Open-weights model surpasses closed Mythos in cyber capabilities; people shrug
airesearch12 · x · 2026-08-14
A few weeks ago people panicked about closed and deeply supervised Mythos. Now an open-weights model surpasses its cyber capabilities and everyone is like :shrug:. This reflects how quickly we adjust to a new normal in AI.
More from AGI Musings
- Karpathy's 1-hour Stanford lecture: build AI engineering from scratch, emphasizing graph-based agent loops — VeryWellVersed · 2026-08-14
- Ben Goertzel's new essay: Why time has a direction, a mathematical exploration with implications for AI — bengoertzel · 2026-08-14
- Another Reason to Leave US Academia: International Interns' Contributions Overlooked — SonglinYang4 · 2026-08-14
- Prediction: 'Labor Shortage' Will Sound Dated by Mid-2030s as Digital Workers Become Unlimited — VraserX · 2026-08-14
- AI Executives Increasingly Push for Recursive Self-Improvement, Raising AGI Concerns — notkilleveryoneist · 2026-08-14
- Open Source AI or Marketing? Audrey Tang on the Line Between Weights and Training Data — 0xsachi · 2026-08-14