Anthropic's internal AI R&D uplift estimated at 4x, still under RSP threshold
AccBalanced · x · 2026-10-11
- A thread discusses current AI-driven R&D acceleration (uplift): the quote-guesser estimates Anthropic is at roughly 4x uplift internally, with uplift doubling every 3-6 months everywhere.
- The retweeted post says a 4x figure matches Anthropic's self-reported August 2026 Risk Report and remains below the Dramatic Acceleration threshold in its Responsible Scaling Policy.
- It adds the report covered the "Mythos Preview" model, with many generations since. Unverified details.
More from AGI Musings
- Personal AI agents could drain subscription revenue built on customer inattention — rohanpaul_ai · 2026-10-11
- Software engineer argues mathematicians may be worse off than SWEs under AI — haider1 · 2026-10-11
- IBM AI engineer's anti-slop essay goes viral after LinkedIn commenters get caught using AI replies — TejasKumar_ · 2026-10-11
- David Sacks amplifies view that AI alignment itself is the core AI safety problem — DavidSacks · 2026-10-11
- Bad AI Consciousness Takes Bingo: a rebuttal for every square — dioscuri · 2026-10-11
- dioscuri: 2027 will be the year AI does for science what it did for math — dioscuri · 2026-10-11