This is a fundamental mispricing in 3 parts:
1) Accounting lag≠econ value (// early $SNOW lumpy rev/cust skepticism)
2) Misunderstanding ramp econs (// VCs hated $CPNG GMs, bc missed $AMZN-like logistics infra cost curve)
3) Underweight optionality (Need to price surface area!)
🧵 Perpetual License → SaaS Subscription → Consumption-Based
This shift is more than pricing – it changes GTM, accounting, sales comp, valuation…everything!
This is how I’m thinking about the transition to usage-based pricing, and what it means for AI + crypto👇
Rundown on where Visa is seeing pmf w/ stablecoins from their CEO, Ryan McInerney:
> 1) Consumer payments -> Visa card issued on top of stablecoin balances. Growing "very, very quickly" and up to ~160 stablecoin-card programs globally (i.e. Whop, Etherfi, Plasma, Redot, $WU soon etc)
> 2) Stablecoin settlement -> "Historically we settle in fiat. We do it Monday through Friday, and we do it once a day. But now we're using stablecoins to offer settlement 7 days a week, multiple times a day across border around the world". Up to single-digit billions (last # I saw $V release was ~$5 billion in Jan '26) and growing "very, very fast"
The latest trend of renewed interest in public sales proves that.
I personally prefer to invest in a new round on @legiondotcc, @buidlpad, @echodotxyz, or @MetaDAOProject rather than farm a new airdrop.
All the airdrops I farm are related to stablecoin liquidity because I just want a higher APR.
Airdrops are dead soon.
In 10 years, many more people will use crypto, but they may not know they're using crypto.
They only need to feel the benefits, not understand the systems behind it. The best tech is often invisible.
The next prize to chase here is physical intelligence - the ability for an AI to 'act with consideration' (e.g. gingerly squeeze an overrip tomato, swirl clear liquid in a cup, use a paintbrush or paint roller interchangeably, or safely handle a hot drink).
Interesting that GPT 4.5 and LLaMa-3.1-405B pass the Turing test (convincingly imitate human convo) when prompted to adopt a humanlike persona.
Source: UCSD Mar-25 study by Cameron R. Jones
To the class of '24/'25:
There's a famous line from @pmarca when he moved to Silicon Valley in 1994. The Internet boom was already under way. He came from a small town in Iowa, and seeing these growing giants like AOL and Cisco, he was worried that he was already too late. That by 1994, he worried the whole Internet thing was basically done.
I felt the same when I first got into crypto VC in 2018. Polychain, Pantera, a16z, DCG, and others were already legendary. Was there really space for another crypto VC? Seemed like maybe it was already too late, and the industry had consolidated.
Don't worry. Barely anything has happened yet.
BTC is still 1/10 of gold. Less than 1% of the world is on-chain. There hasn't been a single law passed in the US about crypto. And there's not a single major central bank in the world owning any of this stuff.
You're not late. Crypto is still tiny compared to what it's going to be in 20 years. So welcome to the crypto—a lot is going to change from here. Come make your mark.
We've reached a major milestone in fully decentralized training: for the first time, we've demonstrated that a large language model can be split and trained across consumer devices connected over the internet - with no loss in speed or performance.
4/ Initially, new entrants may benefit more, but incumbents *can* catch up.
Naturally PE will sell the dream; however, the default complacency is going to wreck returns unless these firms push portcos to optimize 'barrels'
3/ New entrants can more effectively build around AI from the ground up to maximize AI potential.
Recall: Cloud native vs. legacy prem businesses - shifting to cloud vs. nativity is much harder
"You can’t eat spread, or spend spread, or pay pension benefits with spread. For those things, you need returns." - Howard Marks
While I don't typically play in credit mkts, HM is a proven investor generally worth listening to (1/x)