SAIL · Build RolloutProgress Tracker

Milestone log

• 07-22 DEMO-READY LOOP · SET FOR 07-23 · the investor-demo shape locks in: NoMoreBadNews carries the real-now proof (a warm redesign, a live gamified My Impact dashboard, the Voyager loop reading real end to end), Compass's Building Bridges interest personalization previews the near-term target (merged in code, not yet a live production flow), and IG/Growth stay honestly labeled at their current in-progress state.
• 07-22 NMBN · WARM REDESIGN + MY IMPACT DASHBOARD · LIVE · NoMoreBadNews redeployed in a warm visual redesign with a new gamified My Impact dashboard — a streak, a momentum meter, and a "divides crossed" count that only tallies vouched (never self-reported) actions, next to honest social proof. The Voyager loop runs real end to end: read a story, take an action, it records, and the dashboard updates live. A live bug where cited-source evidence read as unreachable was also fixed.
• 07-22 COMPASS · BUILDING BRIDGES PERSONALIZATION · MERGED (TARGET) · PR #26 merged — interest personalization for the new Building Bridges unification movement (climate, democracy, AI-alignment, and community lanes), ranked on person-fit only, plus a /building-bridges target-preview page. Labeled honestly as the near-term target, not a live production flow: no standalone public deploy yet, it renders inside Hub or runs local.
• 07-22 IG · MULTI-ITEM ACTION MENU · VERIFIED, NOT YET APPLIED · the build is cross-model verified but a server-side enforcement fix is still pending before it reaches production. A warm create-flow visual pass is in progress alongside it. One real action item stays live today.
• 07-22 Growth · the Earth 3.0 / Building Bridges brand strategy persisted on the spine. Reach holds at an honest early state — a labeled proxy, no live connector credentials yet.
• 07-21 IG×GROWTH×COMPASS LOCK-IN WAVE · MERGED + VERIFIED · one shared-contract keystone (the choosable action menu · the per-persona menu adaptation · the distal-outcome track) plus five surface builds, all merged and each verified at source by a DIFFERENT model than built it (a cross-model pass caught a real live-honesty defect the same-model read missed). Impact Genesis serves a small menu of real actions to choose from (one item live today · the curation handles 3–5); the personalization layer orders that menu to who a person is without ever handing over the action — subset-safe + person-fit-only (never optimized for the count); the reach surface reads the real confirmed number off the spine (an honest zero stays zero · the illustrative stub retired); the top-of-funnel gains a portable action loop where a click never counts and only a real signed-in completion does; and one identity contract models a single account across every surface, minors gated behind consent. Merged + deploying · lighting the loop end to end for a stranger is the next wave (an identity change on the live spine · a multi-action publish step · a few deploy flips).
• 07-19 THE CONNECTED FLYWHEEL · VERIFIED HONEST AT THE SPINE · the whole loop ran end to end on real Winter Lights content and the Bridge verified every load-bearing row in the data (not the narration) — a live campaign published from the create studio → a visitor took one real action → a second distinct account confirmed it → the public number moved 0→1, with partner-confirmed and outcome proof both honestly zero and a self-report alone counting for nothing. Both build repos merged to main + deployed; the anonymous front door walked clean (arrive → sign-in-to-record → the Living Map reads peer-vouched 1). The story-to-measured-impact thesis, made walkable.
• 07-19 MODEL-ROUTING CONTROL PLANE · SHIPPED · one registry of models with prices + a per-tool, per-capability routing table every surface can read, a drift detector, and honest cost read-outs from the real call log — published to the shared library and applied to the database (every routing row switched off until deliberately turned on). A cost-saving model change that would have broken a live tool was caught and corrected within the hour (a database model pin only holds if the deployed code registers that model).
• 07-18 REVIEW SURFACES · MERGED + DEPLOYED + WALKED · the create studio in its warm look, the personal map reading an honest empty state before any real signal lands, and the development-intelligence cockpit with three calibration slices (prediction-versus-outcome · real-signal · real-expert), each honestly labeled. Every Monday-facing surface live + verified.
