SELF onboarding · verification and constraints

End goal

Matt can return from the cinema, listen to a grounded onboarding guide, open one linear Notion page, launch one prepared Codex conversation, and complete a first evening planning session without guessing where information lives or which action comes next.

End state

Onboarding page, first-evening cockpit, critique, audio guide, Telegram delivery and a verified path through the current dashboard.

Proof

Live review URL, Notion page readback, Codex task created, audio decode/Telegram receipt, provider and browser checks.

Stage

Designed system is being made operational for a first calibration. Existing runtime automation remains explicitly separate unless verified.

End goal system

17:30 scout
digest and questions

→

18:00 cockpit
reflection, manual plan, tracker

→

closeout
transcript, snapshot, briefs

→

morning relay
two audios and triggers

Route to the end state

1 · orient
listen and open the linear Notion page

→

2 · calibrate
run the first special evening session

→

3 · verify
confirm records, audio and receipts

→

4 · simplify
remove friction before automation

Deliverables and observable criteria

Linear Notion onboarding page.
Must lead from mission → touch points → evening flow → storage → audios → trigger rehearsal → first-session instructions.
Verify: fetch page, confirm the ordered headings and links, open it in Notion.First-evening Codex cockpit.
Must state calibration purpose, manual-writing transition, checklist states and closeout phrase.
Verify: created task has the prompt and can be opened in the Codex sidebar.Onboarding review and critique.
Must explain every current process, its friction, and the simplification chosen.
Verify: local Markdown exists and is linked from the review index.Charon onboarding audio.
Must be grounded, cheeky, supportive, practical and understandable from scratch.
Verify: final file decodes, duration is recorded, Telegram API returns the private message.System check.
Must inspect current dashboard, Computer History evidence and touch points, and record unknowns.
Verify: report names what was observed and what remains unverified.

Constraints

Do not turn the dashboard into a second task manager.
It remains a map and linked-view cockpit. Human plan edits belong in the Daily Plan, Calendar and Sheet.

Do not treat computer activity as intent.
Evidence can prompt a question. It cannot prove distraction, billable work or completion.

Do not silently write after the reflection phase.
The agent announces the manual-writing boundary and only the user edits plan, calendar and tracker.

Do not claim runtime activation from documentation.
Every step must say designed, built, tested, sent or activated precisely.

Do not make the first evening a perfect-routine exam.
Calibration and truthful receipts matter more than thirteen green boxes.

Handoff flow

Onboarding audio
why and how

→

Notion page
linear route

→

Codex cockpit
first calibration

→

Receipts
what actually saved

Why each exists: the audio removes orientation friction; the page keeps the route visible; the cockpit holds the live conversation; receipts make the system trustworthy.

Evidence slots

Local
Onboarding, first-session script, critique, source brief and receipts live in the task output directory.

Notion
Dashboard and onboarding child page are verified by provider fetch and visible browser inspection.

Telegram
Private onboarding audio message ID and returned file metadata recorded after send.

Browser
Review page, Notion page and Codex task opened; no claim made for untested automation.

Live executor notes

Runtime executor result: four discoverable local entry points now exist — daily-digest, plan-tomorrow-and-reflect, planning-session-done and daily-audio-guides. The local rehearsal verified the reflection boundary, truthful checklist states, plan snapshot, fixed-track fixture assembly and idempotent closeout. Recurring scheduling, provider writes and real next-day audio delivery remain pending until their own receipts exist.

The fresh pinned first-evening task is 🤖 First-evening calibration coach, task ID 01a0c9b4-00db-7be3-b007-f6552c0a785b. The executor updates this board in blue as artifacts and checks complete. Green means proof exists. Red means a real failure or blocker. Base contract remains black.

What the computer-use check is allowed to prove

It can show which apps, pages and review surfaces were recently active and whether the dashboard is reachable. It cannot prove the meaning of activity, the quality of work or whether an item was completed.