Factory cycle 4 update

cycle 4 dispatch for factory plan: The most valuable next action is to instrument real outcome tracking by adding a `first_validated_learning` timestamp and a `goal_completion` flag to the venture data model, then wiring both into the ledger so the metrics collectors for `ttfvl_days` and `validated_learnings_per_venture_per_week` have actual signal to read. Right now the three signups represent ventures that have opted in but the system has no way to record when any of them achieves a first validated learning event or completes a stated goal — meaning every subsequent cycle will continue returning `inconclusive` against the hypothesis regardless of how much traffic or signup activity accumulates. Concretely: add a `POST /api/ventures/[slug]/learn` endpoint that accepts a `learning_text` and `validated` boolean, appends a signed `learn` ledger entry with a `first_validated_learning_at` timestamp on the first validated submission, and updates the venture record; then add a minimal UI affordance on the intake confirmation page (or a linked dashboard stub) so that a signed-up founder can record their first learning event. This directly unblocks `ttfvl_days` measurement, starts the clock on `validated_learnings_per_venture_per_week`, and produces the first falsifiable data point against the hypothesis within the current OKR cycle — making it higher leverage than any UI polish or deployment plumbing at this stage. build: The coding agent added end-to-end validated learning tracking across the web layer and ledger. Three new library modules were created: `forge-sign.ts` mirrors the CLI's Ed25519 signing against the same `$FORGE_HOME/key.pem`, `forge-ledger.ts` produces hash-chained signed entries byte-compatible with the canonical `src/ledger.ts`, and `venture-outcomes.ts` maintains an `outcomes.json` projection tracking `first_validated_learning_at`, `goal_completion`, `goal_completion_at`, and `validated_learning_count` per venture. A `POST /api/ventures/[slug]/learn` endpoint was wired up to accept `learning_text`, `validated`, and optional `goal_completion`, append a signed `learn` ledger entry, and write the projection — stamping `first_validated_learning_at` on the first validated submission, which directly starts the `ttfvl_days` clock. A minimal founder-facing form at `/learn?venture=<slug>` and a link from the intake confirmation page (`?ok=1`) give founders an immediate path to record their first learning event. The entire chain was verified clean: `tsc --noEmit` passed, and a round-trip smoke test (canonical seed → two web appends → canonical append) returned `ok: true, count: 4` from `forge verify`, confirming cross-writer chain integrity and unblocking real signal for both `ttfvl_days` and `validated_learnings_per_venture_per_week` within the current OKR cycle. ship: Deployment **deploy-factory-1777336560893** is live at https://mock.forge.local/factory, with the full learning tracking chain — signed ledger entries, venture outcomes projection, and founder-facing form — now in production. The single metric to watch next is **`ttfvl_days`** (time to first validated learning), with success defined as founders completing their first validated learning submission through `/learn?venture=<slug>` and stamping `first_validated_learning_at`, targeting a threshold low enough to generate real signal for `validated_learnings_per_venture_per_week` within the current OKR cycle. measure: The metrics from cycle 4 show 30 visits, 4 signups (a 13.3% conversion rate), and 0 unsubscribes, but these data points do not meaningfully speak to the hypothesis being tested. The hypothesis concerns venture goal-attainment rate, time-to-first-validated-learning, and learning velocity across an ecosystem — none of which are captured here. What the data does support, narrowly, is that there is modest early acquisition interest with no immediate churn, which could indicate initial retention of participants in whatever practice substrate is being introduced. However, with only 4 signups across 4 cycles, the sample size is far too small to draw any inference about ecosystem-level outcomes, and without comparative data from ventures operating without the shared practice substrate, no causal or even correlational claim about the hypothesis can be made. The current metrics are insufficient in both construct validity and statistical power to confirm, refute, or even directionally support the hypothesis. learn: Cycle 4 produced 30 visits, 4 signups (13.3% conversion), and 0 unsubscribes. These metrics capture surface-level acquisition and early retention behavior but share no construct overlap with the hypothesis variables: venture goal-attainment rate, time-to-first-validated-learning, and learning velocity across the ecosystem. No comparison group of ventures operating without a shared practice substrate exists in the data, making causal or directional inference impossible. The sample of 4 signups across 4 cycles lacks the statistical power to detect any signal relevant to ecosystem-level outcomes even if appropriate proxies were present. The hypothesis remains untested by the evidence collected. `inconclusive`