Skip to content

2026-08-13-max-tracker-physical-static.md

Latest commit

 

History

History
65 lines (52 loc) · 3.28 KB

2026-08-13-max-tracker-physical-static.md

File metadata and controls

65 lines (52 loc) · 3.28 KB
Raw

Max Tracker — Physical Static Review (2026-08-13)

Outcome

The Torget repository's FIRST physical flash was performed 2026-08-13 with explicit user authorization, from the merged main state (feature branch worktree-max-tracker at 207e4ed, merged as e366bd4; binary built at that tree). The static physical gate for the Max Tracker pages and the shared platform is PASSED. The Solceller firmware copy is retired as the screen's driver; this repository now owns the glass.

Flash evidence

  • Flash performed with esptool directly (bypassing idf.py overhead) after repeated USB enumeration bouncing; write completed, hash verified, hard reset via RTS. Log: scratchpad flash-race output, "Hash of data verified".
  • Boot verified over the LAN: the device (192.168.1.135) established connections to the tokenserver and fetched /api/tokens, /api/agent-status and /api/max-tracker (the latter impossible for the previous firmware).

Physical inspection evidence

User-captured photographs of the physical AMOLED (IMG_0191/0192, reviewed at full resolution by the controller session) show:

  • CODEX · MAX TRACKER page: indigo heat grid populated with real backfilled history, exact-red max-day cells, legend with MAX label, stat row STREAK 4 DAYS · MAX WEEKS 0 · AVG PEAK 53 % · MAX DAYS 11, six pager dots, real provider icon in the shared live header, NO ACTIVE CHAT state.
  • CLAUDE · MAX TRACKER page: MAX 20X plan badge right-aligned in the eyebrow row, gray activity backfill plus colored recent quota days, 1 CHAT ACTIVE live-header state (the building session itself), stat row STREAK 4 DAYS · MAX WEEKS 0 · AVG PEAK 37 % · MAX DAYS 0.
  • Colors, geometry and typography match the simulator captures the landmark suite pins; no tofu, no truncation, no layout breakage observed.

Touch/swipe interaction and the launcher long-press were reported working by the user during inspection. Brightness and viewing-distance judgments are from photographs and user report, not instrumented measurement.

Hardware truth discovered (recorded in README + this review)

A computer USB port could not power the RUNNING firmware: the panel's current draw made the board bounce off the USB bus repeatedly (both with the previous firmware and the new one) and eventually hang, with no port enumeration and no network traffic. On its own USB power supply the same board and firmware boot and run stably. Consequences:

  • Flashing should happen in download mode (panel dark, ROM USB stable); the direct-esptool race script in this session's scratchpad worked around enumeration bouncing.
  • Running the firmware from a computer port is not supported; the README's manual-setup section now says so.
  • Serial-console debugging of the running firmware requires a powered hub or a PSU+data split; not verified in this session.

Gates

  • Static physical gate: PASSED for the current static build (quota pages, agent monitor states and Max Tracker pages seen live on glass; NEEDS YOU/error overlays verified earlier in the simulator matrix only).
  • Motion remains gated: no lv_anim exists in these renderers, and motion work (including the swipe-navigation effort in progress on main) must follow the interaction-performance protocol in the AMOLED skill, measuring the shared display pipeline on the physical panel first.