← Leaderboard
BT
Bret Taylor
OpenAI (chairman) / Sierra (co-founder)
First call on the ledger
No calls filed yet. Quotes in curly quotation marks are verbatim; everything
else is a labeled paraphrase. Every call links its original source. Before you trust the call, check the record.
Where they stand
Views on the record but not scoreable bets: no single number and deadline to grade. Shown in their exact words, never scored.
Responding to Kimi K3: open-weight models 'are not necessarily cheaper to run' (@36:26), 'the frontier models are much, much more token efficient' (@36:38), and 'for a lot of tasks, it turns out these frontier models from OpenAI and anthropic are actually just better' (@37:26). Teaser duplicate of this line at 34:25; in-context occurrence filed.
“at the end of the day it's going to be, you know, how efficient is your inference. And I think the frontier labs are still far ahead here. And I think that's really where the value is going to accrue”
'if companies are thinking they can differentiate on the models, I think that's probably a fool's errand' (@47:40); analogy: 'Having a website is not a competitive advantage' (@49:11). Also predicts a maturing applied-AI market: 'if you fast forward 12 months from now, IT departments will be really sophisticated' (@45:21) and 'where the world is going is paying for outcomes' (@41:58) - the Sierra pitch.
“if you fast forward a few years, intelligence will be widely available. So you're not going to be able to differentiate on intelligence, because you can just license it”
On the new Chinese K3 open-weight model: 'it turns out the frontier models are much much more token efficient' @1:29, and on US labs keeping the lead: 'i think the u.s labs need for every price performance latency cost they need to have the best model available and actually i think that's more likely than not' @2:46.
“I'll say one thing that I think is a little bit overblown about these open weight models is they're not necessarily cheap cheaper to run”
Answering the host's challenge that if models can be built for a lot less 'that's a different game too'; he grounds it in inference efficiency and capability: 'for a lot of tasks it turns out these frontier models from open AI and anthropic are actually just better than these open weight models' @2:19 and '5.6 is leading by most of these metrics' @4:47.
“the frontier labs are still far ahead uh here and i think that's really where the value is going to accrue”
His answer to the enterprise ROI complaints (Khosrowshahi, Karp cited by host @5:23-5:36): 'the market's in kind of a quirky spot where everyone's worrying about paying for tokens' @5:44; at Sierra 'you don't pay per token you pay per outcome you pay per originated loan' @6:32 and 'the solution to a lot of the problems with ai tokenomics is we don't have a mature applied ai market' @7:52.
“i believe where the world is going is paying for outcomes”
Follow Bret Taylor
One short email when Bret puts a new call on the record, and when a verdict lands on one. Plain words, free.