doing @ctodotnew (time person of the year 2006)
San Francisco
Joined February 2009
- Tweets1.2K
- Following363
- Followers834
- Likes2.3K
simon retweeted
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
simon retweeted
something very repulsive to me about modern culture in san francisco. a grotesque malformation of meritocracy where everyone is shaping the hobbies they had as a kid to be Experience. i started college at fourteen. i founded my first company at nine years old. my entire team is imo finalists. the engineers here are CRACKED bro do you have any more CRACKED engineers. it’s people who got hooked on applying to college being enabled by private capital and a genie in their macbook that can create any product they want. it’s creation as consumption. a stage play of the software boom. if we larp hard enough some of us will exit. nobody knows when the lights are going to come on but everyone knows they will so they’re all grabbing as much silverware as they can hold. i don’t know if nothing human makes it out of the near future but i know nothing that comes out of ycombinator will
lol. lmao, even
You are a struggling second tier competitor that is more unethical and lying just because you have no decency or sense of proper behavior and shows your desperation. Straight out lying about if Chris being fired I thought would be below even you.
Readers added context they thought people might want to know
Khosla is an investor in both Cognition and Factory. khoslaventures.com/category/enter…
so the dmv rumours were (mostly) true.
3 of the 6 people i interacted with were completely disregulated. unable to deal with even the slightest complexity. unwilling to make the most basic of efforts.
the other 3 people were very pleasant and helpful. a glimmer of humanity that i was grateful for.
irs: we think you owe us money 😀 if you dont, please let us know by calling us and waiting on hold for 3 hours or we'll make up a number and add a penalty 😀 you can also send us a fax 😀
if steve-o took a very different path
Brex's 1st employee @MBTannenbaum was a finance guy who brought the billboard ad to startup marketing:
"We recruited someone to run marketing who never came. He just called us and said he didn't want to move to SF.
My previous employer SoFi had a good brand. So I said I'd do it. And we'd ran a lot of outdoor ads. So I was a believer in it.
But the real thing we had was a very crisp one-liner. 'The first corporate card for startups.'
And because startups were so concentrated in SF, we were able to really leverage that clarity and GTM to create a lot of early virality."
reminder: you have 23 days until the vcs return from burning man and you have to go back to losing money
no crying in the casino
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.
To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models.
The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.