All posts public domain under CC0 1.0.
We found embeddings like RoPE aid training but bottleneck long-sequence generalization. Our solution’s simple: treat them as a temporary training scaffold, not a permanent necessity.
arxiv.org/abs/2512.12167
pub.sakana.ai/DroPE
We found embeddings like RoPE aid training but bottleneck long-sequence generalization. Our solution’s simple: treat them as a temporary training scaffold, not a permanent necessity.
arxiv.org/abs/2512.12167
pub.sakana.ai/DroPE
1. CRON is inefficient
2. RLM (Recursive Language Models) are extraordinarily powerful
3. Every recursive algo can be implemented as a queue
4. I gave the agent a queue
alexzhang13.github.io/blog/2025/rlm/
kalshi.com/markets/kxca...
kalshi.com/markets/kxca...
it’s not vaporware like NFTs; people do (unfortunately) like interacting with the chatbot and it does have legitimate applications (Claude Code works shockingly well)
but “people want AI agents everywhere”? no
youtube.com/shorts/PraQC...
youtube.com/shorts/PraQC...
There is nothing to suggest the shooting of an unarmed woman in Minneapolis was justified.
This heinous killing must be criminally investigated to the full extent of the law.
There is nothing to suggest the shooting of an unarmed woman in Minneapolis was justified.
This heinous killing must be criminally investigated to the full extent of the law.
Fun fact there are a finite number of posters you can drive off BlueSky before it is smited by divine intervention. I don't make the rules...
Fun fact there are a finite number of posters you can drive off BlueSky before it is smited by divine intervention. I don't make the rules...