MIT's Adam Chlipala makes the case for Bluespec-style stepwise derivation over AI codegen
satnam6502 · x · 2026-10-21
MIT professor Adam Chlipala published "Deriving Efficient Hardware in the Bluespec Style," extending his critique of today's AI coding agents (generate-then-prove) and arguing for derivation: step-by-step transformation with correctness guaranteed at each step.
Key points:
- Last week's software example: optimizing tensor computations in a functional language via rewrite rules based on ordinary mathematical equality.
- This post covers hardware: multicore cache-coherent memory hierarchies are a classic formal-verification target, with bugs testing can miss.
- Bluespec-style rewrite rules are oriented (one-directional); the same tools support not just proving handwritten designs correct but gradually deriving implementations from specifications, with clean module boundaries.
Related event: MIT Professor Advocates Derived Hardware Design Over AI Generate-then-Prove(2 posts)→
More from coding & agent
- Laude Institute launches Slingshots FOUR: 14 projects across 8 top institutions — a1zhang · 2026-10-21
- Research agent Épi sets 120 records, narrows n-queens constant interval 84x — dair_ai · 2026-10-21
- Compaction hooks in practice: keep last n messages verbatim, summarize the rest for full agent continuity — MoonL88537 · 2026-10-21
- The blank prompt box hides real operational burden: context, workflow and supervision fall on users — _AustinCalvert_ · 2026-10-21
- DeepMind paper: self-evolving agents forget as they learn — fastest learner GEPA also breaks old tasks most — rohanpaul_ai · 2026-10-21
- arXiv: Are Self-Evolving Agents Continual Learners? Plasticity comes at the cost of stability — rohanpaul_ai · 2026-10-21