minimaxir

5 ms·

61.
by minimaxir 1mo ago
I see this as well. Odd there is only 6 Pro.
62.
by minimaxir 1mo ago
Also is rolling out to the $100/$200 Codex subscriptions, I got access this morning. Unsurprisingly, it does indeed consume usage at ~2.5x the rate of Sol.
63.
by minimaxir 1mo ago
That's how we get AM.
64.
by minimaxir 1mo ago
Yes, in theory.
65.
by minimaxir 1mo ago
"Chain of Thoughts" is a term from the title of a 2022 research paper "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models" ( https://arxiv.org/abs/2201.11903 ), well before ChatGPT
66.
by minimaxir 1mo ago
ARC has their own writeup on the result, which offers some nuance. https://arcprize.org/blog/astra tl;dr it's 62% when apples-to-apples to other models, which is still notable.
67.
by minimaxir 1mo ago
DeepSeek v4 Flash recently had a similar "more reasoning is cheaper" curve. It's a fun counterintuition.
68.
by minimaxir 1mo ago
IMO headline is misleading because it implies OpenAI is being petty, but in this case the cut off is extremely understandable from a business and legal perspective.
69.
by minimaxir 1mo ago
That's not defensible.
70.
by minimaxir 1mo ago
> They've all been compelled to build the same horribly expensive AI infra, to serve similar models that have no ability to lock-in customers Emphasis added, since having a horribly expensive AI infra allows offering enterprise co
71.
by minimaxir 1mo ago
This submission has citations on where he was wrong, they aren't predictions.
72.
by minimaxir 1mo ago
That's an unfair characterization of tech journalists in general. When BuzzFeed News was around, I worked closely with them on technical aspects and projects to ensure that everything was correct and accurate, and there were very recep
73.
by minimaxir 1mo ago
This very submission is about cases where Ed has not been pretty spot on.
74.
by minimaxir 1mo ago
There was a major material difference between the FT and Ed's post: FT explicitly noted that OpenAI has $billions in cash on hand (which is relevant if you are analyzing data about solvency), while Ed's post obfuscated that data p
75.
by minimaxir 1mo ago
Over the past few years, I had helped Ed with some difficult nuances around the obscure technical aspects of serving LLMs (e.g. benchmarks and model caching); he has also shouted myself and Simon Willison out positively multiple times. I st
76.
by minimaxir 1mo ago
Post on your main account if you care that much.
77.
by minimaxir 2mo ago
Obligatory question: who is still using OpenClaw? It's been a minor meme about how much discussion about OpenClaw fell of a cliff after March; the Google Trends graph is funny: https://trends.google.com/explore?q=OpenCl
78.
by minimaxir 2mo ago
Occam's Razor would suggest that's because Simon posts good content. You can check his domain history to confirm that not everything Simon posts gets upvoted. https://news.ycombinator.com/from?site=simonwillison.ne
79.
by minimaxir 2mo ago
AI tokens are not like tokens at Chuck E. Cheese's. They are not interchangeable.
80.
by minimaxir 2mo ago
Optimization. Why use many word when few word do trick?
81.
by minimaxir 2mo ago
You can click the table headers to sort Ascending/Descending.
82.
by minimaxir 2mo ago
Hy4 apparently has ludicrous traction on OpenRouter already ( https://openrouter.ai/tencent/hy4-preview ), with trillions of tokens processed in a couple days: more than GLM 5.3 in a week. That said, it's relatively
83.
by minimaxir 2mo ago
It is confusing that music companies are suing Anthropic, the one company that does not have any programmatic AI audio input or output through their models.
84.
by minimaxir 2mo ago
Writing code is only one part of software engineering. Making good engineering decisions is the other, and LLMs aren't optimal at that...yet.
85.
by minimaxir 2mo ago
It is weird for them to simultaneously a) apologize and b) brag about the marketing ROI of the trolling.
86.
by minimaxir 2mo ago
Aside from Switch emulators which get C&D by Nintendo due to private key copyright infringement issues, nothing with existing emulators would cause a legal tussle. AI has zero impact on it.
87.
by minimaxir 2mo ago
It's especially annoying that LLMs/agents are a logical starting point for decompiling old games but there is a significant backlash against it as the gaming community is very anti-AI and consumers of the benefits decompiling of
88.
by minimaxir 2mo ago
Specifically, you can set an account setting "only do ZDR" and it will error if you try to use an endpoint that doesn't do it.
89.
by minimaxir 2mo ago
And now pretty much everyone is hosting it, with some matching the 50% discount.
90.
by minimaxir 2mo ago
"HN is down more frequently so therefore it must be AI" does not logically follow.

More ›