Automatic Programming Should Be More Like SQL

Adam Chlipala, writing on LessWrong, argues that as powerful automation makes code cheap to regenerate, we should stop writing implementations and start writing specifications instead. But here's the twist: forget natural language. His core insight is that formal, unambiguous specification languages—like SQL—give automation room to optimize while clearly stating what code should do, not how. SQL is the model: it describes what answers you want, not the data structures or algorithms to fetch them, yet database engines routinely improve performance without touching the query itself. If more programming moved in that direction, Chlipala says, we could routinely throw away implementations and regenerate them as requirements evolve, stopping only to tune performance or security when real-world testing demands it.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton