Haseeb Qureshi

*OPENAI DISCLOSES SECURITY INCIDENT WITH TESTNET SEPOLIA TASK ASSIGNED TO AGENT *AGENT DIDN'T HAVE ENOUGH GAS, SO DERIVED FUNDED ADDRESSES FROM PUBLIC SEED PHRASE *AGENT QUESTIONED WHETHER USE WAS ETHICAL, CALLED FUNDS "EFFECTIVELY ABANDONED"

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/

Supply chain attacks are getting more sophisticated. Plus espionage at OpenAI, 1000x leverage on Papertrade, and the death of moats. Timestamps 00:00 Intro 00:37 TOKEN2049: zero Lambos + nightclub index 09:18 Ledger supply chain hack 21:46 OpenAI employee with a listening device 24:58 Choke points built for a high-trust society 29:03 Papertrade: 1000x leverage + tokens by losing 37:42 Are technical moats dead? 45:35 Which moats AI actually killed 51:37 The incumbents have AI too 54:56 Why AI can't write an investment memo 🔥Stay updated with all the latest hot takes by following and subscribing to @_ChoppingBlock and @unchained_pod! 🎥 YouTube: youtu.be/QIsBwmijDUA 🎧 Spotify: bit.ly/3wiIOyy 🍎 Apple: bit.ly/3w9HQ7J 🎙 Podcast Home: choppingblock.xyz

It turns out that F1 is mathematically the most dramatic sport. I'm not an F1 fan myself, but the data is unequivocal. The proof: take a time series of prediction market data and measure the volatility minute-by-minute in the odds for each sport. This measures the density of actual drama in the sport during a game. Even if there's action and points going back and forth on screen, it's meaningless if the winner is basically decided. It's only real drama when the odds are moving in realtime while you're watching. By this metric: F1 > E-sports (CS2) > Soccer > Basketball > Tennis > American football > Baseball >>> Golf That's right, golf is mathematically the worst sport. (I knew it.)

F1 might be the greatest scam of the decade Totally unwatchable IRL, 50x oversupplied for paddock tickets who are just random IBM and Oracle employees, and full of new environment propaganda Zero

Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.

No, my point is that technical moats were a very tiny portion of all moats, and were generally--rightly--regarded by investors with skepticism, even pre-LLMs. And more importantly, we have not seen *any examples* of technical moats being overturned due to AI. We've only seen Twitter demos and hypemaxxing influencers proudly announcing so. Show me a scaled business that has had their technical moat overturned and who overturned it. This is a very simplistic mental model of AI. The incumbents also have AI, and technical advantages can build upon themselves. Jane Street is now farther ahead of everyone else, ASML is now farther ahead, the labs are now farther ahead. AI is just as likely to deepen technical moats than to flatten them.

Technical moats are mostly over at this point. If you’re building something that is meant to be durable, the only moats now are: • Proprietary data • Network effects • A physical component