Seven agents will plan, query Zetaris, challenge each other's evidence and write an answer you can audit.
No investigation yet. Ask a question above, or pick one of the example cases.
Every number in the answer is computed from the rows the agents retrieve, and every claim that lacks data is struck out rather than asserted.
Hypotheses tested
Each starts untested and is decided only by the evidence retrieved in this run.
Accounts, reconciled across sources
SEC EDGAR filings joined to the SIC taxonomy, ERP billing and support telemetry. Highlighted rows were orphaned by the taxonomy and restored. Negative amounts in parentheses.
| Company | CIK | EDGAR SIC | Reconciled | FY25 revenue | Q3 ARR | Net retention | ARR change | P1 tickets | p99 latency | Timeouts | Q4 weather exposure |
|---|
Evidence ledger
Every statement sent to Zetaris, which hypothesis it served, and whether it returned rows. A struck mark means no evidence, so no claim was allowed.
| Pass | Agent | Purpose | For | Rows | Statement |
|---|
Query the data
Run read-only Lightning SQL against the same tables the agents use. Every statement passes the Scout guard first, and anything that writes is refused.
Results appear here.
Engines
Live state of the four Meterless engines: what memory holds, how much context the bounded reasoning saved, what the shared world model recorded, and which contracts the guard signed.
Memory (H-MEM)
Loading…
Bounded reasoning (Markovian)
Loading…
Shared world model
Loading…
Signed contracts (Scout)
Loading…
Connections
Point AuditWhy at your own Zetaris instance and Meterless Nightshift runtime. Keys stay in this server's environment and are never shown again. Your NVIDIA key belongs in Nightshift's Providers screen, not here.
Status
Loading…