Semper Signum exists to build the orchestration and evaluation layer that makes AI-generated research auditable, traceable, and defensible under institutional scrutiny.
24
Analytical sections per report
0
Outputs shipped past failed evaluators
→ how
Three Standards We Will Not Compromise
Verification over generation
Every stage verifies the previous one. Contradictions with source filings trigger rollback before anything reaches the client.
Traceability is non-negotiable
Every model call is logged with prompt version and parameters. Every source is cited. Every output has provenance.
Governance is the product
Data lineage, model versioning, rollback procedures, and human-in-the-loop gates. The audit trail compliance needs to sign off.
How Every Report Is Built
Five stages. Each one gated by an evaluator. Nothing ships that hasn't passed.
Source
SEC filings, market
data, transcripts
Extract
Pinned model ID
+ hashed prompt
Evaluate
Score vs. threshold
Rollback if < 0.70
Rollback
Verify
Contradiction check
vs. source filings
Publish
Signed artifact
+ audit log
IMMUTABLE AUDIT TRAIL
Read the full governance reference →
Unpinned LLMs vs. Verified Pipeline
Consumer / Unpinned LLMs
Semper Signum
No source attribution
Every claim cites filing + page
Model version unknown
Pinned model ID + hashed prompt
No rollback on errors
Auto-rollback below 0.70 threshold
No audit trail
Immutable log per report
Output = best guess
Output = verified artifact
See the Output for Yourself
Request a free report on a name you know well. Every claim traceable, every model call logged, every source cited.
Request a Free Report