usef-

4 ms·

31.
by usef- 1mo ago
That would be a terrible tradeoff. The ship has already sailed and a lot of public AI content will not be their own. Deliberately making their product worse to reduce identifiability of AI inputs by 25% just doesn't sound worth it to m
32.
by usef- 1mo ago
Note that no one ever quotes the end of Dario's line, either, where he said programmers would still be needed in that 12 months. People think the prediction was more extreme than it really was because all the clips didn't include
33.
by usef- 1mo ago
Yes, as a user you pick what works for you. But it is a reality for them that growth has been huge, and GPU manufacturing is bottlenecked. People were very skeptical about how much investment most companies put into hardware/data cente
34.
by usef- 1mo ago
Commenters here likely haven't used it long enough to give non-superficial reactions. The customer quotes on the release page are all about it solving new problems, fwiw. We'll likely find out in the next few days how it really pe
35.
by usef- 1mo ago
Fable only being temporarily included in cheaper subscriptions was because anthropic is severely GPU constrained. They still are, and it impacts almost all of those unpopular decisions. They did announce from the beginning it was temporary.
36.
by usef- 1mo ago
You're assuming they're training the model to maximize the watermark signal, on top of already adding the watermark. I suspect that would hurt model performance quite a lot, and simply be unnecessary... the watermark tech works we
37.
by usef- 1mo ago
Luna as a doer, with a smarter model planning, can be a good compromise. Using sol for everything can be expensive without much gain, as a lot of steps don't need that sort of intelligence.
38.
by usef- 1mo ago
Out of interest, have you tried the newer models? You are not describing my experience recently.
39.
by usef- 1mo ago
Yes. This feature is brand new: https://developers.cloudflare.com/cache/changelog/ Simon wrote that in 2023: https://simonwillison.net/2023/Nov/20/cloudflare-does-not-co...
40.
by usef- 1mo ago
Exa also has an API for it that has worked well for me, returning markdown for a URL, which means you don't need to render js or anything yourself. It doesn't need an account for up to 1k requests/month, which is more than I&
41.
by usef- 1mo ago
In what way? It has worked well in my experience. It holds up with long context windows, unlike many, too.
42.
by usef- 1mo ago
You can't burn through a week's opus usage in 1 hour. You might be comparing the 5-hour limit of opus with the week limit of codex?
43.
by usef- 1mo ago
The whole point is that it's supposed to be more efficient. But models are also still getting absurdly more efficient every year, so you're likely nullifying much/most of the advantage. 18 months is a long time right now (and
44.
by usef- 1mo ago
You might be thinking of API pricing? The subscriptions give you a lot.
45.
by usef- 1mo ago
> We tell founders presenting at Demo Day, 'If you dress up too much, you will read to the investors as a stupid person.' They're coming to see the next Larry and Sergey, not some junior MBA type. https://www.wi
46.
by usef- 1mo ago
A simpler explanation is that it's just a better naming system. Calling something "small" might make it sound inferior to competitors. And S/M/L gets awkward as soon as you have more than three sizes. This naming sy
47.
by usef- 1mo ago
The things you list seem completely unconnected to me, and not driven by people embracing weird. Stereotype "brogrammers" were very similar to each other.
48.
by usef- 1mo ago
If by "everything considered weird" you mean a small chosen part of it from a decade ago. There are infinite pockets of it.
49.
by usef- 1mo ago
For what it's worth, the things you describe are mostly because they're extremely short of GPUs and growth rates were absurdly high. (Eg. They repeatedly said they'd keep fable in lower subscription plans if they had the capa
50.
by usef- 1mo ago
Your example is not realistic: it won't arbitrarily change a 0.5% token up to 65%. If it did it would absolutely degrade performance in a very measurable way. It's only biasing the randomness for tokens where it does have multip
51.
by usef- 2mo ago
(I had multiple bad autocorrects on this comment, written quickly, didn't reread until now)
52.
by usef- 2mo ago
Exactly. It currently seems to be a ranking of how much safety testing each company does. It's also only ever going to be the companies that publicly disclose it happening. (in the case of hugging face, OAI's hamd was forced to di
53.
by usef- 2mo ago
That's very defeatist. Do you have any concrete reason to think the major providers are lying to every one of their business/API customers about not training or storing the data? The business loss of trust would outweigh any benef
54.
by usef- 2mo ago
I just have one as a symlink to the other. Seems to work, and doesn't need an extra tool call.
55.
by usef- 2mo ago
I'm always surprised how large harnesses are By comparison, the recently open sourced grok harness was 840k lines, codex 950k, pi 250k https://simonwillison.net/2026/Jul/15/grok-build/
56.
by usef- 2mo ago
Comments elsewhere in this thread say other major brands are even more locked down. Edit: > With One UI 8, Samsung removed the OEM-unlocking capability globally, > Samsung uses a physical Knox Warranty Bit/e-fuse. Loading unoffic
57.
by usef- 2mo ago
The things you list not remembering may just be a different focus. I was always amazed by how many movie details other people remembered until I realised they rewatched those movies many times and I never rewatched anything. Similarly, bein
58.
by usef- 2mo ago
What would you recommend?
59.
by usef- 2mo ago
I have a lot of dead blogs in my RSS reader I don't remove for this reason. Maybe they'll start writing again. They've survived many opml exports/imports between different readers over the years.
60.
by usef- 2mo ago
Which chatgpt model?

More ›