Researcher sees no evidence Anthropic's 2x R&D acceleration is pulling ahead
adam_karvonen · x · 2026-11-21
Adam Karvonen says he doesn't see evidence that Anthropic's year of 2x AI R&D acceleration is creating separation: in Nov 2025 Opus 4.5 and GPT 5.2 are equally smart (Opus better at understanding intent), and he expects Nov 2026 to look the same — Opus 5.5 and GPT 6.2 equally smart, with the same intent-understanding edge. Frontier labs stay in lockstep despite the spending difference.
More from AGI Musings
- AI could systematically revisit great math papers only a handful have read — littmath · 2026-11-21
- Ex-Anthropic researcher Jacob Coxon resigns blasting labs; Dario's response sparks backlash — sjgadler · 2026-11-21
- New post: OpenAI and Anthropic should actually pace the frontier, not just talk — mealreplacer · 2026-11-21
- Joe Edelman on Telocracy: purpose-driven institutions accountable to AI auditors — edelwax · 2026-11-21
- Nobel Laureate David Baker Says Protein Design Has 'Only Scratched the Surface' — DaveJuergens · 2026-11-21
- Ex-AI-ethics member reflects: rejecting LLMs cost his community money and influence — KLdivergence · 2026-11-21