time for world to see the world
equity is the elephant. debt is the whale.
bigger, quieter, less covered.
the AI buildout is being financed by the whale while everyone's watching the elephant.
buffett bought gillette because whatever happened to the internet, people would still need to shave.
the AI version: whatever the next model does, will people still need this?
by the time a startup looks obvious, the price already knows.
one paradox of early stage VC: the hottest startups rarely produce the best returns
in fact the most valuable investments are often in companies without immediate signal. The pricing is lower, there are fewer competitors, and the founders tend to know things other people don't.
the challenge for those founders is they need to capture an investors attention long enough to explain what they do because, at first glance, they may not look venture-scale.
xitter.cf/harris/status/21100094…
nobody can call the exact bearing of a technology. the direction is enough.
the people who packed no coat lose whether we land north or south.
most of it, yes. the piece that changed is pace. a seed round used to buy 6 months of learning.
the same money now buys a competitor the same learning in a week.
VCs hyperfocus on TVPI.
TVPI is what a fund is worth on a spreadsheet.
DPI is what actually hits the bank.
We choose to care about both.
Hyperscalers now pay for AI with debt.
- @Amazon's long-term debt: up 81% to $119B in Q1.
- Alphabet: up 111% to $98B in the first half.
Profits can pause. Bonds against a half-built campus can't.
buy early, sell to the long-term buyer once the mark is real. dpi is the product.
Leo Alterman retweeted
We’re proud to announce our $40M funding round led by @eladgil and Verified Capital, alongside @01Advisors and @MarathonMP. This capital will expand our foundation, advancing in-house technology and operations, and enhancing our dining, lifestyle and travel offering.
if a company can't say what the AI does for the unit economics, the AI is doing something for the valuation.
congrats to the @reflection_ai team on Beam. 501B parameters, 23B active, trained from scratch, full weights out this month.
open weights at the frontier change who gets to build on top. that's the part worth watching.
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about Beam: reflection.ai/beam









