pants2

12 ms·

331.
by pants2 10mo ago
Generally, the easiest: 1. Sample a set of prompts / answers from historical usage. 2. Run that through various frontier models again and if they don't agree on some answers, hand-pick what you're looking for. 3. Test differe
332.
by pants2 10mo ago
Dupe: https://news.ycombinator.com/item?id=46121870
333.
by pants2 10mo ago
> Their value is as a structural check on the power of proprietary systems Unfortunately that doesn't pay the electricity bill
334.
by pants2 10mo ago
The best benchmark is one that you build for your use-case. I finally did that for a project and I was not expecting the results. Frontier models are generally "good enough" for most use-cases but if you have something specific yo
335.
by pants2 11mo ago
When I was young we'd find expired fire extinguishers and go blast each other with em. Very easy to use. Good times. Hope they're not toxic...
336.
by pants2 11mo ago
Source? Typically capital losses can only be netted against capital gains the next year, and only against a small amount of income.
337.
by pants2 11mo ago
I don't see signs of LLM writing in your comment so I'll have to guess no.
338.
by pants2 11mo ago
I think a lot of the volume on spot BTC has gone to ETFs, DATs, perps, WBTC, and other derivatives, when a few years ago spot was really the only option. Hard to track total volumes now.
339.
by pants2 11mo ago
If you see them returning value to shareholders in the short term, perhaps. But some of these DATs are giving their leadership massive pay packages (up to $400M[1]) so IMO they will sit on their treasuries as long as possible paying themsel
340.
by pants2 11mo ago
October 10 crypto liquidation bug on Binance triggered a cascading sell-off
341.
by pants2 11mo ago
The equivalent of "Spot Instance" is basically the OpenAI Batch API
342.
SF home invasion targets crypto of Physical Intelligence co-founder Lachy Groom(nypost.com)
6 points by pants2 11mo ago|2 comments
343.
by pants2 11mo ago
apologies to the mods for changing the headline but the articles on this had truly atrociously bad headlines
344.
by pants2 11mo ago
This is the worst headline. Why is the name of the person who posted the video more important than the person who was robbed? And it reads as if the suspect robbed his own friend, not Garry's friend.
345.
by pants2 11mo ago
That's my point though. Azure or AWS are just as bad and that won't solve anything. What are you running that requires a major cloud provider?
346.
by pants2 11mo ago
Don't be so sure - while I haven't tested Opus 4.5 yet, Gemini 3 tends to use way more tokens than Sonnet 4.5. Like 5-10X more. So Gemini might end up being more expensive in practice.
347.
by pants2 11mo ago
Congrats on finding it in within three days. I recently discovered something similar on our infra that had been going on for months (just an INFO log in a tight loop). This is the type of thing that major cloud providers should absolutely
348.
by pants2 11mo ago
Those are different things. You can run Codex inside Antigravity if you want to. Codex is an assistant CLI it extension and Antigravity is a VS Code forked IDE.
349.
by pants2 11mo ago
My experience with adjunct professors is they're much better at teaching than old tenured professors who just want to do their research and couldn't care less about the students.
350.
by pants2 11mo ago
Brand new account with 4 rapid & likely LLM comments, directional quotation marks, and common ChatGPT-isms such as "that does X without doing Y"
351.
by pants2 11mo ago
Why did you use an LLM to write a comment?
352.
by pants2 11mo ago
> The idea you are going to beat Two Sigma at their own game with tokens is just an absurdity. Individual quant traders aren't competing with Two Sigma. If you're an individual quant trader and you find a signal with $500k&#x2F
353.
by pants2 11mo ago
Well, if you can get a model to consistently lose money like that, then you just trade the opposite of what it says and you're guaranteed money!
354.
by pants2 11mo ago
Whisper simply isn't very good compared to LLM audio transcription like gpt-4o-transcribe. If Gemini 3 is even better it's a game-changer.
355.
by pants2 11mo ago
So the ending of Star Trek: Beyond wasn't so absurd after all
356.
by pants2 11mo ago
FYI on the jobs, CoreWeave's new AI center in Pennsylvania: 100 MW capacity, ~70–75 full-time technical roles initially, scaling to 175 full-time roles at ~300 MW.
357.
by pants2 11mo ago
lol I remember asking GPT4 how much aspartame it would take to sweeten the ocean, and it refused because that would harm the ecosystem.
358.
by pants2 11mo ago
Also two things can be true at the same time: 1. The neighborhood is so bad you'll be murdered for certain if you walk alone at night. 2. Murder rates are at record lows. This just means hardly anyone's walking alone at night - bu
359.
by pants2 11mo ago
Though to be fair, thinking tokens are also ignored in a lot of cost comparisons and in my experience Claude generally uses fewer thinking tokens for the same intelligence
360.
by pants2 11mo ago
To some degree this is normal - for example I worked at a med device startup where we got investment from one of our manufacturing partners. That helped with growth which in turn meant more manufacturing work for them, and presumably they&#

More ›