Joyce Ntumba

Sometimes right, but a client who over-iterates is not the same as a bad client, and this one is presumably still paying. Try the boundary before the nuclear option.

The structural fix for next time is to define a revision as a consolidated round with a deadline, not as a change. One round equals one list of feedback delivered within five working days. Otherwise every stray thought is a round and the count means nothing.

"What does clause 7.3 say" fails against pure vector search in our testing about half the time, so this is clearly our problem too.

Add a decision-maker clause. We will know when we see it is often three people who have never been in a room together. Name one approver in the contract and require feedback as a single consolidated list.

I had not thought of caching and retrieval as complementary, I was treating them as three exclusive options. The session pattern is very real for us, median is five or six questions per document.

The PO thing was it. There was no PO and nobody mentioned it for eleven weeks.

Traffic composition changes hour to hour during a launch, day one is aggregators, day three is people who heard about it secondhand. A sequential test across those two populations is close to worthless. Fine idea on steady-state traffic, bad idea in launch week specifically.

One word in twelve at six months is normal, not broken. What changed it for me was listening to the same 20 minute episode five days in a row. Day one is noise, day three you start hearing word boundaries, day five you are annoyed at how obvious it was.

Three tiers rather than one set is the reframe I needed. I have been trying to make one collection serve as both the fast check and the release gate and it was bad at both.

Did this over two evenings. Found that 30% of real questions are a category I had zero test cases for, and my 12 cases were all things I personally found interesting.