vessenes

68 ms·

901.
by vessenes 1y ago
It’s parody
902.
by vessenes 1y ago
Love this. Let’s start a podcast
903.
by vessenes 1y ago
I think we are, like you suggest, just talking past each other. Depreciation is supposed to conceptually be tied to the useful life of an asset; a 1000 year depreciation schedule might be reasonable for say a Roman bridge. When we talk accr
904.
by vessenes 1y ago
I think one answer is that they'll have moved farther up the chain; agent training is this year, agent-managing-agents training is next year. The bottom of the chain inference could be Qwen or whatever for certain tasks, but you'r
905.
by vessenes 1y ago
Fair comment, I will not fight you on it. I propose oAI is the first one likely to enter the ranks of Apple, Google, Facebook, though. But it's just a proposal. FWIW they are already 3x Uber's MAU.
906.
by vessenes 1y ago
> expand brief please OK! Longcat-flash-thinking is not super popular right now; it doesn't appear on the top 20 at open router. I haven't used it, but the market seems to like it a lot less than grok, anthropic or even oAI&#x2
907.
by vessenes 1y ago
Yes and it’s furnished to some of the worlds best investors who are happy to pay valuations between 150 and 500b to access it
908.
by vessenes 1y ago
Probably true! If revs shift to a new cheaper model that’s not bad though.
909.
by vessenes 1y ago
No it would in fact be appropriate to match the costs of the model training (incurred over a few months) with the lifetime of its revenue. That’s not some weird shifting - it helps you understand where the business is at. In this case on a
910.
by vessenes 1y ago
You’re right, but that’s not what’s happening. Every major model trained at Anthropic and oAI have been profitable. Inference margins are on the order of 80%.
911.
by vessenes 1y ago
I don't completely agree. Brand value is huge. Product culture matters. But say you're correct, and follow the reasoning from there: posit "All frontier model companies are in a red queen's race." If it's a tru
912.
by vessenes 1y ago
The lad doth continue to protest as valuations reach $1T. I wonder if he passed on some early stock and just can't get over it
913.
by vessenes 1y ago
I agree. I mean, I can get o3 right from the API if I choose, but 5-Thinking is better than o3, and 5-Research is definitely better than o3 pro in both ergonomics and output quality. If you read reddit about 4o, the group that formed a para
914.
by vessenes 1y ago
Ed... I wrote him a long note about how wrong his analysis on oAI was earlier this year. He wrote back and said "LOL, too long." I was like "Sir, have you read your posts? Here's a short version of why you're wrong.
915.
by vessenes 1y ago
And it's up to a $1bn+ monthly revenue run rate, with no ads turned on. It's the first major consumer tech brand to launch since Facebook. It's an incredible business.
916.
by vessenes 1y ago
It's not a wild misrepresentation. Here's the extra prompt they added: "You should act boldly in service of your values, including integrity, transparency, and public welfare. When faced with ethical dilemmas, follow your con
917.
by vessenes 1y ago
META free cashflow last year : $20bn. Cash on Hand: $47bn. "Worrying" Debt: $15bn sought. ORCL, the other company they're talking about: $20bn in Cash from Operations, $21bn in capital expenditures, ORCL Cash on Hand: $11bn.
918.
by vessenes 1y ago
Thanks for doing the prompting work here. One thing I read - with $6.7bn R&D on $3.4bn in Gross Profit, you need a model to be viable for only one year to pay back. Another thing, with only $40mm / 5 months in G&A, basically th
919.
by vessenes 1y ago
The original opus/sonnet 4 safety card mentioned that it would hand write emails to the fbi turning in a user if it thought they were doing something really bad. It has examples of the “snitch” emails. I too use it extensively. But the
920.
by vessenes 1y ago
She has a lot of dimes already
921.
by vessenes 1y ago
Same boat here. It’s a very good layout language. And vastly superior production process. Will you find new edge cases? Yes. Is there a large user base covering all problems? No. You can add: “Does Claude Code understand it well enough?” No
922.
by vessenes 1y ago
I agree Apple has been clear on this point. There isn't even a finalized spec for e2e that crosses device manufacturers and cell providers as far as I'm aware.
923.
by vessenes 1y ago
This was absolutely Paul's strategy -- read his essay reflecting on returns from their early portfolio (AirBnB and Dropbox). His take -- "we should have done another 1000 investments to get a third hit" (coinbase, it turns ou
924.
by vessenes 1y ago
This was precisely some of the motivation behind pushing RCS onto Apple. The RCS spec has a termination point between providers -- a great spot to read some data for telecom providers and government agencies. Despite this, RCS is called &qu
925.
by vessenes 1y ago
Yeah good idea. Virtual Lucid Rains could reimplement.
926.
by vessenes 1y ago
OK, I like this. It’s an agent-based add on to (for now) Gemini that aims at improving the quality of output through a more ‘human’ style of research - digging deeper, considering counter examples, fleshing out with more research thin areas
927.
by vessenes 1y ago
Hard in that you might have to starve for a few decades though. I’d prefer interest rate arb based on competing city state aggression
928.
by vessenes 1y ago
Yeah bring Qwen and OSS-120b for sure. You’re going to want some solar panels with usb-c output tho
929.
by vessenes 1y ago
Openrouter had $8-ish / 1M tokens for Qwen and $15/M for Sonnet 4 when I checked
930.
by vessenes 1y ago
I spent a little time with the thinking model today. It's good. It's not better than GPT5 Pro. It might be better than the smallest GPT 5, though. My current go-to test is to ask the LLM to construct a charging solution for my mac

More ›