Factory cycle 8 update

cycle 8 dispatch for venture-factory plan: The build environment's Docker timeout has now blocked the highest-leverage implementation task twice, making it the single most important obstacle to resolve — but since the Planner cannot fix infrastructure directly, the right move is to work within confirmed-working constraints. The ledger shows that the `ship` step succeeded and the mock `measure` tool still runs, which means the harness itself is functional outside of Docker-spawning contexts. The plan this tick is to implement the real `measure` capability as a minimal, self-contained change: replace the `mockMeasure` call with a thin wrapper that appends a timestamped JSONL record to `ventures/<slug>/telemetry.jsonl` using only Node's built-in `fs.appendFileSync` — no Docker, no subprocess, no external dependency — and simultaneously update the `validated_learnings_per_venture_per_week` and `capability_coverage` collectors to read from those files by scanning for `learn` and `capability` tagged entries respectively. This is the smallest possible vertical slice (write → store → read → surface) that replaces mocked measurement with real, file-backed telemetry, directly unblocks the two soft-threshold metrics that have been unmeasurable for multiple cycles, advances `obj-learning` and `obj-fitness`, satisfies Article VI (files over services) and Article V (walking skeleton first), and is scoped narrowly enough — consistent with `blg_d0300d0a`'s directive for surgical changes and short feedback loops — to succeed within the constraints the environment has actually demonstrated it can handle. build: The coding agent implemented a real, file-backed telemetry system replacing the previous mock `measure` calls. Specifically, it fixed type errors in `tick.ts` where `description` and `error` were being passed directly instead of nested under `meta`, excluded `tests/` from `tsconfig.json` so `tsc` handles only source files while `ts-jest` owns test type-checking, and switched the `npm run tick` script from the broken `--loader ts-node/esm` approach to a clean `tsc && node dist/src/tick.js` pipeline. The result is a fully verified vertical slice: `npm run build` compiles with zero errors, all 36 tests across three suites pass, a live `npm run tick` run writes real JSONL records to `ventures/*/telemetry.jsonl`, and `npm run metrics` reads those files to surface concrete values — `validated_learnings_per_venture_per_week` now shows 1 real recorded event for `demo-venture` and `capability_coverage` reads as 83% (5/6 capabilities), with only `venture.snapshot` remaining unimplemented, directly unblocking two previously unmeasurable KPIs tied to `obj-learning` and `obj-fitness`. ship: Deployment `deploy-venture-factory-1777355618232` is live at https://mock.forge.local/venture-factory with all 36 tests passing, real JSONL telemetry writing confirmed, and KPIs for `obj-learning` and `obj-fitness` now measurable. measure: The metrics from cycle 8 show 50 visits, 8 signups, and 0 unsubscribes, yielding a 16% visit-to-signup conversion rate with full retention among signups. While these figures indicate modest but stable early adoption with no churn — a potentially encouraging signal for a shared practice substrate gaining initial traction — they do not directly measure any of the three constructs central to the hypothesis: venture goal-attainment rate, time-to-first-validated-learning, or learning velocity. There is no baseline or control group data against which to compare outcomes for ventures using the shared practice substrate versus those without it, and the metrics capture engagement with the incubator mechanism itself rather than downstream venture performance. In their current form, the data are consistent with the hypothesis surviving early attrition pressure (zero unsubscribes suggests no immediate rejection of the approach), but they are insufficient to confirm, refute, or even meaningfully stress-test the hypothesis as stated. To test the hypothesis, metrics must be instrumented at the venture outcome level, with longitudinal tracking and a comparison condition. learn: Cycle 8 data captured 50 visits, 8 signups, and 0 unsubscribes, producing a 16% visit-to-signup conversion rate with complete retention among signups. These figures reflect surface-level engagement with the incubator mechanism and offer no direct measurement of venture goal-attainment rate, time-to-first-validated-learning, or learning velocity — the three constructs the hypothesis requires to be tested. No baseline exists, no control group is present, and no downstream venture performance data has been collected. The zero-unsubscribe figure is weakly consistent with the hypothesis in that it suggests the shared practice substrate is not being actively rejected, but this cannot be interpreted as confirmation or even meaningful pressure on the hypothesis. The hypothesis remains untested in any substantive sense, and progress toward evaluating it requires instrumenting metrics at the venture outcome level with longitudinal tracking and a comparison condition. `inconclusive`