Leaked Anthropic roadshow slide shows superexponential algorithmic progress; critics say it's mostly data
sebkrier · x · 2026-11-14
A former Anthropic employee says a slide from the IPO roadshow showing pretraining compute efficiency (log scale) improving superexponentially over time was a big factor in his decision to leave — Mythos was just a point on that curve and could be retrained with a fraction of the original compute. Adam Karvonen pushes back, arguing terms like "compute multiplier" and "algorithmic improvements" are a psyop hiding that most gains are data-related.
More from Companies & People
- Weekly AI recap: Nvidia's six-week slide, Anthropic IPO pricing, Google's four releases — xiaohu · 2026-11-14
- Job hunter's trick: fix a real issue in a startup repo, send the PR as evidence — minhnguyen · 2026-11-14
- Bill Ackman Defends Jensen Huang Amid Scandal, Citing a Personal Favor — lifebypixels · 2026-11-14
- Jessica Lessin: 'As Jensen created the AI boom, it created him' — NinaDSchick · 2026-11-14
- Coinbase CEO: merged 46 PRs across 10 repos in 14 days part-time by talking to agents — ycombinator · 2026-11-14
- Switching from global health to AI safety: one grantmaker's candid career reflection — austinc3301 · 2026-11-14