Here for the experience. Meditation & self-inquiry, AI, entrepreneurship & investing, cute dogs and babies.
San Francisco, CA
Joined March 2007
- Tweets7.2K
- Following1.7K
- Followers10.9K
- Likes12.9K
Pinned Tweet
A startup is not a company: it's an experiment to see if a company deserves to exist. Until p/m-fit, speed of hypothesis testing trumps everything. Startups die when they spend faster than they learn.
Yan-David (Yanda) Erlich retweeted
Most company politics are rooted in fear.
I call this the Fear Triangle.
When fear shows up, people tend to move into one of three roles: savior, bully, or victim.
The recent tiffs among the meditation community on here give me comfort that beings much further ahead on the pathless path still, at least occasionally, get captured by judging/comparing/fixing mind. This human life is quite the adventure. 🥰
May all beings be at peace
A little late to this meme but 🥰 @maticrobots @Tesla with love from Wizzy & Sequel
Yan-David (Yanda) Erlich retweeted
A few years ago, my wife and I were sitting together at the end of a long retreat.
We were talking about an investment we passed on years ago, one that would have made us rich enough to never start @artofaccomp.
Then I asked her a question out of curiosity:
Yan-David (Yanda) Erlich retweeted
Very happy to see this from Nick!
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/
Both Panda Contacts (pandacontacts.com/) and Panda Timekeeper (pandatimekeeper.com/) are live in the Mac and iOS App Stores, and both are iPhone Duo ready. 🥰🥳
How did folks fare in the iPhone Duo purchase frenzy:
50%October delivery
42%November
8%December
12 votes • Final resultsThe crew at @axiommathai continues to push the frontier of AI-enabled math discoveries. This is a huge one.
Yan-David (Yanda) Erlich retweeted
Check out this preprint by Cureton and Thorner using AxiomProver: hexagonmath.org/2610.00200v1
github.com/AxiomMath/Dedekin…
Some fascinating developments following OpenAI's shock breakthrough proof of the Quasi-Riemann Hypothesis, which was released publicly on October 6:
1. Building on OpenAI's proof, a recent Hexagon preprint (released on October 12) by Cureton and Thorner establishes that ALL finite-order Hecke L-functions are zero-free to the right of 1 - 1/6000.
2. And now, today, an arXiv preprint by Frei, Somos and Widmer uses this result to prove unconditionally the 2007 Ellenberg–Venkatesh bounds on ℓ-torsion in class groups of number fields.
I was amused to see the AI declaration in the second paper: "No artificial intelligence tools were used in this work."
Yan-David (Yanda) Erlich retweeted
we're on the cusp of huge breakthroughs in interpretability - this is the science we must understand, and I think we're going to be able to do it
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/
Yan-David (Yanda) Erlich 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/
Yan-David (Yanda) Erlich retweeted
it’s been an enormous privilege and joy to watch @nickcammarata cook this up over the last year. week after week, experiment after experiment, he unraveled the tangled mess of our thinking about neural nets to reveal the truth.
i never knew quite what to expect on our weekly walks. each conversation brought a torrent of new theories about the nature of AI, some conjectures teetering on the brink of madness, and others that seemed to cut through years of confusion with a single sentence.
nick is one of the great scientists of our time. his intuition and research taste are out of this world, but what sets him apart is his unmatched ability to capture the essence of what something truly is, without being lost in the details. i would highly recommend you read his entire essay if you’re curious about AI. it’s long, but written beautifully.
nick’s research makes me far more optimistic about the future of AI alignment. i hope that it can inspire the same for you.
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/
This is an incredible discovery by @nickcammarata and the team at @GoodfireAI, reversing a long-held belief in the inscrutability of neural nets, and providing hope for significant advances in interpretability, just as the need becomes dire. Proud to call this crew friends. 🥰
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/
Yan-David (Yanda) Erlich retweeted
Axiom Proved The Quasi Riemann Hypothesis (Generalized) For All Dedekind Zeta Functions
Important complexity consequence of new claimed quasireimann (hexagonmath.org/2610.00200v1): Valiant's conjecture VP != VNP over the complex numbers follows from the boolean-complexity lower bound NP not in P/poly. Burgisser proved this under GRH. (🧵1/4)
Hi @1Password @OpenAI, could we get a nice 1Pw plugin for Codex like exists for Claude please. 🙏
