AlDente Pro weekly calibration - live observations
Date: 2026-08-23 (Europe/Bucharest, EEST / UTC+3)

System
- MacBook Pro 13-inch, 2019 (MacBookPro15,4), Intel Core i5-8257U, 8 GB RAM
- macOS 15.7.9 (24G830)
- AlDente Pro 1.38.1 (98)
- Battery condition reported by macOS: Normal; cycle count: 198

Test context
- AlDente's weekly calibration was in progress and was not interrupted.
- The AlDente dashboard showed Calibration Mode, step 5/5, discharging to 70%.
- The Mac was on battery. macOS Low Power Mode was disabled.
- AlDente preference onBatteryEnergyMode was 0 (disabled).
- The privileged helper held DisableInflow, PreventUserIdleDisplaySleep, and
  PreventSystemSleep assertions during the calibration.
- AlDente and its privileged helper remained running throughout the observations.

Observed controller and benchmark states
- 10:43-10:48: CPU_Speed_Limit=25%, CPU_Scheduler_Limit=40%, with no
  thermal or performance warning reported by macOS.
- 10:47:57: /usr/bin/time -lp openssl speed sha256 produced 24,137.64 kB/s
  for 8192-byte blocks. The controller was 25%/40% before and after the test.
- 10:50:21: without quitting AlDente, disabling a setting, or performing a
  sleep/wake cycle, the controller temporarily returned to
  CPU_Speed_Limit=100%, CPU_Scheduler_Limit=100%.
- 10:50:34: the same SHA-256 test produced 227,300.10 kB/s for 8192-byte
  blocks. Immediately after the test, CPU_Speed_Limit was again 25%, while
  CPU_Scheduler_Limit was 74%.
- 10:53-10:56: the controller repeatedly reported 25%/40%.
- 10:57:19: it reported 28%/100%.
- 10:57:41: it reported 69%/100%.

Important caveats
- The low controller state is intermittent/dynamic even while the same
  calibration phase remains active.
- A Firefox RDD process briefly consumed approximately 113% CPU during one
  early low-performance sample. Later samples showed the controller limits
  varying without that process load.
- Screen Sharing was active for part of the observation and increased
  WindowServer/ScreenSharingAgent load. The two benchmark figures above are
  therefore useful as same-machine state comparisons, not as clean hardware
  baselines.
- System-wide memory free was 49% at 10:50, with 228.75 MB swap in use;
  macOS memory pressure level was normal. No current memory-pressure event was
  observed during the captured reproduction.

Recovery/transition observation
- In contrast to the earlier reproduction, the CPU controller temporarily
  recovered during this calibration without quitting AlDente and without a
  sleep/wake transition, then dropped again after load. This does not identify
  the root cause, but it shows that the controller state is not permanently
  latched in every occurrence.

Calibration completion
- The final discharge phase completed without intervention at approximately
  11:12:25 EEST. AlDente's lastCalibrationState changed from 4 to -1 and
  lastCalibrationTimestamp changed to 1787472745.880276.
- At 11:13, macOS reported 73% UI charge, while raw capacity was
  3534/5040 mAh (70.12%). AC power resumed and the battery was not charging.
- AlDente released its DisableInflow and sleep-prevention assertions.
- CPU_Speed_Limit and CPU_Scheduler_Limit both returned to 100%, with no
  thermal or performance warning.
- AlDente remained running. No sleep/wake cycle and no application quit were
  used for this transition.

App-generated diagnostic
- Generated after calibration completion at 11:15 EEST using AlDente's
  "Generate and attach debug file (recommended)" option.
- Filename: aldentepro-debug-2026-08-23-11-15-15.pasta
- Size: approximately 262 KiB
- SHA-256: c39c2a15c120af077f2171a076231af8d46fa36cbaf317535ba43ec309ba849a
