"AI hasn't solved coding": easy tasks hide the bumps, warns dev in Anthropic debate
gerardsans · x · 2026-10-17
Pushing back on "AI has solved code," gerardsans argues straightforward code isn't the end of the difficulty spectrum—move to harder tasks to feel the bumps. He cites Anthropic's own service for fixing code it just generated: training data has gaps, harnesses fail, context windows drift, budgets end, and AI hits dead ends.
More from coding & agent
- Engineer's long-form guide: building your own personal software factory — maddiehfaulkner · 2026-10-17
- 45-year game coding veteran: agentic coding is the biggest productivity jump I've ever seen — IgorCarron · 2026-10-17
- PKU & Samsung: slow model coaching lifts fast agent's live game score from 13.51 to 24.19 — rohanpaul_ai · 2026-10-17
- Scale AI's AgentEnv RL environments framework lands on Hugging Face Hub — ben_burtenshaw · 2026-10-17
- CoPE adds one forecasting loss to GRPO for 2x sample efficiency in agent RL — youjiaxuan · 2026-10-17
- LeanGraphSearch mines Mathlib dependency graphs, lifting proof success to 81% — kaiwei_chang · 2026-10-17