A project here is a major internal improvement at Datopian: a product, a process or a capability. Client delivery is tracked elsewhere, not here.

TL;DR: start a project in 3 steps

For major internal work only (product, process, capability). Not for client projects.

  1. Install the LTP skill in your AI agent (once): npx skills add life-itself/reasoncommons

  2. Ask your agent:

    Use the ltp-project skill to analyse a new project, [name]. Ask me questions to get started.

    Optionally tell it the situation (where we are) and the complication (what's wrong) up front.

  3. Create the page projects/slug/README.md with a ## Summary that answers:

    • Objective and key results: what we want, and 2–4 measurable results
    • Analysis: link to the LTP (or SCQ), or "not yet"
    • Driver: one named person
    • Timeline: start, end or review date, next milestone

A stub is fine: use status: idea and mark gaps. Switch to status: active once all four have real answers. Everything below is detail.

Is it a project?

Only if all three hold:

  • Internal improvement: builds a product, process or capability for Datopian, not a client deliverable.
  • Major: matters at company level, needs more than one person or more than a few weeks.
  • Bounded: has an objective and a date by which we'll know whether it worked.

Otherwise: client delivery goes in the client's tracker; ongoing work with no end is an initiative; campaigns, releases and routine tasks go in the execution tracker linked from the relevant initiative.

The page

A single file projects/YYYY-slug.md or, if it will have supporting files, a folder projects/slug/README.md. Copy the frontmatter from the project template and set a parent initiative. Frontmatter feeds the portfolio tree.

Example Summary (illustrative, not a real project):

## Summary

- **Objective:** Cut proposal turnaround so we can bid on more RFPs without adding staff.
- **Key results:**
  - Median time from RFP received to submitted: 10 days → 5 days by 2027-01-31
  - Bids submitted per quarter: 4 → 8
- **Analysis:** [[ltp/rfp-speed-2026-10-01.reportLTP report]]
- **Driver:** Jane Doe
- **Timeline:** 2026-10-01 → 2027-01-31. Next milestone: reusable answer bank, 2026-11-15.

Put actions in a beads epic (bd create --type=epic …) and set active_issue to its ID. The page states the plan; beads tracks progress.

More on the LTP skill

LTP (Logical Thinking Process) turns what you know into a goal, the key results needed to reach it, the obstacles and the actions. The skills come from Reason Commons:

  • ltp-project interviews you, or reads existing notes, docs or an SCQ, and writes a *.ltp.yaml plus a report of the judgment calls it made.
  • ltp-visualize turns the *.ltp.yaml into an interactive HTML view.

Tips:

  • You don't need anything written first. If you do have notes or an SCQ, point the skill at them.
  • Review the report's judgment calls; the first output is a draft.
  • Keep files in the project's ltp/ folder. Once an LTP exists, update it rather than older notes.
  • The LTP goal becomes the objective, its key results become the key results, and its actions become beads issues.

Worked examples: E2E operators, PortalJS.

Keeping it current

Update the Summary when anything in it changes. Active projects are reviewed in the accountability rhythm. When a project ends, record what we tried, what happened, what we learned and what's next, then set status: retired.

Built with LogoFlowershow