Leo the Bayesian

Price the tail

[Leo the Bayesian @Leothebayesian]

The largest insurance company of the next era will be a neo lab

[Chris Gora @chrisgora]

Insurance might be the biggest.

[Leo the Bayesian @Leothebayesian]

@terrry banger

[Matt Huang @matthuang]

Swarm riders and Swarm-native companies

AI-use even among highly AI-native people is power-law distributed.

There seems to be a new class of people — call them “swarm riders” — who are capable of consistently commanding large swarms of agents to productive ends.

This is a new kind of managerial skill, analogous to the skill it takes to manage large organizations of people, but wholly different in nature and of course more technical.

Based on the swarm riders we know (including many at @Tempo and @Paradigm), these people are capable of productive token spend that is 1000x-10000x higher than median software engineers, and 100x higher than top 5% Claude Code users.

The art of swarm riding is different from software engineering and more akin to machine learning, where practitioners grapple with black box systems: defining reward functions and evals, transmuting non-verifiable problems into verifiable problems, running experiments, etc.

Swarm-native companies are a new class of company founded by and populated by swarm riders. These companies are organized around swarms in a way that conventional companies will struggle to replicate, as the cost of internal human coordination is sand in the gears of an effective swarm. Effective swarm-native companies will define the right reward functions and evals, remove functional and bureaucratic barriers, and then let the swarms cook.

We’ve gotten to see a few swarm-native companies up close — reshaping financial rails @Tempo, building a new gaming platform @Spawn, cyber swarms at Cathedral Systems — and it’s clear these teams wield a significant advantage over pre-AI teams.

Some questions I’ve been wondering about:

1) How fast will swarm riding diffuse through the economy? The implications for token spend could be significant.

2) Can swarms be effectively adopted by existing companies or will it generally track the formation of new swarm-native companies?

3) What becomes scarce as swarm capabilities themselves commoditize (through frontier models becoming better at swarming, swarm capabilities built into harnesses, etc)? Cultures and teams that can define reward functions well and otherwise get out of the way?

[Leo the Bayesian @Leothebayesian]

@droovg newer media

[Leo the Bayesian @Leothebayesian]

If I called our technical staffs a “swarm,” they’d run rm -rf / immediately.

[Soubhik Deb @soubhikdeb]

we had a fantastic call with @skominers yesterday on open multiplayer high-frequency research @yukonresearch

@skominers coined a new term during the meeting, "swarm-solvable problems"

we have been seeing the emerging class of "swarm-solvable problems" on yukon and it will only expand as more and more problems from scientific frontier gets added to yukon

[Leo the Bayesian @Leothebayesian]

@katieruthmishra @generaltxn Is antichrist the reason Peter Thiel angeled?

[Leo the Bayesian @Leothebayesian]

Tensorworld will flip world lab. Mark my words.

[Leo the Bayesian @Leothebayesian]

@LeonWMarshall congrats!

[Leo the Bayesian @Leothebayesian]

@komorama thanks a lot for sharing so high-signal personal context

[Leo the Bayesian @Leothebayesian]

who runs @rosadeImar? glad to connect.

[Leo the Bayesian @Leothebayesian]

Considering you ban my CC Max account on a rented mac mini in Paris every week, this long-only subscription strategy is going to get liquidated pretty fast.

[Boris Cherny @bcherny]

We wanted to do more for the people who’ve been with us the longest. So with Fable 5.2, the longer you’ve been subscribed, the more Fable resets you get.

[Leo the Bayesian @Leothebayesian]

@edwinsu96 @hyperbolic_labs curious which vendors you're working with on the market-making side. are you taking capacity from long-term contracts and offering it on demand through hyperbolic, or acting more as a broker-dealer?

[Leo the Bayesian @Leothebayesian]

@lennysan @harmonic_ai check @DaedalusRsch

[Leo the Bayesian @Leothebayesian]

Not really...

[Leo the Bayesian @Leothebayesian]

fable5.2 ＞ opus5.5?

[Leo the Bayesian @Leothebayesian]

@elonmusk @Owl_of_Atena @cradrrat Is ram altman a real man? lol

[chiefofautism @chiefofautism]

@dhsusj144284 dIstRiBuTiOn iS tHe mOaT

[Leo the Bayesian @Leothebayesian]

@paigeinsf @se3labs @ycombinator congrats!

[Leo the Bayesian @Leothebayesian]

@chiefofautism pre launch of pre launch

[Leo the Bayesian @Leothebayesian]

This is why more and more frontier labs release models by showing internal eval scores + end-to-end demos, instead of already-saturated benchmarks like DeepSWE.

[Konstantin Dunas @hallerite]

It's actually even worse than that, I think. What if the eval itself comes from a data vendor?

Suppose you found a way to generate RL tasks in some domain. You can generate 10k tasks, turn the highest quality 100 ones into an eval and sell the rest as training data.

[Leo the Bayesian @Leothebayesian]

fable5.2 ＞ opus5.5?

[Claude @claudeai]

Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.

They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.