Dev argues AI trained on all papers could design programming languages better than humans
zack_overflow · x · 2026-10-08
Developer zackoverflow argued that almost none of TypeScript's ideas are novel — structural subtyping, intersection/union types, dependent types, gradual typing all come from decades of research (Rust's type system likewise borrows from FP languages). TypeScript succeeded by finding the right combination of existing ideas that integrated well with JavaScript's dynamic base. His takeaway: a model trained on every research paper would likely do this better than a human.
More from coding & agent
- Did Anthropic quietly drop Claude subscription access for Agent SDK apps? Dev finds doc change — iannuttall · 2026-10-08
- 3 AI agent-generated optimizations merged upstream into SGLang-Omni — ChengleiSi · 2026-10-08
- DuckDB shows a database with no data: a few hundred KB catalog over your data lake — RexDouglass · 2026-10-08
- Open-Source Async Agent: Keep Chatting While Tools Are Still Running, With ReAct Emerging From an Event Loop — wuj · 2026-10-08
- Dev open-sources smoke effect from Replit-built Drift game with agent prompt workflow — techartist_ · 2026-10-08
- Engineer Pushes Back on AI Code Analogy: Topology-Optimized Parts Look Cool but Are Unusable — rms80 · 2026-10-08