Bryan

tech and other cool things

[Bryan @byintes_]

@uhmclaire jester maxxing

[Bryan @byintes_]

if youre new to the open source:

kimi = best at pre train
glm = best at post train
deepseek = best at optimizations

[Bryan @byintes_]

@uhmclaire spoken like a true New Yorker

[Bryan @byintes_]

citadel securities raised from sequoia too btw it’s completely sensible to go to YC even with some crazy alpha

[Michael Wang @michaelwangelo]

hey, Michael from Prodigy Research (YC S26) again

Our launch has made the rounds and produced a lot of discourse. Much of it is around a few particular questions: if you have alpha, why raise money at all? and why do YC in particular? I’ll tackle these in order.

First, why raise money at all if you have alpha?

The short answer is that we are raising money for the same reason any business raises money.

There is a lot of what I would call “magical thinking” about trading. Many people have a mental model where a profitable trading strategy is like a magic faucet. You turn it on and unlimited money comes out, simple as. If this were the case, it would be irrational for the owner of such a faucet to raise money. If more money is needed, simply get it out of the faucet.

But in fact, it takes capital to convert alpha to dollars! Prodigy is training models and building the machine that generates alpha. Capital allows us to convert alpha to dollars faster and more efficiently, do bigger and better training runs with more compute, conduct more model inference and research, build infrastructure that allows us to do more demanding kinds of trades, and generally grow the pie faster in a win-win way.

So we could recast this question as: “if you have [the core of a really profitable business], why raise money?”

You will notice that this is an argument against raising money in general. A business can be profitable today (as Prodigy is) and yet have extraordinarily high-return opportunities in which to invest marginal dollars. If a dollar spent on compute today can create far more than one dollar of long-term business value, then it is much more rational to raise that dollar than to deliberately grow more slowly to avoid fundraising.

That brings us to our second question: why raise from YC in particular?

Prodigy can build and grow much faster with more compute, and raising equity capital gives us the flexibility to do that. We could have raised outside of YC, but YC offers us things like access to incredibly ambitious founders, partners who know an enormous amount about building companies, and many other resources from which we have already derived a huge amount of value.

We’re heads-down building right now, and we’re super excited about scaling quant trading beyond human limits.

[email protected]

@realProdigyCorp

[Bryan @byintes_]

judging by the spike of Claude ads it seems like the rumors of suddenly slowing growth might be true?

[Lawrence Chen @lawrencecchen]

It’s genuinely insane to see cmux growing via word of mouth

I go to events in SF and people are using software that I built!? I 100% need to be doing more marketing, but even so, cmux is still growing

cmux is almost at 100k DAUs!

This is just nuts

brew install --cask cmux

[Bryan @byintes_]

kimi always has a nice big model smell, makes it the best open model to post-train on

[giatt @1owroller]

girlfriend overheard me and started yelling at me asking me “who the fuck is kimmy” “why the fuck are you talking about chinese models”

[Bryan @byintes_]

am p sure that SSI is actually a quant shop

[🍓🍓🍓 @iruletheworldmo]

‼️BREAKING

ssi are set to release a model superior to fable across the board.

amid growing investor pressure ilya’s hand has been forced to release.

expect a release at the earliest this month but most likely mid august.

things are hotting up indeed.

[Tony @0xRacist]

Let him cook

[Bryan @byintes_]

There is no better marketing for Harvey/Legora, than Kirkland trying to build their own Harvey/Legora for $500m

[FleetingBits @fleetingbits]

some thoughts on kirkland building its own harvey

1) kirkland is spending $500m over four years in order to build its own internal ai legal tools; kirkland intends to spend $100m this year

2) i suspect that kirkland is doing this because they have told themselves that they have valuable data and because they want to appear differentiated

3) i think the first issue is that kirkland probably does not have differentiated data from other elite law firms; at least, not at the level a harvey would absorb

4) all the elite firms probably have similar internal workflow data and so long as some of them defect, that is enough to commoditize the data kirkland wants to use for its platform

5) and, to the extent that they do have different internal workflows, harvey and legora will end up representing a better version of them and this will put kirkland at a disadvantage

6) moreover, companies like kirkland will have difficulty building their internal legal platforms because they do not have experience with software development

7) and, there are both cultural and structural issues with them managing software developers, like they cannot give non-lawyers equity in the firm due to regulation

8) so, i think firms like kirkland are better off using tools like harvey and legora and then looking to focus on where their value really is now: client relationships, local knowledge (litigation, regulation) and legal r&d (novel structures, etc...)

9) anyway, this seems to me like a phenomenon that ai creates across a lot of industries, where firms that were previously vertically integrated become unbundled due to ai because part of the intelligence gets moved to the labs or otherwise gets commoditized

10) and so, a new set of companies are created whose job it is in order to provide services complementary to the labs: forward deployed like harvey and legora and data providers like mercor, surge and handshake

[Bryan @byintes_]

they’re going to reduce rate limits aren’t they…

[Sam Altman @sama]

come for the rate limits, stay for the best model

[Bryan @byintes_]

we’ll soon learn if open source is 6 months behind sota or if it takes 6 months to distill sota

[Bryan @byintes_]

elliot management

[litquidity @litcapital]

“Do you work in good finance or bad finance?”

[Bryan @byintes_]

using Claude to clean up my desktop files is great but can I use Claude to clean up my Claude chats please

[Bryan @byintes_]

In absence of more details coming out, seems like smart financial engineering to get an edge to lock in the largest mega fund platforms despite what everyone says.

There’re maybe 20-30 of such platforms - locking them into a JV, getting them to deploy capital (some of which will surely go towards hiring implementation FDEs!) is a very nifty way of getting even more leverage. It’s like off balance sheet financing is moving from CAPEX to OPEX!

Would put money on the guaranteed returns being equity kickers on qualifying exit rather than being underwritten by OpenAI balance sheet though.

Everyone involved highly sophisticated and knows what they’re getting into.

[Andrew Curran @AndrewCurran_]

OpenAl is offering private-equity firms a guaranteed minimum return of 17.5%, as well as early access to models not yet in public release.

[Bryan @byintes_]

@kimmonismus actual reason why this happens is that chain of thought isn’t passed as context into the next turn

[Bryan @byintes_]

@Entschul1999 @itsandrewgao But if that is the case wouldn’t the qn be “after the 100th flip”? Now I’m probably as confused as Claude is

[Bryan @byintes_]

@tszzl @PradyuPrasad give us normies fast codex too please