Shipped & live12
Capabilities built and running in the last 30 days. We recommend logging these in Jira as completed, so the board reflects what actually exists. Several map onto existing epics that can simply be updated.
1
Company Knowledge Base — built and live
Live
Outcome The company now has one searchable, trustworthy knowledge base that both staff and our AI tools read from — replacing scattered personal notes.
Definition of Done- Live at kb.life-labs.dev with a friendly reader/editor and working full-text search.
- Every page is structured as sources → topics → traceable claims, so facts can be checked back to where they came from.
- Seeded with the first ~75 company and contemplative documents; search restored end-to-end.
2
Automatic file & document capture into the KB
Live (manual); auto-sources pending setup
Outcome Staff can drop in a file and it is safely filed into the knowledge base — and the plumbing to pull in Google Drive and Slack automatically is built.
Definition of Done- Drag-and-drop upload → automatic clean-up → a quality & secrets check → published and searchable.
- Automatic Google Drive and Slack capture is built but not yet switched on — see 'Turn on automatic Drive & Slack capture'.
3
Weekly Research Radar — live
Live · first digest 22 Jun
Outcome The team gets a ranked weekly digest of the most relevant new scientific papers, plus a reading inbox — so we stay current without manual searching.
Definition of Done- Automatically gathers new papers each week, ranks them for relevance, and posts a Slack digest.
- An in-hub reading list lets people mark papers important/read and send them to the KB.
- First scheduled digest goes out Monday 22 June 2026.
4
Meeting transcription tool — live
Live
Outcome Upload a meeting recording and get a clean, speaker-labelled transcript filed into the KB — privacy-safe (processed in the EU, no AI training on our audio).
Definition of Done- Audio upload → transcript with named speakers → published as a KB meeting document.
- EU data processing; audio deleted after use.
- Automatic Zoom capture is a separate next step — see 'Automatic Zoom meeting capture'.
5
Internal Slack assistant (LIFE-bot) — live
Live
Outcome Staff can ask one Slack bot a question and get answers from the knowledge base, research radar, benchmark results and golden questions — while only ever seeing what they're allowed to.
Definition of Done- Answers from all four data sources directly in Slack.
- Respects each person's individual access rights; verified working end-to-end.
- Two finishing steps remain — see 'Finish the Slack assistant rollout'.
6
Company activity dashboard (who did what)
Built & live; staff-use pending sign-off
Outcome Leadership can see activity across all internal tools — including uploads — in one place, useful for adoption and operations. End-user meditation data is deliberately excluded for privacy.
Definition of Done- All internal apps report activity to a single dashboard with an uploads view.
- Automatic data-retention limits applied.
- Switching it on for staff visibility needs a privacy sign-off — see 'Privacy sign-off'.
7
Per-person access control across all tools
Live
Outcome Each person sees exactly the tools and data appropriate to their role, and senior reviewers get curator access — closing around 30 security gaps in the process.
Definition of Done- Role-based plus per-person permissions enforced across every connected app, with tamper-proof identity.
- Verified live (e.g. our senior reviewer now has Data Lab curator access).
8
Safer, more reliable operations (AI Ops)
Live
Outcome Production can no longer be changed by accident, and an automated monitor fixes common infrastructure failures on its own — less downtime, lower risk, less manual work.
Definition of Done- All accidental auto-deploy paths removed — production is only ever updated on purpose.
- A 24/7 monitor detects and auto-resolves build/hosting/database issues and security-reviews every change before it ships.
9
Benchmarking & AI-judge system (Data Lab) — v0.3
Live · v0.3 in progress
Outcome We can systematically measure and improve the quality of the meditation AI (“Wisdom”) using a panel of expert AI judges — the engine behind the v0.3 benchmarking work.
Definition of Done- A secure Data Lab where reviewers run candidate answers through expert AI judges with live scoring.
- The judge engine runs as its own reliable service, separate from the live app.
