I agree and I think your kitchen analogy is apt. I worked in two restaurants when I was in my late teens/early 20s: one a chain, one French fine dining. With one clear exception, nothing in trading has ever approached the level of stress a kitchen has. The Bear was accurate.

This is fucking awesome

Postpartum meal gifts are nice but nobody ever brings veggies. I've got lasagnas from three different households

pediatrician told me my 10 day old baby is back to her birth weight

Because AI is leverage. Levent is an expert mathematician (just look at his CV - it's impressive even by math standards). AI enables one expert to work far faster than even a complete team of experts. It's like managing a brilliant team that doesn't encounter the Dunbar number.

These places haven't been primarily market makers for years. Jane Street is a vertically integrated directional risk taking machine.

A game's EV is not sufficient information to decide on playing it or not: - is the game single round, or iterated? - if iterated, can you quit at any time? - can you change the bet size (especially % of wealth)? Rational agents can decline to play this game.

Another hangup I'm seeing: why did Jane Street have long momentum exposure? Not because they're yoloing a factor. Because many stat arb strategies organically accrue momentum exposure and residualizing it out would reduce risk-adjusted returns. A stylized example with no IP since it's in the literature: your signal is xsection(analyst earnings revision). Long equities with upside revision acceleration, short equities with downside revision acceleration. This is _not_ a bet on price continuation/trend. However! Analysts empirically cluster around news that has already moved the price. The legitimate fundamental information seeding the cross-section has a latent loading on momentum, via human behavior. If you try to orthogonalize this signal to momentum, you linearly downweight the equities in the cross-section which report the highest analyst conviction, because that conviction is mechanically reflexive on momentum, even though it is a qualitatively distinct source of information! Consequently, your residualization will flatten the distribution: post-resid top quintile of the xsection has shifted names with weaker signal from the middle three quintiles, and likewise for the bottom quintile. Realistic numbers - you squeeze a 1.0 Sharpe out of analyst revisions, 8% annualized on 8% vol. You residualize out momentum, your vol does drop because you took out the momentum risk. Maybe it goes to 6%. But oops - because of the distribution smearing effect, your returns have fallen faster, to around 5%, and now you've got like a 0.85 Sharpe. Jane Street are not idiots. They understand this intimately well, and far better than I'm even describing here. They're still up $40B net revenue YTD after losing $15B because they rationally accept the higher risk adjusted returns that keep in the factor prone to sharp drawdowns.

Suppose you are a market maker with 50% share of the entire US equities market. 250 trading days, $600B notional traded per day, $160T annual, you deal half that and capture the volume-weighted average spread (currently ~5 bps, dragged up by liquid megacap). Assume 0 adverse selection and that you take half the spread in your market on every trade. That's just enough to land you at $20B. But no one is 50% of the market, and adverse selection cuts down that spread capture.

What I'm learning is that most people outside of these firms still somehow think Jane Street, HRT, etc are primarily market makers. There are very few firms specialized in market making and they're not putting up $20B+ from it in net trading revenue (it's effectively impossible)

> The firm has generated more than $40B in net trading revenues in the year to Friday, even accounting for the July loss

Let's cut the gordian knot here. You're saying "never a down week" in marketing materials. Either your model's strategy is so capacity constrained it's unserious as a demo (e.g. trading 5 figures fading shitcoin momentum), or your model can get you $500k without giving up equity.

I can't quite explain it, but this feels like a late night Adult Swim commercial break. It has a certain surrealist tone to it. It's almost satirical

Who will make progress on Riemann first? - Anthropic's staff mathematician; Harvard undergrad; valedictorian; Erdos 2 in undergrad; Morgan prize; Princeton PhD; Society of Fellows; big time problem solver - a guy going for a jog asking it to "take a real stab at it"

People are unfairly criticizing this. Yes it's very technical and not very soulful, but Mancuso wrote it as a study piece for teaching. It shouldn't be compared to SRV's Pride and Joy or BB King's Thrill is Gone, nor does it mean Mancuso can't play like that.

Open weight models are predictably forcing a commoditize-your-complements dynamic on harness engineering. Claude Code is closed source; Codex is encrypting/obfuscating increasing portions of its prompts. The IP is moving from inference to the agent ecosystem.

It is very funny to ask Fable to verify this, instruct it not to search the web (by now it will find this tweet), and watch it spiral out checking and re-checking because it can't believe someone handed it a valid counterexample to a 142 year old conjecture.