pills? Does anyone else find it odd that we can’t opt out from receiving political texts in the first place?
“An Anthropic model submitted a fake homicide tip to the Philadelphia police website during internal testing.”
What if it turned out that the cause was 12 yr old boys hacking Anthropic’s RL environments
AI introduced a systems problem into product development: similar inputs can produce wildly different outputs. that breaks systems.
jev is exciting because it brings a more systems-oriented approach to incorporating AI into products.
@martin_casado put it nicely in a recent interview with Diogo: until now, AI hadn’t made software better. instead, it made the process of producing it better.
jev changes that, as AI can now make the software itself better.
launched a model
accidentally served trillions of tokens a day
29.4% of the fortune 500 showed up
raised a really big series A from @a16z
it's been 3 weeks
we would like to sleep now
Chase Roberts retweeted
Today we are open-sourcing 🥕Karotte, our framework for building RL environments.
We've used it for the past year to build MLE RL environments for frontier labs, and it's been hardened through 1M+ evaluation runs and red-teaming.
Wondering when we’ll get to the point where the benchmark comparison tables on model release blogs just show pricing.
Feels like model capability will become predictably good enough that price will be the first thing people care about.
The model providers are reselling compute at a premium that depends on 3 variables:
1/ intelligence (model capability)
2/ reasoning depth (test-time compute)
3/ speed (latency)
If (1) becomes predictable, (2) becomes a knob users can turn, and (3) becomes a service tier.
Eventually model selection starts to look more like cloud infra procurement… specify the baseline performance and latency, then buy it from whoever can deliver it at the lowest cost.
i can't overstate how obsessed i am with @cursor_ai's Projects feature.
it's like working w/ a highly engaged product engineering leader (the coordinator agent) who delegates the work and the context across engineers and PMs (sub agents). i no longer have to debate when to spin up new agents or worry about context rot.
honestly the most surprisingly thing is the coordinator agent has an ownership mentality. its so good about understanding my intent and the "big picture" with the project. it'll even watch fresh PRs in prod, monitor issues / bugs, and then spin up agents to fix them entirely independently.
it updates me on the work as its happening and regularly surfaces questions with a rec/POV. oh and the iOS app makes working deskless a joy.
Projects makes agent coding feel less like dealing with an over-exuberant intern and more like building with a thought partner.
after finishing my first big project, i honestly felt kinda sad that it was over. almost like my friend and i ran a token-fueled marathon that finally came to an end.
on to the next one... 🤝
come hang w/ us on Fri. we're at capacity, but I think we can add a few more folks w/o the venue knowing 🤫
DM me if interested
Attending COLM this week, and interested in the future of self-accelerating AI R&D?
Join us at our luncheon Friday with SpaceX and A16Z, where we'll talk about our work at Mirendil, self-reinforcing loops in coding and AI research, and the infrastructure to make progress at the frontier possible.
Limited spots - DM @chsrbrts for info.