In film, "we'll fix it in post" is what you say when something went wrong on set and you don't want to redo it. AI research has made it our entire …

The argument is that post-hoc patching of trained models, analogous to fixing film mistakes in post-production, is not a viable scientific methodology for AI. The ICML oral paper contends that alignment and safety properties need to be designed into training rather than retrofitted, which has direct implications for how practitioners structure model development pipelines.