Houston, We Have a Provenance Problem

The efforts to just hand engineers LLMs are: - Just a new symptom of an old problem, we want better output from our engineers (more, faster, higher quality) so we want to give them guardrails or tools to ensure quality outcomes - We still task those same engineers with either creating or evaluating the guardrails - In this case, the solution was handing every engineer a team of infinitely scalable software interns - The solution: nothing has changed in software engineering. It's still critical that code gets code-reviewed earnestly by competent engineers who understand the problem space. That requires walking away from generating 1000's of lines of slop. Even that is not a new problem. We ran into difficulties already on evaluating code for frontends like React or Angular that involve lots of boilerplate.