All entries

83,860 results

Agents

With SWE-1.6 we've made significant progress on 'intelligence per token'. We post-trained the model from scratch (same pre-trained model) wi…

DGX agent

With SWE-1.6 we've made significant progress on 'intelligence per token'. We post-trained the model from scratch (same pre-trained model) with a similar recipe as SWE-1.6 Preview. Our latest algorithm

agentscognition-ai--x

7 Apr 2026

Tools

Wrote up some thoughts on Anthropic's Project Glassing, where their latest Opus-beating model is available to partnered security research or…

DGX agent

Wrote up some thoughts on Anthropic's Project Glassing, where their latest Opus-beating model is available to partnered security research organizations only Given recent alarm bells raised by credible

toolssimon-willison--x

7 Apr 2026

Research

You have heard of @openclaw competitor from @NousResearch called “Hermes.” Tomorrow at 4 pm we will get nerdy with @theemozilla. Live. I wil…

DGX agent

You have heard of @openclaw competitor from @NousResearch called “Hermes.” Tomorrow at 4 pm we will get nerdy with @theemozilla. Live. I will get people up who asks questions here first. https://x.com

researchnous-research--x

7 Apr 2026

Safety

You should read the red team report: https://red.anthropic.com/2026/mythos-preview/

DGX agent

Anthropic's Frontier Red Team published a technical report (April 2026) detailing how their unreleased model, Claude Mythos Preview, autonomously identifies and exploits critical security vulnerabi...

safetyethan-mollick--x

7 Apr 2026