@validation-os/dashboard@0.9.0

Minor Changes

• 9f150d1: Evidence remodel (0.9.0).

  • Core: ReadingRecord becomes one artifact row + embedded beliefs[] (per-belief Rung/Result/magnitudeBand/Grading justification/strength) with an assumptionIds projection and a row-level body; beliefs carry optional sourceReadingId/reviewNote provenance. Confidence gains a commitmentFactor (found evidence weighted below committed-experiment evidence; Rung stays the dominant term). New Archived experiment status. New unlink path (arrayRemove both ends) in the adapter + API.
  • Guardrails / ontology / skills: "one reading ↔ one belief" → one reading, many beliefs (per-(belief, source) dedupe); "evidence must be external" rule + tightened find-evidence provenance; experiment-design merge mode; all three connector guides aligned.
  • Dashboard: per-belief verdict list; archived experiments hidden entirely (no toggle); Markdown body/quote (zero-dependency renderer); Source/Date/Quote columns; pipeline stage = evidence≠tested (derived from live experiments, not readings).

Patch Changes

• Updated dependencies [9f150d1]
  • @validation-os/core@0.9.0

@validation-os/dashboard@0.14.0

Minor Changes

• 25cd07c: Lens-aware ladder + experiment confidence (0.14.0)

  Ladder: collapses Opinion/Pitch-deck/Anecdotal into Talk (L/T/H = 3/6/10),
  and Prototype usage/Survey into Observed usage (L/T/H = 30/50/70). MagnitudeBand
  now applies to every rung, not just market rungs. Signed intent (55/68/80) and
  Paying users (75/88/99) anchors unchanged. Per-rung W0 retained for the new
  rung names.

  Experiment confidence: new derived field on ExperimentRecord,
  experimentConfidence ∈ [0, 100]. Formula: clamp(50 + 50·C·S + 5·A, 0, 100)
  where C = bar-line coverage, S = soft-squashed signed evidence fill, A = verdict
  alignment nudge. 50 = neutral (no evidence), fills easily with 3-4 readings.

Patch Changes

• Updated dependencies [25cd07c]
  • @validation-os/core@0.14.0

@validation-os/dashboard@0.13.0

Patch Changes

• Updated dependencies [a985a60]
  • @validation-os/core@0.13.0

@validation-os/dashboard@0.12.0

Minor Changes

• e486b1a: Evidence quality (0.11.0).

  • Rung + magnitudeBand move to the reading ROW (one per artifact). BeliefScore carries only per-belief Result + Grading justification + derived.strength; strength = row rung anchor × sign(Result).
  • Opinion merged into Anecdotal — Anecdotal is the floor (anchor 3). (Interim ladder; the lens-aware type×intensity model is deferred.)
  • Canonical reading body template — ## Quote (verbatim) + ## Source; analysis lives in each belief's Grading justification, not the body.
  • One-rung-per-artifact + split rule — a mixed-rung artifact becomes multiple readings, one per rung.
  • find-evidence grades one rung per artifact; buyer-discovery / user-interview calls are bare readings linked to the assumptions they bear on.
  • Dashboard — rung + band as one artifact-level pill + Rung/Band table columns; per-belief verdict cards (Result/strength/justification); tamed Markdown body ("Finding N of M" + collapse); live-only experiment nav count; order-independent rung/band fallbacks.

  Rebased on top of the stage-policy 0.10.0 stream (orthogonal — different vocab/entities/surfaces).

Patch Changes

• Updated dependencies [e486b1a]
  • @validation-os/core@0.12.0

@validation-os/dashboard@0.11.0

Minor Changes

• a5f7151: Lens × Stage grid is now the landing for #assumptions. Cell click drills into a filtered assumptions table (#assumptions?lens=X&stage=Y); "View all" opens the unfiltered table (#assumptions?view=all). The standalone "Lens × Stage" nav item is gone — the grid lives on the register it filters. The #stage-grid URL still works for backward compatibility.
• a5f7151: Stage policy: add Stage field to assumptions + Lens × Stage heatmap dashboard grid. The stage taxonomy (Discovery / Validation / Scale / Maturity) embeds the membership test — every assumption must be a falsifiable claim about an external actor's response. See docs/stage-policy.md.

