labbing @hakosystems :: prev @nascent

open source
Joined January 2022
Messing around with my android phone and codex. It’s kinda crazy that you can plug a phone into your computer and AI can start hacking on it immediately. We’ll see if I can get the phone controlling itself
12
11
133
14,118
I’m not so sure about this new feature @nikitabier
This Post is unavailable. Learn more
4
259
Replying to @plotchy @0x_Osprey
hmm i think raw dogging my password into their site is too weird
1
1
59
I'm winning chess games you can't access with pieces you've never heard of @PixieChess
3
3
618
Multiplying numbers used to be an LLM gotcha but now is nearly solved. For this competition I ran a grid search over all mainstream models and gpt-oss-120b stood out scoring ~93% correct on 18digit*18digit multiplication! Insane! eg: 364826485628193748 * 492816485726395817 = ...
𝚐𝚙𝚝-𝚘𝚜𝚜-𝟷𝟸𝟶𝚋 seems to be weirdly good at more "computational" problems like sudoku and multiplying numbers ~76 higher scores/$ ratio:
4
1
13
2,892
Sora 2 has him acting different
4
4
4
67
13,367
Was it worth it? I'm still not sure
6
2
1
39
6,146
Excited to be able to aura farm 🏆
Excited to finally launch what I've been working on the past 3 months! Introducing Peval, a competition to see who can write the best prompts, and which models are the best at different evals come beat me: peval.io/competition/chess
5
827
Replying to @IamEmily2050
consistently good
219