Status 2026-W41 (6 to 11 October)

Progress. Engineering for milestones M0 to M3 is finished: simulator and survey, GNSS fault injection, trust engine, three-vehicle allocator, backend with audit log, recommender, policy gate, operator console, scenarios, ground-truth evaluation (CI regenerates the table) and a held-out validation (7 of 8 expectations met). This week also added: - A calibrated position bound: 3 m, held in 100% of 4,040 development and 99.72% of 5,308 held-out GREEN reports. The first figure shown held in only 30.8%. - Fleet-level mission completion: with one of three vehicles lost, 100% of the area was covered when the re-task was approved and 87% when it was denied (3 runs each, scripted operator). - The product documents and the how-this-was-built page, edited by me; a first risk log. - Usability round 1 (2 participants, 11 October): neither participant acted on the decision card unaided; one approved after a hint. The findings note is in docs/usability/findings-2026-10.md. In response: protocol v2 (a role and a goal in the briefing, screen recording required) and a console redesign (a red "Decision required" band, flat operations layout). Not yet re-tested. - Publication: git history cleaned (54 MB to 3.4 MB), security and contribution files, Dependabot, CodeQL, the repository made public, a documentation site on GitHub Pages and a README summary that replaced a stale "M0" status line.

Next. - Usability round 2: 2 to 3 new participants on the redesign (issue 16 stays open until then). - Demo video recorded (https://youtu.be/L7DrhOYMQ5A), then tag v1.0. - Rewrite this status, close issue 1.

Risks and decisions. See docs/programme/risks.md (R4 and R14 mitigated). Decisions recorded this week: ADR 0001 (separate detectors), ADR 0002 (scenarios follow PX4's failsafe), ADR 0003 (LAND for a frozen GNSS position). New risk to add: the redesign may not fix the decision problem (see round 2).

Re-plan. The calendar plan (M0 due 18 October to M4b due 6 December) is about eight weeks; the engineering was done in the first three days of the repository (created on 8 October). The remaining schedule risk is people and writing time: recruiting round 2 and recording the video. Corrections this week: a frozen GPS position makes the vehicle fly away, not stall (the recommendation changed to LAND), the displayed position bound was invalid (replaced by a calibrated one), and the first usability briefing did not tell people a decision was theirs to make (replaced).