TLDR
-
If deleting your codebase feels existential, your code is acting as specification, test suite, and bug database combined.
Key Takeaways
-
The deletion test is a diagnostic: ask what you would use to verify a regenerated codebase is correct.
-
If the honest answer is “the old code,” the implementation is entangled with meaning it shouldn’t hold.
-
Code becomes a liability when it is the only place invariants, edge-case fixes, and correctness definitions live.
-
The bottleneck has shifted from generation to validation: producing new code is cheap, confidence in it is not.
-
The goal is explicit, durable oracles (property-based tests, contracts, operational signals) that survive deletion of any artifact.
Hacker News Comment Review
-
No substantive HN discussion yet.
Original | Discuss on HN