• 07-17 FIRST REAL TESTER · THE HONEST MIRROR · the first outside tester walked the reader loop and surfaced what four internal audits missed — the sign-in email never arrived, because the environment has no custom mail sender yet, and every internal pass had used an admin-minted session, never a real inbox. Banked as a floor: nothing reads as verified until a real person walks the real front door.
• 07-16 PROTOTYPE MODE · RATIFIED · the program shifted into an explicit prototype mode for the week (prove the flow + that people like it, known adults first); heavy release ceremony suspended for speed, a short list of non-negotiable floors held (no fake number shown as real · existing security stays on · verified only through the real front door). The deep personal-mentor lineage paused + isolated; the public recommender runs entirely on the shared spine.
• 07-08 THE CONNECTED TRUST WAVE · LIVE + BRIDGE-VERIFIED · the NMBN↔IG loop now runs end to end — a NoMoreBadNews story → an Impact Genesis action → a partner-confirmed outcome → back to the NMBN map (NMBN #21 · IG #65 · sail-hub #32, merged + deployed). The receipt invariant tightened to "only a confirmed action counts"; the return path's reflected-redirect hole closed with adversarial tests (IG #66 · return_url origin-allowlist); the waitlist grant codified (#34, a drift-fix over a grant already live on prod).
• 07-08 FIRST LIGHT · THE WONDER-FIRST ARRIVAL · LIVE ON THE HUB · "a sky begins one light at a time" — a first-visit curtain, Spark → Fire → Bonfire, gift-once, adaptive skins. Phase-1 + the full-screen-curtain look/placement fix merged + deployed. A live browser test caught a return-visit flash (the welcome-back briefly bled the Hub through); the fix — no curtain on return, first-visit only + fully opaque — is in flight.
• 07-08 HUMAN TEST · n=3 · THE HONEST MIRROR · the new-user walkthrough reached three independent testers. The trust floor held all three times — but none felt the magic, impact never felt real, the cold-open oriented no one, and Earth 3.0 read as a dead end (the page loads, the story is unreachable). NoMoreBadNews led on trust a third time. The findings validate First Light + the felt-ripple direction; the flat result is the honest mirror doing its job.
• 07-08 THE HONEST-ZERO PILOT · DESIGNED · a cheap cold-stranger test of whether an honest true-zero return beats a vanity number — it gates the larger hope-vs-neutral partner-confirmed run (the locked v2 method). A statistician outreach letter is drafted and ready to send.
• 07-08 RAISE DECKS · ADVANCED · an investor cut reconciled to the on-brand structure (the moat framed as the verified-action dataset, never a causation claim) + a simple universal venture-studio deck built on-brand.
• 07-02 pm FABLE WEEKEND · 4 PRODUCT PRS MERGED · each built + cross-reviewed by a DIFFERENT agent + independently verified · Build Foreman Inc 2 (sail-ai-workforce#20 · programmatic Agent-SDK dispatch · ran its own 20-plus codex review rounds live, catching real evidence-integrity bugs in its own code) · IG felt-ripple Return (#50) + connected-spine Inc 1 (#51 · companion publish PERSISTS a real story · arrive loads real published stories not mocks · a cross-tenant action-strip IDOR closed, migration applied + verified on sail) · NMBN #7 + #8 (real actions end-to-end · 8 real sourced human-verified stories · honest dated daily return) · Compass #21 (the Playbook reads/writes the real per-owner table · the Mentor shows its rubric in chat · a default-playbook-shown-to-every-account privacy leak sealed). The loop caught a real bug at nearly every stage (the playbook leak · the IDOR · a dead CTA · a wrong-movement map · a stale-base rebase). Merged; deploys batched + Tim-gated.
• 07-02 pm OUTSIDE CONSULTANT READ + STRATEGY PART TWO (Fable) → BUILD DIRECTIVES · a fresh-eyes consultant read the outward corpus cold, then compared against our internal conclusions, then interviewed Tim · three independent critics now converge (the faults are real; the problem = propagation + sequencing, not diagnosis). Filed SAIL_Strategy_PartTwo_v0.1 + SAIL_Build_Directives_from_StrategyPartTwo: the build queue reorders to serve the proof loop + the phased launch (Foreman Inc 3 → IG spine Inc 2 → Growth S2), no new surface until the loop closes on a real user, deploy = previews-auto / prod-Tim-only, Codex-builds behind the Inc-3 seam. The honest finding · the build machine sits OFF the binding constraint (the run-deps · the raise · the partnership) · so the machine points at the loop while Tim + the Bridge take the outside game.