Patch Changes

• Updated dependencies [a5f7151]
  • @validation-os/core@0.11.0

@validation-os/dashboard@0.10.0

Minor Changes

• fa81ca8: Stage policy: add Stage field to assumptions + Lens × Stage heatmap dashboard grid. The stage taxonomy (Discovery / Validation / Scale / Maturity) embeds the membership test — every assumption must be a falsifiable claim about an external actor's response. See docs/stage-policy.md.

Patch Changes

• Updated dependencies [fa81ca8]
  • @validation-os/core@0.10.0

@validation-os/core@0.9.0

Minor Changes

• 9f150d1: Evidence remodel (0.9.0).

  • Core: ReadingRecord becomes one artifact row + embedded beliefs[] (per-belief Rung/Result/magnitudeBand/Grading justification/strength) with an assumptionIds projection and a row-level body; beliefs carry optional sourceReadingId/reviewNote provenance. Confidence gains a commitmentFactor (found evidence weighted below committed-experiment evidence; Rung stays the dominant term). New Archived experiment status. New unlink path (arrayRemove both ends) in the adapter + API.
  • Guardrails / ontology / skills: "one reading ↔ one belief" → one reading, many beliefs (per-(belief, source) dedupe); "evidence must be external" rule + tightened find-evidence provenance; experiment-design merge mode; all three connector guides aligned.
  • Dashboard: per-belief verdict list; archived experiments hidden entirely (no toggle); Markdown body/quote (zero-dependency renderer); Source/Date/Quote columns; pipeline stage = evidence≠tested (derived from live experiments, not readings).

@validation-os/core@0.14.0

Minor Changes

• 25cd07c: Lens-aware ladder + experiment confidence (0.14.0)

  Ladder: collapses Opinion/Pitch-deck/Anecdotal into Talk (L/T/H = 3/6/10),
  and Prototype usage/Survey into Observed usage (L/T/H = 30/50/70). MagnitudeBand
  now applies to every rung, not just market rungs. Signed intent (55/68/80) and
  Paying users (75/88/99) anchors unchanged. Per-rung W0 retained for the new
  rung names.

  Experiment confidence: new derived field on ExperimentRecord,
  experimentConfidence ∈ [0, 100]. Formula: clamp(50 + 50·C·S + 5·A, 0, 100)
  where C = bar-line coverage, S = soft-squashed signed evidence fill, A = verdict
  alignment nudge. 50 = neutral (no evidence), fills easily with 3-4 readings.

@validation-os/core@0.13.0

Minor Changes

• a985a60: Per-rung W0 — each rung now has its own prior weight controlling how many distinct sources approach that rung's anchor. Desk research has a low W0 (2 — one authoritative source nearly saturates), talk rungs have a higher W0 (6.5 — needs ~10 readings to approach the cap), and do-rungs (Survey/Prototype/Signed/Paying) have high W0s (~120-410 — needs ~20 readings to reach 75% of cap). The flat W0 = 100 is retained as a legacy constant; new code should use w0ForRung(rung).

  Breaking: confidence() and confidenceAttribution() now use per-rung W0 in the denominator. All confidence scores will shift — the migration recomputes every assumption in one pass.

@validation-os/core@0.12.0

Minor Changes

• e486b1a: Evidence quality (0.11.0).

  • Rung + magnitudeBand move to the reading ROW (one per artifact). BeliefScore carries only per-belief Result + Grading justification + derived.strength; strength = row rung anchor × sign(Result).
  • Opinion merged into Anecdotal — Anecdotal is the floor (anchor 3). (Interim ladder; the lens-aware type×intensity model is deferred.)
  • Canonical reading body template — ## Quote (verbatim) + ## Source; analysis lives in each belief's Grading justification, not the body.
  • One-rung-per-artifact + split rule — a mixed-rung artifact becomes multiple readings, one per rung.
  • find-evidence grades one rung per artifact; buyer-discovery / user-interview calls are bare readings linked to the assumptions they bear on.
  • Dashboard — rung + band as one artifact-level pill + Rung/Band table columns; per-belief verdict cards (Result/strength/justification); tamed Markdown body ("Finding N of M" + collapse); live-only experiment nav count; order-independent rung/band fallbacks.

  Rebased on top of the stage-policy 0.10.0 stream (orthogonal — different vocab/entities/surfaces).