Unified the headers across all four cockpit tabs. [research-index.html](command-center/research-index.html) and [ai-engagement-dashboard.html](command-center/activity-log/ai-engagement-dashboard.html) now match [index.html](command-center/index.html) and [factory.html](command-center/factory.html): CE logo + "Chris Explains AI" brand + dash-name on the left, live clock + date + Dallas weather on the right. Research dash-name corrected from "Research Reports" to "Research". Changes made in the generators ([build-research-index.py](command-center/build-research-index.py), [build-ai-engagement-dashboard.py](command-center/activity-log/scripts/build-ai-engagement-dashboard.py)) so future republishes preserve the look.

Evolved [.claude/skills/sync/SKILL.md](.claude/skills/sync/SKILL.md) from a dashboard-reconciliation pass into an end-of-day cleanup command. Added bedtime trigger phrases ("I'm going to bed", "sync everything up", "wrap it up for tomorrow"), HANDOFF.md reconciliation (remove items the logs prove are done; does NOT regenerate the file — that's /handoff's job), ask-before-changing mid-flow when uncertain, plus a new scan step that surfaces Idea Factory items parked at a human gate and learning/build-candidates.md entries whose trigger is now met. /sync and /handoff remain distinct commands.

Updated [HANDOFF.md](HANDOFF.md): Active focus rewritten so the Track B opener now points at the master slide template as the first production step. Track B planning entry added under "Done this session" with the full decision summary (pricing, delivery shape, tooling, asset inventory). Spec is at [course/format-spec.md](course/format-spec.md).

Track B planning session ended with [course/format-spec.md](course/format-spec.md) locked end to end. Pricing decided: **$229 waitlist / $279 post-launch** (was undecided). Delivery shape: ~110 per-section videos with a next-button UX (~80 section videos + 22 face-only module intro/outro + 8-12 demo videos), ~5.7-6.0 hr total runtime at Chris's measured 150 wpm pace. Format: Loom-style webcam-in-corner over slides, draft IS the script with a light pre-record edit pass for slide/demo cues only. Tooling locked: slides via python-pptx → upload .pptx to Google Slides, Speakflow floating teleprompter pinned under the webcam, two-stage edit pipeline (Claude runs ffmpeg batch preprocessing for trim/normalize/encode/Whisper captions, Chris does creative cuts in CapCut). Mic still open (Chris will pick; Shure MV7 default recommendation). Asset inventory rewritten to match what the course actually contains (zero fully-written worked prompts in the draft, only structural skeletons and labeled examples): 17 per-section reference-card downloads, Master Reference Pack PDF, Module Summaries PDF, sales trailer video, welcome email video, sample-lesson preview. No case studies, no practice tasks, no community at launch. Section 6.9 (13.2 min) stays as one video. Slide density loosened from "headline + 3 lines max" to content-adaptive (bullets, tables, prompt snippets, diagrams allowed); slide-and-script pairing built into the deck-plan format. Next concrete step (separate chat): design the master slide template, 8-12 worked sample slides covering all 5 patterns + brand colors + webcam safe-zone.

First real Factory run completed end-to-end on `video-reverse-engineer`. Gate 1 closed with locked answers in [clarify.md](ideas/video-reverse-engineer/clarify.md); 4 parallel Sonnet researchers wrote dimension files synthesized into [research.md](ideas/video-reverse-engineer/research.md); 5-persona stochastic-consensus eval + Opus synth produced [evaluation.md](ideas/video-reverse-engineer/evaluation.md) (4-1 build-later, 68/100); Chris redirected on new "about-to-start-posting" context; second round wrote [evaluation-v2.md](ideas/video-reverse-engineer/evaluation-v2.md) (4-1 build-now, 77/100, full inversion with 3 personas flipping on merits and devil's-advocate holding); detailed Opus plan at [plan.md](ideas/video-reverse-engineer/plan.md) (~6,200 words, implementation-spec depth, claude -p locked as the synthesis mechanism). Parked at Gate 3 awaiting Chris's call to build. Also shipped: dedicated Idea Factory tab [factory.html](command-center/factory.html) on the cockpit with unified pages-nav across all 4 dashboards (Command Center / Idea Factory / AI engagement / Research); active tab highlighted indigo. Factory SOP updates: Gate 2 auto-pass rule (score ≥ 70 -> straight to plan), detailed-plan spec for all future plans, Expand-skip allowance when Gate 1 locks scope, Gate-redirect protocol (context-update + v2 personas + v2 synthesizer). Bug fix: [factory.py](scripts/factory/factory.py) `cmd_gate_open` now idempotent on both state and queue halves; resolves the `set stage awaiting-*` -> `gate-open` collision hit during this run. Publish hardening: [publish-cockpit.py](command-center/publish-cockpit.py) now sets `PYTHONIOENCODING=utf-8` on subprocess env + `errors='replace'` on decode + UTF-8 reconfigure on its own stdout, eliminating the Unicode error path on Windows. Smoke-tested kill/park path with a dummy idea, verified cockpit rendering, cleaned up. Lab notes added to [lab-notes.md](.claude/rules/lab-notes.md): 2 anti-patterns (the gate-open collision + sub-agent prompts must include prior-file anchor when running redirects), 2 validated patterns (stochastic-consensus + redirect-rescore worked first try, Expand-skip when scope is locked). Cockpit pushed live multiple times (final commit c355a4e). Factory declared ready by Chris.