Co-founder/Partner @hanabicapital Prev Partner @IndexVentures, Infra Eng/PM@ Palantir + MemSQL

Santa Cruz, CA
Joined July 2009
The metric to track seems to be total compute as a percent of revenue. Or, more conservatively, compute as a percent of net revenue (less marketplace fees). Perhaps with a quarter of two lag for the portion of compute that goes to training.
1
1
49
Implied gross margin is a bit flattered as it does not seem to include any training compute, any data spend, or any marketplace fees in COGS.
1
13
Bear or Bull, one cannot deny that this is an impeccably timed IPO.
8
534
The ANT S1 is by any measure incredible. Only interesting Q is timing. Power is in the Q2 numbers. Peak tokenmaxxing, peak advantage over OAI, pre-OS takeoff, pre-price cuts, at a time when they were still conservative on compute. Will be interesting to see next few quarters.
1
5
770
One of the funniest things about the Anthropic S1 is that virtually every relevant metric was already leaked.
8
442
The question isn’t just demand. It’s can you deliver against that demand in a given timeframe in an economically viable way.
1
7
496
lol and it got worse.
Give it up for the Baltimore Ravens!
179
And for my next act, I will light your house on fire and sell you the fire truck.
4
431
Bryan Offutt retweeted
It is very interesting that AI is quite bad at many things, but specifically good at many of the most dangerous things (hacking, code, bio, etc).
1
10
485
Bryan Offutt retweeted
Excited to announce Agent Mode. 1/ Right now, agents are struggling to use websites on the internet. As a result, personal agents take too long to reply, and don’t get the information they need.
69
53
24
225
21,378
I really enjoy that Jensen just refers to them as "computers". I am not sure what else he would call them, but it just feels incredibly quaint and old school.
1
1
154
Bryan Offutt retweeted
At some point you need to stop playing covers and start playing your own songs.
You cannot execute someone else's strategy. Nobody thinks about your business nearly as much as you do. No one can obsess over the problem more than you. You have more data than all of them combined. You're the one who has to execute on all the space in between. Every decision downstream of a framework you don't actually believe in gets wobbly. You just accumulate drift. When things don't go perfectly right (and at a startup, they won't), you'll start second-guessing the whole thing and think you probably should have just done it your way. So yeah, just do that from the start.
2
1
34
1,878
Single breadcrumb is dropped on table. All of X: this breadcrumb means *insert sweeping assertion about the loaf that it came from*
1
172
There’s a difference between idea market fit (promise market fit?) and product market fit.
1
2
284
Another terrible Ravens 4th quarter implosion in the books.
225
It’s unbelievable how bad NFL officiating is.
184
The movie is as good as everyone says it is. It is the only documentary I have watched where had to regularly remind myself that what I was watching was real. Genuinely thought provoking.
1
312
I wonder to what degree lab revenue is improved by releasing models more often simply because everyone has to run evals against the new models.
1
8
1,006
One of my strongest beliefs is that the term "conviction", its overuse, the way it's venerated, and the incentives that veneration creates has done considerably more harm than good in VC. There is an important distinction between belief and feigned (or retroactive) certainty.
7
482