workflow-optimization 3

Run a blameless post-mortem on this entire conversation to improve future performance. 1. Document the entire process so far (what you did, how, what you found, next steps, etc.). 2. List successes: techniques / approaches you discovered that worked well. Examples: tools, code snippets, prompt structures, planning techniques. Share what change to the environment / prompts will make it easier to repeat these successes in the future. 3. List problems faced: failures, inefficiencies and mis-alignments. E.g. commands that failed or behaved unexpectedly, corrections, more steps than necessary, where you adhered to the letter not spirit, took shortcuts that compromised quality, etc. Dig deep for root causes. Mention the PRACTICAL impact. Suggest pragmatic & safe fixes (if any) to prompts, skills, or environment (e.g. tools, .env) at a root cause level - preferably that resolve **entire classes/patterns of failures**, not just a specific instance. Create or append to `notes.md` as `## Post mortem (%d %b %Y)` with today's date.

A software services firm once asked us, “How come we are not able to staff projects quickly, even though we have a lot of people on the bench?” There were a bunch of reasons, but among those, we found something interesting. They were implementing two policies that were logical on their own, but disastrous together. (The bench is where programmers sit when they are not on a project.) Here’s how they work. When a project starts, the project manager requests resources (people) for the project. HR passes on matching CVs to the project manager, who approves or rejects them, in consultation with the client. ...

O’Reilly hacks.