- Golden-prompt test reports are reproducible and access-controlled.
10
Central control & history for all AI prompts
Live
Outcome Every AI prompt across LIFE is centrally tracked and version-controlled — no more silent changes — and it's the foundation for letting Alex safely edit Wisdom's behaviour.
Definition of Done- All prompts resolve through one tracked registry, visible at /admin/prompts, with full version history and verified backups.
- Current Wisdom prompt (v0.8.3) is live through the registry.
11
Predictive profiling — testbed & validation tooling built
Tooling built; real-user verdict pending
Outcome The tooling to test whether a personal profile makes Wisdom's responses better is built — together with an honest audit that caught and corrected an earlier over-claimed result.
Definition of Done- A Data Lab testbed that predicts user reactions (with vs without a profile), an offline validation harness, and a training-data export pipeline — all built and verified.
- A thorough audit confirmed and fixed 20 serious issues, including a misleading “confirmed” result that turned out to be synthetic.
- The real-user verdict is still pending — see the profiling items under 'What's next'.
12
Stages-of-Meditation evaluation tool (SOME) — live
Live
Outcome Our meditation-stages evaluation tool is live, rebranded, on company single sign-on, and security-hardened.
Definition of Done- SOME live on company login with security issues fixed.
- The larger phenomenological-assessment dashboard is a separate, unbuilt idea — see 'What's next'.
What's next7
Work that is gated, half-shipped, or awaiting a call. The three flagged above need a decision from you; the rest are ready or near-ready to finish.
1
Profiling: make it provable on real users
Not started (top priority)
Outcome Before we can claim a user profile improves Wisdom, the app has to measure whether users actually feel better and keep using it — then we run an honest study on real, consenting users.
Definition of Done- Step 1 — Build a lightweight in-app mood/check-in plus simple 7-day return measure, so there's a real outcome to test against (this does not exist today and is the blocker).
- Step 2 — With a consented group of users and a success bar agreed in advance, run the validation and report a clear go / no-go.
2
Finish shipping two completed quality upgrades
Built & tested; not switched on
Outcome Two finished pieces of the quality program are built and tested but not yet turned on in production.
Definition of Done- Let reviewers safely author and AI-steer Wisdom's prompt inside the Data Lab (built, tested — needs deploy).
- Upgraded v0.3 AI-judge rubrics (built, verified — needs deploy + a re-run of the benchmark baseline).
3
Turn on automatic Drive & Slack capture into the KB
Built; needs admin setup
Outcome Automatically pull Google Drive documents and a Slack channel into the knowledge base — the engine is already built.
Definition of Done- Grant the Google Workspace permission, register the Drive watch channel, and set the Slack token — then content flows in automatically.
4
Automatic Zoom meeting capture
Blocked on a Zoom workspace
⚑ Decision: purchase a company Zoom workspace
Outcome Meetings recorded on Zoom get automatically transcribed (with real names) and filed into the KB — no manual upload.
Definition of Done- A company Zoom workspace and integration so completed recordings ingest automatically with named speakers.
5
Privacy sign-off to switch on staff analytics
Drafts ready; awaiting CFO
⚑ Decision: David (CFO) to finalize the privacy assessment & staff notice
Outcome Activate the activity dashboard for staff visibility — legally and transparently.
Definition of Done- David finalizes the drafted privacy assessment (DPIA) and staff notice; then staff visibility is switched on.
6
Finish the Slack assistant rollout
Code ready; finishing steps gated
Outcome Complete the switch to the new Slack bot and enable its richer assistant mode.
Definition of Done- Retire the old bot (final, irreversible cleanup) and flip the Slack 'AI assistant' setting on.
7
(Optional) Larger meditation-stages assessment dashboard
Idea / spec only
⚑ Decision: build it or shelve it
Outcome A research tool for the team to build a cross-tradition map of meditation stages and their markers — a substantial future build that is only specified today.
Definition of Done- Decide whether to build; if yes, scope it as its own epic.