still somewhat loves computers β€’ some exp with llms β€’ there are no claude-like models β€’ tokens per petrodollar

Joined June 2025
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
I'm going to be presenting my research at the Latent Space paper club tomorrow, thanks @yikesawjeez. Link: luma.com/k8u3voz2
1
2
1
25
1,080
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
The discourse was split on this a year ago but I think a lot of classical procedural generation could in fact be materially improved by some LLM intervention
14
129
10
2,454
75,987
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
Good morning, Photon 2.2 is out with support for optimized local inference across NVIDIA Ampere, Ada, Hopper, and Blackwell GPUs, including A10/A10G, A100, RTX 3090, L4, H100, B200, and RTX PRO 6000 Blackwell.
12
6
102
4,911
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
everyone getting sniped by the personal drama, but missed the more interesting unanswered question: can these labs see all your work and scoop you when the stakes are "high enough"? πŸ€·πŸ»β€β™€οΈ
139
280
139
2,723
678,578
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
Ah yes my Reward hacking technique I haven't used since the Heian Era
1
1
23
377
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
no head steps in the same residual stream twice; for it is not the same stream, and he is not the same head
1
7
124
π‘˜π‘’π‘Ÿπ‘›π‘’π‘™π‘‘π‘Ÿπ‘–π‘π‘˜ retweeted
There's an old saying in Schellingβ€”I know it's in Hegel, probably in Schellingβ€”that all great world-historic facts and personages appear, so to speak, twice. He forgot to add: the first time as... as... tragedy... the second time as... as... it's funny if it happens again
100
5,564
163
27,184