deaux

20 ms·

391.
by deaux 6mo ago
I'm super surprised that not a single comment here is bothered by this straight-out-of-Opus article. I guess the shared fury at GoDaddy is enough to overrule that emotion, or something. Because there are sure such comments on other art
392.
by deaux 6mo ago
I know exactly the feeling you mean. I get a much stronger feeling of that when I talk with friends who frequently take a plane for a 250 mile trip which has a world-class comfortable high-speed train connection with very frequent trains,
393.
by deaux 6mo ago
Yep. Wonder how long they'll wait with removing the 1x models, especially Gemini if it takes a long time, though they might just grab at an overdue reclassification of "preview" to "GA" to push it through.
394.
GitHub Copilot: GPT-5.5 7.5x more expensive under promotional pricing than 5.4(docs.github.com)
4 points by deaux 6mo ago|2 comments
395.
by deaux 6mo ago
You got caught out by AI-generated slop news.
396.
by deaux 6mo ago
Funny, but in this case it will be the opposite. If you tell an LLM to find potential regression, it will lean towards "finding" it even where there is none.
397.
by deaux 6mo ago
No, all I need to demonstrate is a country that saw no significant increase, not necessarily a decline. From everything I know, the US states as well as the Netherlands that all decriminalized it in the 70s didn't see local use increas
398.
by deaux 6mo ago
> The converse is also true It isn't true, at least not as a hard and fast rule. Post-legalization changes in demand differ greatly per country. It completely depends on contemporary cultural factors of the country in question.
399.
by deaux 6mo ago
Ah yes, the Western economic model forcing individual American companies like Amazon , Youtube and Uber to become profitable after.. checks notes _14 years_ for Uber, 9 years for Amazon, many years for Youtube.
400.
by deaux 6mo ago
https://pi.dev/
401.
by deaux 6mo ago
That's super interesting, isn't Deepseek in China banned from using Anthropic models? Yet here they're comparing it in terms of internal employee testing.
402.
by deaux 6mo ago
Vibes > Benchmarks. And it's all so task-specific. Gemini 3 has scored very well in benchmarks for very long but is poor at agentic usecases. A lot of people prefering Opus 4.6 to 4.7 for coding despite benchmarks, much more than I&
403.
by deaux 6mo ago
Not at all, it's more than enough for a large range of tasks. As for slow, that's just a function of how much compute you throw at it, which you actually control unlike with closed weights models.
404.
by deaux 6mo ago
Agreed on both, especially hallucination. That's what makes their chat app even worse, it's very opaque about web search and sources, so you can't tell whether it's a hallucination.
405.
by deaux 6mo ago
I've never ever had Gemini over the API switch languages in translation tasks and that's across more than 10 language pairs and 6 figures of calls, across both short and long outputs. Maybe your languages are even lower resource o
406.
by deaux 6mo ago
There are no real benchmarks of how "natural/idiomatic" output is in a multitude of languages. "Multilingual benchmarks" are usually something like "How good is it at a multiple choice exam like the SAT in lang
407.
by deaux 6mo ago
ctrl+f "cutoff, 0 results" Surely it doesn't still have the same ancient data cutoff as 5.4 did?
408.
by deaux 6mo ago
And they're incredibly good for their size.
409.
by deaux 6mo ago
They flatout gate any API access of the main models behind Persona ID verification. Entirely.
410.
by deaux 6mo ago
They had this ready and timed it for GPT 5.5 announcement. Zero chance it's a coincidence .
411.
by deaux 6mo ago
> The entire reason I keep a long-lived session around is because the context is hard-won — in term of tokens and my time. Silently degrading intelligence ought to be something you never do, but especially not for use-cases like this. Ha
412.
by deaux 6mo ago
Interesting, IME it's all beans. At my go-to place, the baristas are pretty bad but the two owners are super dedicated to roasting their own beans, you can always see them putting a lot of time and effort into it. Result is much better
413.
by deaux 6mo ago
It's time to reclaim the word "Meta" for Greek
414.
by deaux 6mo ago
> If I had to guess the pro and flash variants are 5x to 10x smaller than opus and gpt-5 class models. I really doubt it, especially Pro. If anything I wouldn't be surprised if their hardware lets them run bigger models more cheaply
415.
by deaux 6mo ago
Does that mean you're using API pricing rather than subscription? Seems like itd get expensive very quickly for a small team.
416.
by deaux 6mo ago
I've been thinking about making something like this myself. Afraid to tell you that half the stuff in there is already outdated. Models have their own archetypes. Since early this year almost every vibecoded website is Opus, which has
417.
by deaux 6mo ago
No, it is still indicative of a low quality product. And I say that as someone who has probably been agentic coding longer than you have. Indicative in my dictionary doesn't mean definitive. It just makes it much more likely. You can m
418.
by deaux 6mo ago
It took about 9 years of losses before they got profitable, which is the exact mechanism I talked about.
419.
by deaux 6mo ago
The whole Silicon Valley VC industry and the majority of the net worth of SWEs on HN is based on dumping. "Burning VC cash" is transparently dumping, and it's squarely what the US big tech dominance is founded on. Amazon, Ube
420.
by deaux 6mo ago
> Dumping is typically used in the context of international trade. This is dumping and it is international trade. Maybe you don't realize it because you're American and have internalized it as business as usual.

More ›