RPL 3304/2025 · 22 pages · C36-SPECIALIST-R33
Every preserved statement proposition was passed through the current context/inference triage. The result is a reproducible review map—not a machine verdict on honesty, guilt or credibility.
These are inherited provisional working classifications; they are not 276 adjudicated findings and are not converted into a score.
Each tile is candidate statements / statement propositions on that page.
R33 expressly says the 2023 classification judgment is under appeal. The first-instance judgment exists, but the filing does not present it as final. This is why the machine must be able to clear a candidate, not merely accuse.
The filing uses hostile characterisations of Gil Marer and reproduces project material. In the reproduced image, the project itself visibly marks several matters as “No probado todavía”. A later summary should preserve that qualification when it is material.
The 21 September 2020 withdrawal and apology are genuine adverse evidence and remain visible. The larger inference—that repeated lawyer changes share the same reason—requires mandate-by-mandate proof. Existing v3 review records the broader causal step as incomplete/unsupported, not the apology as nonexistent.
Exact words → use event → restored material context → contrary/lawful alternative → scope/causation test → dependency review → changed or unchanged decision → rerun.
New evidence reopens the affected proposition; repetition on another host or in another AI summary does not become independent corroboration.
Open machine-readable public metrics