“A drifting accelerator doesn’t end up somewhere between the two approaches; he ends up a vibecoder by default, without the harness.”
The drift is the honest observation here. Nobody decides to stop reading their code. They skim one diff because it is late, then another, and six weeks later nobody on the team can explain why the system does what it does. Chotoborski is right that the model in your head is the asset. Ship the code and lose the model and you own a codebase you have to re-derive from scratch every time it breaks.