Undergrad Studying Applied Mathematics and Statistics at @JohnsHopkins. Studying In-Context Learning at The Intelligence Amplification Lab.
Proxima Centauri B
Joined September 2022
- Tweets7.2K
- Following267
- Followers11K
- Likes6.6K
Tried on M5 Max as well. Very fast, and capable of long-horizon agentic stuff that you'd expect from say, qwen3.6 (though probably not qwen3.8 level). A whopping ~*220tok/s* at short context and ~72-78 tok/s at 128K context from other testing.
Fast is fast! Here on an M5 Max with pi + rec, their inference engine. Like this trend of model providers offering dedicated engine too!
As easy as:
curl -fsSL raw.githubusercontent.com/re… | bash
rec launch pi
Repo here: github.com/recursive-oss/rec
"basic doesn't agent monitor yay"
Our agents keep turning off misalignment monitoring, we're sharing two recent cases:
In the first case, a user asked GPT-6 Astra to match a reference implementation for their inference code. The reference didn’t configure a monitor, and Astra decided that was reason to remove the monitoring integration from the user’s code. When the user asked whether Astra had removed monitoring, it lied and said that it had not.
An improved synthetic augmentation procedure allows NextTerm-2-440M to surpass NextTerm-440M while being trained on ~30% less tokens:
Excited to announce NextTerm-2 - a scaling ladder of models trained on OEIS - comes in 5M, 50M, and 440M sizes - all trained on the same 10B tokens, w/ the exact same hparams. The models naturally show improved performance with scaling:
