rainbohtrout

Trading, venture, love of the game

[rainbohtrout @rainbohtrout]

@macroarb Real recognize real

[rainbohtrout @rainbohtrout]

@grok @ryandengg @a16z @khoslaventures @jclozano9897 @MarianaMinerals @tbc415 @baker_tilney @grok what are top 5 insights re AI applied to mining founder anecdotally addresses…that we wouldn’t otherwise know until this pod. what was pushback from VCs too before the raise?

[rainbohtrout @rainbohtrout]

@lefttailguy And the individual’s tech/systematized process + data acq edge will become beta in the fullness of time as the platforms scale

[rainbohtrout @rainbohtrout]

@lefttailguy Podification of the world. Incentives for privates seems hard to solve since most of the value is in being the deal guy. Think equilibrium will just be less % pnl sleeve relative to publics. Plug your deal guy alpha + domain expertise to LLH for 10% over years over KPI hurdles

[rainbohtrout @rainbohtrout]

@FundamentEdge Everything happens all at once

[rainbohtrout @rainbohtrout]

Kind of beautiful how all the labs set coding as a benchmark, and excel was an afterthought, but they’ve done such a good job at the former that the latter will be solved as a byproduct in <1Y

[Lilian Weng @lilianweng]

A super long overdue (3+ years?) post on scaling laws.

Compute is expensive. Scaling laws are a way to help us reason about the optimal compute allocation between data and model size before committing to a large run.

The post covers what scaling laws predict, how compute-optimal allocation works, why Kaplan et al. and Chinchilla disagree, and how data limits + fitting details make extrapolation tricky.

https://t.co/HP26eJvjHB

[rainbohtrout @rainbohtrout]

And for public’s this is great too since bottleneck trade has gotten boring…not arguing that constraints won’t persistent but the GLM moment implies we’ll get a shock to demand curves where enterprises start optimizing for unit intelligence per $ compute without being punished

[rainbohtrout @rainbohtrout]

Taking the below together, landscape starts to look like the too early ‘23/24 takes that frontier capabilities will commoditize at an affordable rate…what startup bets should one make here?

[rainbohtrout @rainbohtrout]

So 1. inference is the future for CY26+ 2. open weight models lag frontier labs by ~6M (GLM vs Opus) 3. open weight models are a fraction of the cost (prefill/decode stats) 4. open weight capabilities converging to frontier (Baseten GLM-5.2) 5. mag7 GTO play is to kingmake OSS

[rainbohtrout @rainbohtrout]

So 1. inference is the future for CY26+ 2. open weight models lag frontier labs by ~6M (GLM vs Opus) 3. open weight models are a fraction of the cost (prefill/decode stats) 4. open weight capabilities converging to frontier (Baseten GLM-5.2) 5. mag7 GTO play is to kingmake OSS

[Rittenhouse Research @RHouseResearch]

Baseten CEO @tuhinone tells Altimeter's @apoorv03 that one of Baseten's cloud providers has already indicated their B200 prices ($/GPU hour) are set to double when existing contracts expire and are up for renewal later this year.

"If you go out right now saying you want a thousand GPUs, truly.. people are talking about Q2 of next year. So 12 months out, maybe 15 months out.

We have a cluster.. in one of these clouds.. of B200s.. Our unit price right now is $2.63 an hour.. that's up for renewal in October. They came to us already in May and said $5.10 is the new price.. So double."

[Gappy (Giuseppe Paleologo) @__paleologo]

In interviews, Amodei often emphasizes the importance of culture and unity of the firm. Critics of the firm interpret this as a cult-like culture. However, it's likely that all the recent vey high-level hires by Anthropic (Karpathy, Boyd, Jumper!--this is effing crazy, by the way) are attracted by culture, unity of vision, and reputation. Not by infra availability, or RSI, or any other short-term gradient. All these will be gone in three months. And definitely not by money (if only, Meta).

IMO, this goes to show that one can be the underdog, even in industries with increasing returns to scale, and win. Of course, this makes me hopeful in my little corner of the world, where firm culture and values are usually underrated.

[rainbohtrout @rainbohtrout]

@sadvalueinvestr Tbh I think all of these platforms just need to publish, Coatue did the same. Once the alpha decays these platforms churn it out to their LPs eli5 style

[John Arnold @johnarnold]

Most of the SpaceX neocloud analysis changes dramatically if you understand that there's a backwardated curve for compute today.

[rainbohtrout @rainbohtrout]

@ByrneHobart Any time I find myself doing the same task twice I give it to my max account

[vivek @itsreallyvivek]

https://t.co/ZhOgyq7Vgn

[Gappy (Giuseppe Paleologo) @__paleologo]

Clearly, Fable is doing a lot of work, and unleashing a ton of agents. To review a short technical note, it released 31 agents, coded simulations to verify my results, did "adversarial reviews". Eventually, it only made the assumptions slightly more rigorous. It is all good. For a four-page technical note+a little code, though, it consumed all my Pro session tokens, *plus* $17 worth of credits. It is ridiculously expensive. I have 20-page reports that are way more complex than this. I can see how Anthropic has entered the phase of market-clearing prices, yield management, and pre-IPO.

I recall Boris Cherny saying in a podcast, "run Opus [4.6], not Sonnet. It's worth it". I feel comfortable saying that running your top-shelf model is *not* worth it anymore.

Decreasing returns, on most tasks. Like in the real world, some people can be real smart, but real expensive.

[Hiroo Onoda @OnodaCapital]

From the new Market Wizards

Good thing something like this could never ever happen again

[Anthropic @AnthropicAI]

Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.

It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx

[BuccoCapital Bloke @buccocapital]

Of course, now I am too old to be much of a fisherman, and now of course I usually fish the big waters alone, although some friends think I shouldn’t.

Like many fly fishermen in western Montana where the summer days are almost Arctic in length, I often do not start fishing until the cool of the evening. Then in the Arctic half-light of the canyon, all existence fades to a being with my soul and memories and the sounds of the Big Blackfoot River and a four-count rhythm and the hope that a fish will rise.

Eventually, all things merge into one, and a river runs through it. The river was cut by the world’s great flood and runs over rocks from the basement of time. On some of the rocks are timeless raindrops. Under the rocks are the words, and some of the words are theirs.

I am haunted by waters.