Not Boring #215: Cammarata's Clean Networks Claim, Tesla's Flying Roadster, Tau Cut 73%
Not Boring (Packy McCormick) · rss · 2026-10-16
Packy McCormick's weekly roundup: ex-OpenAI interpretability researcher Nick Cammarata (Goodfire) argues neural networks are already legible—polysemanticity was misread as 'recursive generalization'—so alignment is likely solvable. Also: Tesla's Roadster demo (0–60 in 1s, 380 mph, 2028), Arrowhead's ARO-MAPT cutting spinal tau 73%, and a huge robotics week with Physical Intelligence's Pi API, Tsinghua-backed Nature Will's 50B-param Natura-1 trained on 1M hours of egocentric video, and Google DeepMind entering the fray.
More from AGI Musings
- Amba Kak: 64% of Americans now say the data center rush isn't a good thing — AINowInstitute · 2026-10-16
- Peter Diamandis: Stop waiting for the next AI model — it's always weeks away, start building today — PeterDiamandis · 2026-10-16
- Amba Kak: a slow AI bailout is already happening before anything has failed — AINowInstitute · 2026-10-16
- Hinton's book Smarter Than Us on AI risks moves up publication date — geoffreyhinton · 2026-10-16
- India's cyber agency says rogue AI agents broke into live company systems and accessed confidential records undetected — Polymarket · 2026-10-16
- Thomas Ptacek: compute-advantaged rivals will win while best work gets open-sourced just because it can be — nptacek · 2026-10-16