Current checks
On October 21, 2026, 451 existing automated cases passed across private source copies: Anime 337, Upscaler 70, Compressor 38, Conflict Monitor 3, and Health Monitor 3. Tests used disposable state; the original applications and personal media were not test destinations.
All 704 files in the bounded source/document inventory matched their starting hashes after the audit. This scope excludes live data, media, dependencies, credentials, and generated trees. It is not a full-disk integrity claim.
Different kinds of evidence
- Source inspection establishes what the implementation contains, not that every path works.
- Retained measurements describe a past input, runtime, and method. They are not automatically current benchmarks.
- Fresh tests establish the tested cases under stated conditions, not complete coverage.
- Demonstrations show actual UI behavior with safe fixtures. They are not performance comparisons.
Measurements keep their context
Compression results use final-file bytes and decimal GB. Encoding throughput is separate from the recorded successful-stage throughput. Lossy size reduction is separate from visual quality. Dashboard report counts are separate from distinct-source counts and factual accuracy.
The historical Anime artwork report has detailed methods, but its linked raw temporary files are missing. That result is explicitly labelled historical on the case study.
Privacy and attribution
Public assets use generated test patterns, disposable profiles, or the applications’ built-in fictional records. Private source, personal libraries, credentials, raw logs, patient data, and copyrighted episode clips are excluded.
Portfolio limitations
The case studies are curated explanations. Private source and original logs are not public, so these explanations alone are not independent reproduction of every result. The case studies identify the gaps rather than implying otherwise.