vessenes

35 ms·

811.
by vessenes 1y ago
Curious - what do you do on your Linux (or FreeBSD natch) box that your m1 couldn’t?
812.
by vessenes 1y ago
I feel old! I never knew this existed; probably because my Unix shell experience was mostly built in the early 90s and zsh was not often used or distributed in large academic installations at the time; I believe ksh cost a licensing fee as
813.
by vessenes 1y ago
this belongs at the top; early culture was straight Mcdonell Douglas reportedly, and extremely ineffective.
814.
by vessenes 1y ago
OK, AI playing video games is cool. But you know what's really really cool? It looks like SIMA 2 is controlling the mouse and reading the screen at something approaching 30+fps. WANT. Computer use agents are so slow right now, this is
815.
by vessenes 1y ago
The scene just before you describe has the user write "ripe tomato" in the description - you can see it in the video. The summary elides it, but the "ripe tomato" instruction is also clearly part of the context.
816.
by vessenes 1y ago
Potato Potatoh. I think Apple is the largest game platform in the world, or ate least iOS is.
817.
by vessenes 1y ago
> That’s not necessarily a bad thing. I think it is in that it gives censorship power to a large corporation. Combined with close-on-the-heels open weights models like Qwen and Kimi, it's not clear to me this is a good posture. I th
818.
by vessenes 1y ago
Nope - it's "Hey everyone, this is possible everywhere, including open weights models."
819.
by vessenes 1y ago
They don't have to disclose any of this - this was a fairly good and fair overview of a system fault in my opinion.
820.
by vessenes 1y ago
Kimi K2 could easily be used for this; its agentic benchmarks are similar to Claude's. And it's on-shore in China, where Anthropic says these threat actors were located.
821.
by vessenes 1y ago
The Tesla on fsd13 is considerably more courteous and safer than at least the bottom quintile of drivers. It’s much safer than my teenage drivers were in my family. Failure modes in city driving tend to be stopping, not plowing through pede
822.
by vessenes 1y ago
I like this plan, but don't you already have this from the input vector in the prompt, at least if the inference is 'chunk wise' - generating a latent space vector, decoding it, outputting it, doing the next one. What if you
823.
by vessenes 1y ago
quick note -- I believe you need a separate key to get biometric data out of the passports, but it's been a while since I looked at passport digital infrastructure.
824.
by vessenes 1y ago
Yes, I had total PTSD reading that in the announcement. Whether it's just evolving a tone so that we don't get fatigue or actually improving, I'm happy we're moving on. My audio (still 4o I believe) interactions are madd
825.
by vessenes 1y ago
I don't use it much, but I tried it out with okara.ai and loved their interface. No other connection to the company
826.
by vessenes 1y ago
I'm sure it is. That said, they've also increased its steering responsiveness -- mine includes lots about not sucking up, so some testing is probably needed. In any event, gpt-5 instant was basically useless for me, I stay default
827.
by vessenes 1y ago
Exactly my reaction. I hate this. I hate that someone thought of it and that it exists. I have no idea if it will sell, but I was like "no way, that cannot be a real launch from Apple"
828.
by vessenes 1y ago
Yeah in my mind it was smaller -> faster. But seems like it might be 'a bit smaller and batched' to hit the price target.
829.
by vessenes 1y ago
Meh. I do a lot of beta testing, and enjoy it. I think it's fun to watch the evolution of this tech over the last seven or eight years. But you don't have to feel that way.
830.
by vessenes 1y ago
Qualitative reactions matter. I didn't say it was genuinely safe. I can say that it was smooth, reacted better than most adult drivers to the situations I witnessed, and was significantly better at those things than the prior version.
831.
by vessenes 1y ago
Sorry but while I am a happy Waymo user, this is overblown and just incorrect. Waymo has datacenter oversight. (which, who cares, the product is great). Tesla FSD 12 -> 13 was a massive jump that happened earlier this year. 14 is still r
832.
by vessenes 1y ago
have you tried fsd 13/14 yet? I just upgraded to 13 from 12 and it's a massssive improvement. Not sure what 14 s like. It's definitely added a 9 in the last year.
833.
by vessenes 1y ago
This sounds amazing. A demo video would help me finish sign up - I can’t try it without hooking it up to real data, and I don’t want to for a test.
834.
by vessenes 1y ago
I think I got us there, when I mentioned that they had no means of enforcement, and so the statement was meant as you say - to be absurdist in practice underscoring the national sovereignty at the heart of the discussion. Is that what you t
835.
by vessenes 1y ago
Wow, that looks completely different than how Safari rendered the site. Thanks for the link. I like the look of it hosted at Google.
836.
by vessenes 1y ago
Looks like a leak: https://platform.openai.com/docs/models does not list it, and codex-mini-latest says that it's based on 4o. I wonder if it will be faster than codex; gpt-5-nano and -mini are still very slow for
837.
by vessenes 1y ago
You're aware this exists, yes? It's called the OFAC list. It's not generally used to list technologists or web sites related to speech, though, as far as I know.
838.
by vessenes 1y ago
Preston talks about this in an earlier blog post. It's a matter of UK law and, I'd bet, many countries, that they have the sovereign right to legislate about matters in and outside their borders. "If Parliament says smoking i
839.
by vessenes 1y ago
Thank you for this huge time saver. Now, about the paper-that’s super interesting. I imagine the dream here is to distil down into a “reasoning” core. Or maybe reclaim space for more generalization. Lots of interesting use cases.
840.
by vessenes 1y ago
I think you need a custom benchmark -- have you considered making one out of the excel world championships?

More ›