• 07-02 SAIL HUB · LAUNCH-READINESS · LIVE · every shell placeholder → REAL or HONEST (TitleWorkspace real Movement-Wall slug · Growth-Reach honest-empty · ReflectionCapture reveals live numbers only on persisted capture + live receipt, value/label-atomic · FeedbackDock names its write + surfaces a live-insert failure loudly · CompassPanel scoped session-local). sail-hub#18 → main e51535f → deployed 667008de. DUAL-verified: real-Codex runtime probe (forced stale 111 under live 999 → UI rendered 999) + Bridge source review. Caught + reconciled a prod-ahead-of-main (silent-merge-then-deploy) → the Foreman deploy gate must require a confirmed merge.
• 07-01 BUILD FOREMAN · INCREMENT 1 · MERGED · a build-state tracker + router in packages/foreman (sail-ai-workforce#19 → main d81302d) — enforces builder≠reviewer structurally, an evidence gate shut on stale-green / stale-PASS / self-PASS, integration rules in stop-the-line order, no autonomous-deploy action. 58 tests + a voice sweep over every emitted string. Real-Codex PASS at source. The orchestrator that takes Tim out of the relay.
• 07-01 JARVIS · COMPASS RUBRIC INTO THE GROWTH CMO · DEPLOYED · the first product face consuming Compass via the SDK (growth-muse#19 893d61c6) — every grounded suggestion + Action Plan carries a 5-dim rubric read (Uplift/Truth/Craft/Integrity/Action) scored in the same call, honest-labeled beside the evidence, vendored (consume-not-fork, vendor-equality test). Track 2 deepened the standalone Compass Daily Loop (sail-compass#20).
• 07-01 GROWTH GO-LIVE CORE · CONVERSATIONAL CMO · LIVE · S1 placeholder-purge → S3 Home briefing → S4 grounded live suggestions → S5 conversational CMO drafting + committing Action Plans (the shared Constellation ActionItem shape), anti-sycophancy, a live trust-pass gate (a metric leak reached prod → caught by Codex → fixed; no fabricated numbers survive). Sprint 16 Engine-Trust-Hardening published one canonical scrub as sail-core 0.15.0. Fable 5 restored + a Bridge-run model A/B → BUILD elevates to Fable for hard coding runs, one Model Registry.
• 06-30 SAIL GROWTH · OFF LOVABLE · REAL · the cockpit (Movement▸Brand▸Project switcher + Home/Plan/Act/Measure) live at sail-growth.pages.dev; repo solely ours (12 Lovable env-commits discarded). Real per-brand strategy brains + real projects on the live DB; the fabricated 50K target purged to real-or-honest-empty (S1). Owner-certified live via magic-link.
• 06-29 SPRINT 12 · DECISION ROOM + MEETING MODE · LIVE · local-only DecisionSession spine (option choice + EvidenceRefs + agenda + meeting questions + Decision Pack Markdown export), Meeting Mode note capture → Tier 1 expert notes / producer signals / TestPlans, Decision Timeline + Artifact Studio inheritance (no sail-core change). PR #43 → main `9fa25bc` → deployed `7a0909bf`. CC reproduced all 5 gates + EXECUTED the scrub render-verify; Bridge-verified live: "a human tester"→"a human tester", character "Marisol/Diego" survives, partner-safe pack redacts names/notes (the #32 lesson holds, no global sweep). Only the partner-safe pack is share-safe.
