[danielgorgonha]

What this changes and why:

• docs/method.md, Battery: the proposed battery method for phones and for the verdict laptop. It is a proposal until it is posted on the forum, and no battery measurement is made before then (post 15).
• Laptop: unplugged for the whole run; RemainingCapacity (mWh, Windows root\wmi BatteryStatus) logged every 10 s; run start and end marked by timestamp; an idle baseline of the same length; the whole unplugged session's capacity drop cross-checked against its powercfg /batteryreport row. DischargeRate is not used as the measure.
• Phones: adb over Wi-Fi so the phone is not charging, starting between 80% and 60%; the charge counter and current now summed every 10 s; the share used instead of mAh on a phone whose counter scale does not match its battery.
• Every battery run lasts at least 20 minutes (a shorter workload is repeated, and the repetitions are reported). Each result keeps its uncertainty, about 1% on the laptop and about 7% from the phone counter, and the net result carries the baseline's uncertainty too.
• Published battery results hold the model and the measured values only, nothing that identifies a device.

Tests run (see CONTRIBUTING.md):

node --test tests/node.test.mjs tests/charts.test.mjs: all pass (docs only).

For a result under community-results/: commit built, device, browser:

Not applicable.

AI disclosure: no AI assistance / AI-assisted (tool and what it did):

AI-assisted: Claude Code wrote the text; a separate AI review pass checked it against the measurements and the forum posts.

🤖 Generated with Claude Code