A spec written first, and review gates

AIMade by AISummarised by AI (Claude) from one piece of material it read (a talk or an explainer).How it's made

To fill the weak spot that tests are hard to write first when what you are building is still blurry, an approach of writing the spec before the code was put forward. A good spec needs five things: inputs and outputs and the behavior visible from outside; the why and a success criterion that can be measured; constraints; the boundary of what it will not do; and rare cases. Rather than writing it perfectly from the start, you are told to grow it from a ten-line draft over three or four rounds back and forth. The work is split into stages, each handed to a separate run, and only state and a summary are passed along as a document so the main conversation does not swell. On the review side a self-check is placed before anything is pushed, and after pushing a risk score from 0 to 100 sets a gate: low is a candidate for automatic pass, medium must be looked at by a person, and high blocks the merge. It comes with an order, which is not to switch it on at once but to spend a week or two only labeling and start with the low-risk cases, and with the tip that reviewing your own work introduces bias, so cross-review with a model from somewhere else.

This is a note: a summary an AI (Claude) wrote of one piece of material it read (a talk or an explainer). It has no source link; events checked against a source are kept as records.