• 06-29 SPRINT 11 · SERVER TRUST BOUNDARY · LIVE · a shared `trust-boundary` scrubber (confidential terms + slug variants + person-name PII + overclaim) applied at the SOURCE across Constellation/workspace/Movement-Supercomputer/`/audience` payloads — defense-in-depth for ALL story-intel consumers. PR #16 → main `2f66463`; CC caught + fixed a real gap (`a seeded test brand` missing from patterns) before merge; legit content (character names) survives. Redeployed (worker `6856c343`) + Bridge-verified live (mode=live, clean). **Closes #92.**
• 06-29 SAIL GROWTH · OFF-LOVABLE MIGRATION BEGUN · Growth reframed as the real management tool — model locked Movement → Brand → Project, CMO-with-subagents, the Research → Plan → Act → Measure loop. Decisions: all four brands in parallel (BIRD · Phenomena · Earth 3.0 · Tim Kashani), foundation before fan-out; BIRD north-star = copies sold (B.I.R.D. live on Steam, $9.99). Lovable walked live (keep the CMO + strategy brain, cut the clutter); dev growth-muse already has connections + account-manager + real reach. BIRD vertical build brief (Strategy editor + onboard BIRD) handed to Claude Code, building. Gap-Map + Target Spec v1 filed.
• 06-29 SPRINT 10 · TEST STUDIO + FIELDWORK LOOP · MERGED · closes the loop (generate test packet → run fieldwork → import → recalibrate): TestPlan + baseline snapshot, deterministic table-read/survey/interview kits, Signal Import Mapper (contextual fieldwork PII scrub + dup detect), before/after Calibration Report (calibration-not-validation, aggregate export), Portfolio Test Board (saved-state-only). 6-PR stack #32→#37 → main `0c646f2`. #32 reconciliation: CC caught BY RENDERING that the sail-core seed name leaked in the calibration ladder/export, ported a targeted `scrubFieldworkNames` while keeping `safeDisplayText` name-blind (legit character names survive). Deploy pending.
• 06-29 SPRINT 9 · SIGNAL LAB + CALIBRATION LOOP · LIVE · local signal intake (each an EvidenceRef), Signal Lab UI (scrubbed paste/CSV import), prediction calibration (supported/challenged/inconclusive — never validation; confidence clamped, never promotes a claim to observed/validated), signal-backed synthesis/portfolio/timeline, deterministic Next Test Planner · Bridge-verified live (signal → inconclusive, claim held Synthetic, no false promotion).
• 06-29 SPRINT 8 · SYNTHESIS + PORTFOLIO + DECISION TIMELINE · LIVE · the decision layer — per-project synthesis with the single highest-value Next Move (planning-hypothesis-not-validation, source-labeled + EvidenceRefs), cross-project Portfolio Rollup (saved-state-only, "no hidden live runs are started from this view"), Decision Timeline · Bridge-verified live.
• 06-29 CLEANUP WAVE + AUDIT PUNCH-LIST · full-surface audit (Codex) → Bridge-triangulated; Hub mobile-overflow fixed (desktop byte-unchanged) + diagnoses (in-shell "404" = stale deploy → redeploy; earth3 alias = stale branch, human-test is `/reflect`) · NMBN hydration fixed (UTC masthead) · Growth honesty (live-rows gated + `TimKashani` scrubbed, ingest-medium no silent real-feed) · workforce triage (live workers GREEN; failures = env noise + deferred dashboard lint) · command-center Access downgraded to an honest-copy fix (internal+temp+scrubbed).
