@silverarm0r wow, congrats! it's a great short story.
btw, is that a graphgear 500? looks like it but I don't think they make them in white?
English
~
1.8K posts
Hahaha
So they haven't built anything with opus
This all makes sense now
Lisan al Gaib@scaling01
ANTHROPIC HAD MYTHOS INTERNALLY SINCE FEB 24
English
@synthwavedd bro, "BrowseComp: Claude Mythos Preview scores higher than Opus 4.6 while using 4.9× fewer tokens."
English
Mythos scores 77.8% on SWE-Bench Pro (vs 53.4% for Opus 4.6), and 82% on Terminal-Bench 2.0 (up from 65.4% on Opus 4.6)
leo 🐾@synthwavedd
🚨BREAKING: Anthropic has "no plans" to release Mythos, will instead make it available to 40+ companies for cybersecurity work
English
@georgejrjrjr Or the training, more broadly, no? Could be an environment
English
@nosilverv When change becomes easy, the potentiality becomes closer to essence. A fat person is now almost not fat in essence.
English
@humancompressed so it does better than the worst (llama.cpp) quantization baselines out there? at least choose a good baseline
English
The question, of course, is whether a good 2 bit Qwen 3 4B does better than 1-bit Bonsai 8B, which I'm almost certain is true.
PrismML@PrismML
This scatter plot shows the Pareto frontier of intelligence vs. size, defined by models like Qwen3 0.6B, 1.7B, 4B, 8B, and Ministral3 3B. The 1-bit Bonsai family shifts that frontier dramatically to the left. This changes the tradeoff itself: models no longer have to be large to be capable.
English
@moravaginelol wow, he looked like a great boy. sending thoughts and prayers.
English
@silverarm0r Wait, why is that rock emanating light? Can you pick it up?
English