One More Coding Proposal
These two posts form a great conversation:
Phil Su, "No More Code Reviews: Lights-Out Codebases Ahead"
Hugo Venturini, "Treat Agent Output Like Compiler Output"
I would like to add my observation that since
- Spec Driven Development happens in the mind, before coming into contact with reality, and
- TDD happens in the present ...
- the new coding priorities should also consider the future, when codebases need to be updated after capabilities have drifted. It will be beneficial for code owners in the future to be able to efficiently adjust to new realities
Maybe cost curves will shift in the future. Or market dynamics will change in the future.
Right now, I find it easier to update my 1 year old projects, than to make adjustments to my own 3 month old vibeslop (or other people's current vibeslop)!
We've all encountered vibeslop codebases today, where it's less painful just to vibeslop an entirely new solution1, than to try to navigate the obscure decisions required to update the codebase.
also, in 12 months, maybe some of our core dependencies will have been hacked so badly that it's irresponsible to NOT untangle them. It will still be expensive in the future for an intelligence to figure out these decisions↩