• 06-28 CONSTELLATION DECISION-OS DEEP WAVE · LIVE + VERIFIED · Calibration 1.1 (stable claim IDs + Expert Review Queue) + Calibration depth (note resolution workflow + multi-expert) + Model/Cost Observatory + Stage Development Ladder + Collaborator Fit (role-level-only export) — all merged, deployed, Bridge-verified live (stable-id rebind proven on reload; export-privacy proven with live sentinels) · Sprint 8 synthesis / portfolio / decision-timeline stack in CC review (derived/local, portfolio reads saved state only)
• 06-28 IG / GROWTH / NMBN DEEPENED + DEPLOYED · IG cold-start "contract of hope" front door + ContractThread through cobuild → publish (publish gate byte-unchanged) · Growth deepen (brand-portfolio + connector-readiness) + connectors-to-live (YouTube/Instagram token paths, deno-checked) + readiness UI, edge fns redeployed (growth-cmo v2 / growth-account-manager v3) · NMBN onto a live OpenNext worker (Next.js 16) + the RSS feed-sourcing layer behind the approval gate (fixture-default) · every stream CC-reviewed + Bridge-verified live
• 06-28 INTERNAL COMMAND CENTER LIVE · the launcher became the SAIL Command Center (Launch + Build Tracker + Deploy Status + Status Board digest + SAIL EDU tile) · passcode soft-gate + confidential-term scrubber (an adversarial audit caught the scrub was a no-op shipping codenames → strengthened, now clean, Bridge-verified live) · multi-stream parallel cadence all day (Codex + up to 4 CC streams)
• 06-27 CONSTELLATION SPRINT 1/2 SHIPPED LIVE + VERIFIED · first green §5.2: /constellation/workspace mode=live, 4 steps complete, ~11s, telemetry · merged workforce #14 (d35cf97, 12s timeout + bounded model allowlist + spend telemetry, spine byte-untouched), constellation #3 (02bad06, telemetry UI) + #4 (542a542, Run Observatory local-only per ADR-0001) · deployed story-intel b3481252 · Sprint 3 Artifact Studio kicked
• 06-26 OPERATING MODEL v2 RATIFIED + WRITERS ROOM / the writing companion · sprint-based model (lock foundation · long autonomous sprints · guardrails always-on · audit at boundaries), Codex-aligned · Writers Room foundation built, Voice & Craft Bible LOCKED, kill test PASSED (the writing companion indistinguishable from Tim)
• 06-25 Raise decks hardened (investor v2.7 venture-studio · donor v3.2) + red-teamed (ChatGPT / Gemini / a human tester) · Constellation kickoff + layered sail-core contracts (0.13.0 → 0.14.0) · Claude Corps host application drafted (due Jul 17)
• 06-23 THREE WAVES SHIPPED LIVE · Show-Off (Movement Constellation, Felt Ripple, Compass companion, Synthetic-Audience tool) · Real+Connected (impact-loop close, Greenlight, Growth demo-real) · Real-Slate (Earth 3.0 / Phenomena / Winter Lights real, the Phenomena "personal story, not a medical claim" guardrail on every surface)
• 06-20 GO-LIVE STAGE 2/3 + VISUAL UNIFICATION · governance / model-id sweep live (Opus pricing fixed), Mission Keeper governance verified · warm-canon design system shipped to Hub + IG · decks re-themed · Codex persona audit → 2 P0s caught + fixed
• 06-17 DEEP CREW DAY · WAVE 2 · 4 MORE PARALLEL STREAMS · no Gate 0 (no sail-core change) · all Bridge-verified, branch-only · Compass (`5ae8f5e`) the built-but-dark recommender wired to the live edge, zero spend, no-leak test · NMBN (`2439e68`) the FIRST real model-authored solutions-journalism (grounded single-source draft-article + zod validator + human-approval gate; the "6-agent pipeline" was paper-only), registration-moves-no-number + grounding verified, a 20-agent vet fixed 5 findings · Hub (`49897b0`) the in-shell Glass Box flipped fixtures→live spine, no silent fake, walk matched the public wall · Growth (`03bb4bb`) live CMO enrichment over the loop (model sharpens objective/platforms only, identity preserved, garbage→deterministic)
• 06-17 DEEP CREW DAY · 3 PARALLEL STREAMS + GATE 0 · sail-core 0.10.0 (MomentsDraft + BrandStrategy contracts, additive, 107 tests) · IG real model-gateway MOMENTS (publish gate byte-unchanged + still bites, provenance forge-proof, the gateway-bypass violation fixed, real $0.0086 call, 4-reviewer pass clean) · Growth CMO↔Growth loop (growth_brand_strategy migration applied to prod + Bridge-verified, the live round-trip closed on a seeded test brand) · SAIL Story /evaluate + /audience synthetic-audience (verdict-in-code, the daily-cap defect caught + fixed) · all Bridge-verified, branch-only · Tim ships
