I've reviewed every PDF editor out there - then I had ChatGPT build me a better one

I've reviewed every PDF editor out there - then I had ChatGPT build me a better one ...

Context compression finally works in production: new research cuts LLM input 16x without the accuracy hit

LCLMs compress LLM context before decode — 8.8x faster at 16x compression, beating every KV cache method tested. Open-sourced by NYU and Columbia.
eLife

Distinct sensorimotor encoding in tuft dendrites and somata associated with action, correction, and learning

This important study reveals distinct representations of task-related information in the dendrites and somata of cortical neurons during sensorimotor learning and behavioral adaptation. The evidence ...