BYBLOS

Decision infrastructure for AI systems.

AI can reason.

BYBLOS determines how decisions are made — across policy, jurisdiction, evidence and human authority.

Turn probabilistic intelligence into governed, reproducible and auditable decisions.

Explore BYBLOS Get early access

Decision architecture / representative trace

A decision, from evidence to record.

See how evidence becomes a governed decision — and how every transition remains inspectable.

Trace / D-240826-071Inputs hash / 87bd…93fa

Evidence87bd…93fa

Observationeditorial / commercial

RuleR-BR-004 · v18

Authorityhuman required

Decisionadapt

Outcomedecided

Ledgerrecord sealed

Authority requiredAwaiting actor

Actorhuman

Actionratify

Actor refpending

Statusrequired

ProposedRatifiedActiveApplied

Evidence enters with a parent.The source is mandatory. Rights and provenance arrive before any model is asked to reason.

Stage01 / 05

AuthoritySource owner

StateBound

Select any stage to inspect its effect. The automatic trace represents causal order.A model may propose. It cannot silently grant itself authority.

Ledger principle

Every decision can answer:

What happened.
Why it happened.
Which evidence supported it.
Which rule governed it.
Who authorized it.

Ledger / durable decision record

Technical credibility lives in the record.

A governed decision is only useful if its evidence, rule, authority and outcome can be reconstructed later.

Ledger event / representative interfaceRecord sealed

Decision: D-240826-071
Market: BR
Evidence: 87bd…93fa
Rule: R-BR-004 · v18
Authority: human / ratified
Verdict: adapt
Provenance: complete
Rule state: applied

Causal chainComplete

Evidenceevidence_ref / 87bd…93fa
Observationeditorial / commercial
Rulerule_ref / R-BR-004@v18
Authorityactor / human · ratified
Decisiondecision_ref / D-240826-071
Outcomedecided · verdict adapt
Ledgerledger_event / append-only

Models change. Your organization shouldn’t start over.

Models are replaceable infrastructure.

The compounding asset is what your organization learns around them: Knowledge. Authority. Causality. Verified execution history. Operating procedures.

BYBLOS preserves this institutional intelligence independently of the models producing it.

As models improve, providers change and agents are replaced, the organization retains what it has learned — what was decided, why it was decided, who authorized it, what happened next, and which procedures proved reliable.

Model independence / controlled comparison

Change the model. Keep the intelligence.

Classification is probabilistic. Decision structure is not.

Probabilistic layer / replaceable

Confidence: 0.82
Classification: editorial / commercial
Reasoning trace: trace / mt_41a
Interpretation: commercial context
Proposed output: adapt

Classification is probabilistic. Decision structure is not.

BYBLOS / institutional layerFixed across model change

Evidence identity: 87bd…93fa
Rule version: R-BR-004 · v18
Human authority: ratified / verified
Policy state: active
Prior decision: D-240826-071
Ledger history: retained

The probabilistic layer changed. The institutional layer did not.Baseline / Model A

BYBLOS