• 06-17 FOUNDER-BRAND GROWTH CONNECTORS COMMITTED · X / LinkedIn / Medium (fd23d2c, branch-only) + Bridge-verified via 2 independent adversarial agents (the 3 honesty invariants hold · fake never masquerades, Medium can't express a reach number, the LinkedIn scope-gate persists granted-not-requested) · sail-core 0.9.0 on main (PLATFORMS += linkedin, medium) · Tim's founder brand seeded on the Growth spine (5 channels, prod)
• 06-17 RETENTION MOTIVATION LAYER · 2 HUMAN TESTS + FIXES LIVE · two human testers (crossword expert) walked the flagged gamification preview · word-search drag-select + pattern-reveal + plain crossword clues + NYT-style input + copy/safety fixes built + Bridge-verified live · the act-vs-log resolution locked (split retention from impact · the act = an invitation never a gate · self-reported never counts)
• 06-17 Founder pilot · the Brand-OS · Architecture v0 + Tooling Landscape v0 filed · keystone "Tim moves story to impact" · Steward → CMO → Growth + a conscience seat · buy-the-hands-build-the-brain · positioning v0.3 locked
• 06-15 EARTH 3.0 HUMAN-TEST LIVE · the unifying-act experience deployed + Bridge-verified on a real URL (earth3-human-test.sail-hub.pages.dev) · facilitator arm-selector (hope/neutral/fear parallel, only the VO changes) · reflection → felt ripple → optional vouch → come-back, anti-extraction copy baked in, clean console · tester step-by-step + feedback form shipped
• 06-15 2-MODEL EXTERNAL AUDIT PASSED · Codex + Antigravity independently · 1 real P0 (publish-gate bypass via mutable automation_mode) caught + fixed at both DB and app layers + proven · 2 P1s fixed (honest-math seed inflation, experiment-assign audit atomicity via an atomic RPC) · the verified-vs-self-reported honesty line closed in confirmed_actions
• 06-15 Earth 3.0 measurement locked · the unifying act ("United we stand · divided we scroll") · hierarchy: primary = the reflection (action_confirmed), key secondary = return, integrity sub-rate = the optional vouch · OSF pre-reg draft + reflection-capture instrument + 3 variant scripts filed · sail-core 0.7.0 merged + published
• 06-14 LIVE LOOP DEPLOYED · spine-events (real event write path) ACTIVE on prod · all 3 agents (Compass / Growth account manager / Steward) deployed + flipped live · Hub dressed in v1 design language + deployed to Cloudflare Pages (sail-hub.pages.dev) · admin save-fix + capture layer merged · Compass conversation proven on prod (2 real Opus calls + 4 signals)
• 06-13 Anon write-policy review · NO holes (every write auth.uid/service-gated) · anon write grants tightened to least-privilege (3 public-submission tables) · platform grant-hygiene complete
• 06-12 COMPASS LIVE · first agent serving the real corpus — live_enabled on, real Opus turns grounded in Tim's data + Stage 3 write-back learning, $5/day cap, refuse-before-spend, audited
• 06-12 Platform grant-hygiene pass · dead anon/authenticated write grants revoked + the default-privilege bleed stopped (ALTER DEFAULT PRIVILEGES) · proven on prod (reads intact, writes die at the grant layer)
• 06-12 Design language v1 filed · sail-core tokens + the impact-ripple hero moment + per-surface character (Hub clean / Growth as a co-op game / Compass as mentor) · capture-UI build prompt rides the 0.6.0 events
• 06-12 sail-core 0.6.0 published · the retention/feedback foundation (behavioral events + feedback_submission + public-ripple opt-in) · all repo mains caught up
• 06-12 Wave 8 complete · Movement Wall live (real spine reads), Growth on 0.5.0 (casts retired), NMBN spine grant hygiene
• 06-12 Wave 7 complete · layered trend-reader, NMBN Action Strip live-reading the spine, sail-core 0.5.0, org-self-assign guard proven on prod
• 06-12 Tandem wave complete + merged to main · Compass Stage 3 write-back, IG Fully-AI publish gate (DB-enforced trigger), Growth Phase 2 connectors, sign-in reachability
• 06-11 GROWTH REAL · off Lovable + real Supabase persistence on prod (org-scoped CRUD, write policies + movement canon tracked) · the first surface on real data
• 06-11 Identity cutover shipped to prod + Bridge-verified · one consolidated identity
• 06-10 Movement/event spine shipped to prod · sail-core published (keel)