tough

10 ms·

211.
by tough 1y ago
or the router was routing them to the best model all the time prior to launch. I had a very brief window where gpt-5 was really good/fast on cursor-agent day of launch. also horizon-alpha-beta on open router im pretty sure they where g
212.
by tough 1y ago
but unlike cuda there's no custom kernels for inference in vllm repo... I think
213.
by tough 1y ago
We're the AI's Robert's Ducks
214.
by tough 1y ago
Yes I think so, but I meant mostly as inside ChatGPT, like o3 without search is the deal. I hope they add back models selectors mostly
215.
by tough 1y ago
Right it was an out of context presentation of their own company and some very rough AI generated content and years ago before the existence of LLMs but yeah sensationalism and all and people don't do research so unless you remember we
216.
by tough 1y ago
there's no more data to throw compute at to 1T tokens is a lot
217.
by tough 1y ago
I miss o3 greatly too, and could be one of the few reasons to keep me on the PRO plan to give me access to it via ChatGPT optionally
218.
by tough 1y ago
gpt-oss-120b can be used with gpt-oss-20b as speculative drafting on LM Studio I'm not sure it improved the speed much
219.
by tough 1y ago
Doesn't google have TPU's that makes inference of their own models much more profitable than say having to rent out NVDIA cards? Doesn't OpenAI depend mostly on its relationship/partnership with Microsoft to get GPUs to
220.
by tough 1y ago
nice! auto-updates like open code could help to not have to remember to update loving the animations and todos so far also gpt-5 is just great at agentic stuff
221.
by tough 1y ago
token entanglement is an interesting emergent behaviour https://owls.baulab.info/
222.
by tough 1y ago
I read someone that adding -nsfw- or such words to the prompt made it go away reliably funnily enough
223.
by tough 1y ago
agi clause comes to mind?
224.
by tough 1y ago
well it's difficult to trust the people selling it in the first place. They're too biased to not lie It's hard to make a man understand something standing between them and their salary
225.
by tough 1y ago
Words We Haven't Seen - Search unseen words made me chuckle
226.
by tough 1y ago
you realise now google is plugging gemini to all their queries and giving you summaries and stuff so maybe not so much anymore? would be true if it was -pure- search on google's part but it isn't anymore
227.
by tough 1y ago
yeah in that regard it seems that apple tastefully does so on apps where there's mostly remote roundtrips already like the app store or music app, so I agree that there makes sense to reuse the web infra but swift ui apps are great and
228.
by tough 1y ago
so microsoft gave up and the web won? swift ui apps have some webkit views like the app store, music app etc
229.
by tough 1y ago
what if that contribution benefits me personally in any way whatsoever?
230.
by tough 1y ago
dictatorships are the worst. but true, even professional athlete's when travelling defect etc
231.
by tough 1y ago
there was recently this soham dude on SV/YC doing the rounds a few weeks ago if all you do everyday is interview, you obviously get great at it
232.
by tough 1y ago
Its quite presumptuous to assume that 1. Every citizen will have ties or family to -think of- 2. The current regime (which is now on its third generation) will last 7 generations more. 3. You have any descendants at all
233.
by tough 1y ago
if we get AGI and a post-scarcity age what makes these people think they -reaching- AGI will make them kings. seems like governments will have a thing to say about who's able to run that AGI or not. GPU's run on datacenters which
234.
by tough 1y ago
on gh someone reproduced but paper lacks total gpu hours and their benchmark results where 10-20% lower (read on gh issue)
235.
by tough 1y ago
and what if you don't work there or have a salary but happen to own some equity?
236.
by tough 1y ago
Cursor added a UI for todo list and encourages it's agent to use it (its great ux, but you can't really see a file of it) kiro from amazon does both tasks (in tasks.md) and specs. Too many tools soon, choose what works for you
237.
by tough 1y ago
wasnt apple running a bot now for AI/llm stuff
238.
by tough 1y ago
how did you do that? cursor seems very hard for me to modify to use local llms
239.
by tough 1y ago
remarkably both Cursor and Zed do this (GUI, not CLI)
240.
by tough 1y ago
lmao weirdest stuff ever on X and it seems like nobody cares anymore?

More ›