jsnell

19 ms·

331.
by jsnell 1y ago
You're moving the goalposts, given the original complaint was not about research costs but about the marginal cost of serving additional users... I guess you'd be surprised to find out that Meta's R&D costs are an order
332.
by jsnell 1y ago
Blogspam, dupe. Original research: https://news.ycombinator.com/item?id=44522772 (750 points, 478 comments) Another large thread: https://news.ycombinator.com/item?id=44526912 (276 points, 272 comments)
333.
by jsnell 1y ago
You're either overestimating the cost of inference or underestimating the cost of running a service like Facebook at that scale. Meta's cost of revenue (i.e. just running the service, not R&D, not marketing, not admin, none of
334.
by jsnell 1y ago
OpenAI has half a billion active users. You don't need every individual request to be profitable, just the aggregate. If you're doing a Google search for, like, the std::vector API reference you won't see ads. And that's
335.
by jsnell 1y ago
They'd be profitable if they showed ads to their free tier users. They wouldn't even need to be particularly competent at targeting or aggressive with the amount of ads they show, they'd be profitable with 1/10th the ARP
336.
by jsnell 1y ago
Every indication we have is that pay-per-token APIs are not subsidized or even break-even, but have very high margins. The market dynamics are such that subsidizing those APIs wouldn't make much sense. The unprofitability of the fronti
337.
by jsnell 1y ago
But the thing is, LLMs are already incredibly cheap to operate compared to the alternatives. Both for trivial things and for complex things.
338.
by jsnell 1y ago
How about you let the browser makers decide whether they need to have their own engine to compete? The fact is that Apple makes tens of billions in pure profit from Safari, and by closing off one of the principal ways of browser differentia
339.
by jsnell 1y ago
I think the argument is that as long as 3p browsers are forced to be just thin WebKit wrappers, it's harder for them to compete against. Why even bother switching from the default when it's going to be the same slop with a differe
340.
by jsnell 1y ago
The purpose here isn't to deal with sophisticated spoofing. This is setting a couple of headers to fixed and easily discoverable values. It wouldn't stop a teenager with Curl, let along a sophisticated adversary. There's no c
341.
by jsnell 1y ago
> after losing the chrome anti-trust trial? There hasn't been such a trial.
342.
by jsnell 1y ago
Also: https://news.ycombinator.com/item?id=44510913
343.
by jsnell 1y ago
The Claude Code privacy policy[0] is pretty explicit that by default they train on neither the prompts, usage data, or even explicitly provided feedback data (presumably /bug?) that can be used for other product improvements. > By d
344.
by jsnell 1y ago
By considering the bomb location notations you have made on the board. Are they all valid, and do they prove that the location you're clicking on is empty? Then you're not guessing. Are they invalid, or do they not prove that the
345.
by jsnell 1y ago
I doubt it. A guaranteed-solveable minesweeper scenario isn't just about the global board setup, but about what information is available when, and what order the solution is gone through. Bombe[0] is to my mind the definitive explorati
346.
by jsnell 1y ago
I'm telling you that the article you wrote doesn't match what you now claim is the real issue. Your article is entirely concentrated on the monetary harm, and ludicrous demands for repayment. I don't particularly care about w
347.
by jsnell 1y ago
If it wasn't about money, you would not have written an article that concentrates just the amount of money you spent.
348.
by jsnell 1y ago
This genre of writing is quite hard to engage with productively. It is really long due to throwing everything at the wall. Doesn't matter if the argument is obsolete, a lie, unprovable, or actually a legit and we'll researched com
349.
by jsnell 1y ago
I don't see how indexing is related to this? The question is about how to get the LLM to reliably apply the edit it wants to make. Even when the full current version of the file is in the context, this is one of the flakiest bits of th
350.
by jsnell 1y ago
Because the title is a blatant lie.
351.
by jsnell 1y ago
You seem to be either implying the deal doesn't actually exist, or that it is with some entity that won't actually be able to meet the commitment. The former would be outright fraud. The latter would be a really bad idea on Oracl
352.
by jsnell 1y ago
It doesn't seem like a very fresh insight though. I did exactly that for sb-sprof in 2006 [0], and doubt it was the first time anyone had that idea. It falls out pretty much automatically from having both a bump allocator and a samplin
353.
by jsnell 1y ago
Thanks, but where are those accessibility settings? /help shows nothing related to settings other than auth and theme, there's no related flags, and there's a ~/.gemini/settings.json that contains just the auth type
354.
by jsnell 1y ago
What's up with printing lame jokes every few seconds? The last thing I want from a tool like this is my eye to be drawn to the window all the time as if something had changed and needs my action. (Having a spinner is fine, having chang
355.
by jsnell 1y ago
The costs are already far, far below that level. The only reason the consumer-facing businesses are not profitable is that nobody is yet showing ads, i.e. providing service to hundreds of millions of people with no monetization at all. LLM
356.
by jsnell 1y ago
LLM inference has large economies of scale. Properly batched requests are tens of times cheaper than individually processed ones. And it's going to be quite hard for a self-hoster to have enough hardware + enough usage to benefit from
357.
by jsnell 1y ago
I mean, I've mostly been trying to establish what the specific alleged behavior even is and what the factual basis is for the allegation. I certainly can't agree with the maximalist interpretation of what you're ascribing t
358.
by jsnell 1y ago
> begins demanding something Yes, those alleged demands by "leechers" are exactly what I've been asking about. From the lack of specifics, it is starting to be pretty obvious that they do not actually exist. I'm not s
359.
by jsnell 1y ago
That doesn't look like Google, Apple or Microsoft though? It's some random guy.
360.
by jsnell 1y ago
> If you contribute nothing and demand security fixes, then you’re not a leecher — you’re a parasite, because your demands exceed your contributions; There's the "demand" word again. Who is that demanded something from t

More ›