Reproduce a battery depletion calculation: independent review welcome

[mohammadrezwankhan]

I maintain Datacenter Twin Lab, with Codex assistance. The new alpha is a synthetic, local-first power-continuity simulator. I am inviting a bounded technical reproduction; independent external review has not yet been obtained.

For a repeatable command-line review, use the tagged v0.3.0a0 source and the release-pinned capsule below. The current browser lesson 3 runs JavaScript first; Verify against Python loads the reference Python engine on demand. The live course can change after a release, so record the source revision or input hash with any reported result.

Lesson 3 disables charging for both the 100 and 50 kWh cases. A 100 kWh battery at 0.90 discharge and 0.95 distribution efficiency delivers 85.5 kWh to a 1,000 kW IT load. This gives 307.8 seconds of ride-through. The outage begins at 300 seconds, so depletion occurs at 607.8 seconds; utility recovery is at 900 seconds. The charging-disabled 50 kWh challenge depletes at 453.9 seconds. The CLI generator_failure preset allows pre-outage charging: its 50 kWh variant is a different input and must not be compared as if charging were disabled.

A useful review would:

1. Reconstruct that result without using the engine's reported residual as the proof.
2. Run python -m datacenter_twin simulate --preset generator_failure and record the exact revision, input hash, environment, and any mismatch.
3. Check the separate surviving-path example: 700 kW gross × 0.95 = 665 kW delivered, leaving 335 kW unserved.

The reproducibility capsule includes report artifacts and hashes. The review worksheet separates reproduced behavior, mismatches, ambiguities, and out-of-scope claims. Please share only synthetic or sanitized inputs.

Agreement would support software behavior for these fixtures. It would not establish facility calibration, AC transients, protection coordination, cooling/GPU performance, certification, or secure shared deployment. First-run failures and documented uncertainty are useful feedback too.