You don’t have to take the description on trust. Read the tests, open the source material, and see what the results establish—and what still needs review.
01 / Tests & results
What has been tested?
These records show what went into a test, how it was run, and what came out. The dates and limitations stay with the results.
A recorded BriefWise test
Run it again. Compare the result.
In this local test, BriefWise DI² repeated its checks on a saved, fictional legal matter. It reproduced the evaluation and approval result, with an exact match to the saved replay identity.
The test also rejected a deliberately altered copy. The panel shows the recorded checks; the source explains how the test was run and what it does—and doesn’t—establish.
9 September 2026 · Local kernel 0.2.0. The record uses a saved synthetic matter with live release disabled.
What this test doesn’t establish The same local rule implementation was used. The original source bytes and history of authorization remain unverified. This test does not establish legal correctness, identical model generation, private model reasoning replay, or independent legal evaluation. External independent verification remains unverified.
The public package includes a package-integrity/event-chain checker. It does not include the kernel implementation used for the recorded run.
Two fresh runs produced the same decision record. A separate local checking program repeated the calculations and caught deliberately altered calculations.
Per-run dates in source; total corrected 21 September 2026
Status
Source-reported local checks
The published series reports 83 / 83 prompt-level local passes and 188 / 188 local checks across 83 positions (80 distinct source keys and three repeats). Official-evaluator executions are recorded for 27 of the 83 positions; no independent external verification is reported.
The 188 local checks cover 149 official instruction entries: later records count individual keyword terms separately. These are bounded local results, not an official 188-instruction score or a claim of general performance.
Why the total changed from 187 to 188
Book 10, reconciled edition v1.1, page 24, identifies an arithmetic error: the 83 source rows already sum to 188. The repository corrected its README on 21 September 2026. No recorded response or per-row count changed, and the reconciliation is not a new evaluator run.
The five technical PDFs previously hosted here have been withdrawn from the current Site release. The public exhibit collection remains linked to its repository.
The evidence and decision library
Application
Law, weather, mathematics, and governance
Record date
September 2026 collection
Status
Source collection / individual record statuses vary
The collection brings together nine books and 260 source pages on the history of the work, mathematics, weather, law, and governance. The catalog will help you find the right title and edition.
The number of pages doesn’t establish a result. Check each record’s date, limitations, and review status.
Search by field, application number, or title. Each entry has its own date and status. Filing an application does not mean a patent has been granted or that performance has been proven.
Look in the September catalog for “Before Reality Resolved — DI Weather Station.”
September 2026 collection · Release Candidate 1
Not publication-approved. PDF pages 10–11 and 16 distinguish local observations from same-site verification. No forecast guarantee or independent operational validation is implied.
Each interactive demonstration focuses on a particular check. Its own notes explain the setup and limits; using a demonstration is not an independent test.