v3.23.0 — What's new

The pipeline could carry a product all the way to a live URL without ever asking
whether it makes money. It asks now, before it decides to build.

• product-economics — the question no gate was asking. Every stage of this
  pipeline answers "can this be built safely": architecture reviewed, tests
  green, security signed off, deployed. None of them answered "should this be
  built at all, at a price someone will pay" — and silence there read as
  approval. The skill puts three numbers into the brief, before the
  recommendation that rests on them:

  • contribution margin — price minus the variable cost of serving ONE user.
    For anything LLM-backed the token line decides it, computed at p95 usage
    rather than the mean: flat-rate plans are priced by the tail, and the tail
    is what arrives. A negative margin cannot be fixed by volume.
  • the price's basis — cost-plus, competitor-anchored (named, with the date
    it was checked), or value-based. One of them: the one that actually decided.
  • reachable buyers, bottom-up — enumerated, at a realistic price, times a
    fraction with a NAMED channel. A slice of someone else's market report is
    arithmetic, not evidence.

  It reuses the brief's own [source: …] / [assumption] markers rather than
  inventing a second vocabulary, because arithmetic launders provenance: two
  assumptions multiplied together produce a margin that reads like a
  measurement. A number nobody knows does not become an assumption — it becomes
  a kill-criterion with a threshold and the cheapest test that resolves it.

  Building at a loss stays allowed. Discovering it in month four does not.

Full notes, including every fix and the reasoning behind it

Full notes for this release — what changed, and why each change was made.