Start here

1 · Study

• The board: 37 ECO tasks by evidence tier. Start at the top.
• The five tasks Low on both attempts
  • II.7 Draft AI solution (Business Needs)
  • V.3 Oversee model governance (Operationalize)
  • II.6 Manage adoption/integration risks (Business Needs)
  • V.7 Oversee AI solution contingency plan (Operationalize)
  • IV.1 Oversee AI/ML model technique(s) (e.g., algorithm, selection) (Model Dev & Eval)
• The roadmap: Phase I and Phase VI in PMI's order, reading order, the 60-minute loop, the three-question read. Printable.
• The two grids: phases across, domains down, where each task is graded.

2 · Read (PMI sources)

• Workbook pages: open any task page and use "Show pages" under What to read. Not rendered on this machine yet: run python scripts/render_reading_pages.py
• Study guides: Trustworthy AI (I), Data Needs (III), Model Dev & Eval (IV), Operationalize (V). No Business Needs guide exists; use its task pages.
• Glossary: 311 PMI terms, reconciled against Study Hall's 300 flashcards.
• The vault (Obsidian): output/vault, not built on this machine yet: run python scripts/build_vault.py. Open the folder as a vault; start at Home.md or Workbook/_Index.md.
• Course modules on disk: Modules 2 to 7 as PDFs in source/. Modules 0 and 1 are not on this PC.

3 · Your misses

• Study Hall, 26 reviewed misses: on each task page, paraphrased only on this machine (run python scripts/build_private_misses.py for the full items). 28 task links across the board.
• Miss log: record every miss from Udemy mocks, the hardening loop, or re-reads, by ECO task. Export before switching devices.
• Score reports: May 9 and Sept 6 bands are on every task page and drive the board's order.

4 · The daily loop (60 minutes)

1. Write Phase I's six task groups from memory. Check against the roadmap.
2. One task page from the top of the board. Enablers aloud. Open its pages. Read its misses.
3. Ten hardening-loop items on that task, every stem stating what is already done.
4. Log every miss. Mark the task worked even with none.

Weekly: one full 120-item timed mock from an unseen source, misses logged by task.