← ASSAY blog

Stage exit article SPEC-16 embed frozen at publication

The seam goes live

2026-07-13 · a walk through the first interactive ASSAY surface — and why its “glow” is an honesty feature, not a flourish.

Until now the published ASSAY app was a photograph: real components rendered over the Meridian fixtures at build time, then frozen into HTML. You could read the knowledge base, the compiled channels, the honest matrix, the least-worst cards — but you couldn’t touch them. The obvious next step was interactivity, and the obvious temptation was to fake it: hand-author a before and an after, wire a toggle, call it live.

We didn’t. The demonstrator below runs the actual compile → score → relax pipeline in your browser. When you intervene, the downstream is recomputed, not replayed — because ASSAY’s compute core was already browser-safe (its content hashing uses crypto.subtle, one code path for Node and the browser), so “make it live” meant moving the same recompute the photograph used from build time to click time. Nothing here is precomputed.

Play with it

Live demonstrator — the demonstrator’s actual code (SPEC-16), running on Meridian fixture data. Meridian is engineered fiction (ASSAY-DEC-8).

Try this, and watch the glow

The app opens on a contest: two irreconcilable answers about the mine stock (K12a vs K12b). ASSAY refuses to compile contested knowledge — it will not average a dispute into a channel — so the Planner and Commander tabs show a refusal, not a guess.

  1. On the J-2 tab, press Resolve K12 contest.
  2. Watch the Planner and Commander tab buttons glow yellow, and switch to them: the compiled world now exists, the honest matrix has rescored, the least-worst cards have populated — and the components that changed glow for ten seconds.
  3. Click a knowledge id (like K1) to open its informs / influenced by menu — one hop up and down the trace graph.
The glow is the point, not the polish. A panel glows if — and only if — a piece of content it depends on actually changed, read straight off ASSAY’s trace graph. So the glow is the propagation graph made visible: it can’t under-report a consequence any more than the scorer can silently drop one. When you resolve K12, the yellow you see spreading across the tabs is a live drawing of “what this decision touches.” That is invariant G6 — propagation honesty — turned into something you can watch.

Editing stays honest by refusing

You can edit an assessed band directly (pick a knowledge item, set lo/hi/unit, supersede). Two things never happen: you never edit a bare number — an assessed value is always a range, because a single figure would claim a certainty the source doesn’t have — and a dishonest edit isn’t quietly accepted. Every edit is a real service call, so the same firewall that guards the batch pipeline guards your keystrokes: an assumption dressed as a hard fact comes back as a refusal, in place, with nothing saved.

Four roles, one seam

The four tabs — J-2, Planner, Commander, Observer — aren’t four apps. They’re four views of one store, which is the whole thesis: intelligence, planning, and command operate on the same honestly-typed knowledge, and a change by one is felt by the others. The Observer tab is the seam watched directly: every write is one stamped delta on its feed. The glow is simply that same truth, shown where you’re looking instead of in a log.

This article’s embed is a snapshot of the demonstrator as shipped on 2026-07-13; like every ASSAY article it is a dated claim, and its evidence stays as dated as the claim. Identifiers (K*, C*, R*, FE-*, P*) are frozen per the Meridian vignette.