IS Home About Work Verity Books Contact
Infinitive Storytelling  ·  The verification engine

VERITY

A verifiable second brain — audit-grade, governed knowledge infrastructure, delivered as a service.

See how it works
Scroll  ·  The structure beneath the story
FIG. 01 — The problem

Most "second brains" are a search box over a pile of documents

No guarantee an answer is sourced. No way to tell good data from broken data. No record of what changed. No owner for the AI quietly touching the data.

Verity is built around those gaps, not the demo. Every finding is schema-validated, every source is frozen and traceable, the pipeline reproduces identical output on re-run, and the system audits itself before it's trusted.

UNSORTED — GOOD MIXED WITH BROKEN ? SEARCH BOX — NO PROVENANCE “answer” SOURCE? — THE TRAIL GOES NOWHERE
FIG. 02 — Part 01 / What was built

Source → schema → validator → migration → roll-up

A pipeline where nothing enters the brain unless it passes the gate

A complete pipeline: raw documents land in an inbox, get validated against a pinned JSON-Schema contract, and migrate into a frozen, read-only canonical store. From there, two independent passes run against the corpus — a stateful spread/fingerprint analysis and a separately-logged closure red-team — each writing only to its own folder, so every stage is re-runnable and auditable on its own. One Bash-orchestrated command runs the full loop and produces identical checksums on every re-run.

INBOX VALIDATE SCHEMA GATE CANONICAL ANALYSIS RED-TEAM ROLL-UP SELF-AUDIT LOOP — CLOSURE BLOCKED UNTIL LIMITATIONS ARE DOCUMENTED
FIG. 03 — Part 02 / Why it's different

Verifiable. Reproducible. Governed.

Three properties most AI knowledge tools can't claim

03.A — Verifiable

Gated, not hoped for

Schema conformance is enforced at the door. Malformed data is excluded, never quietly mixed in with good data.

03.B — Reproducible

The same answer twice

Run it twice, get identical output. Identical checksums on every re-run — not a chatbot that improvises a new answer each time.

03.C — Governed

Owned and risk-tiered

Every component is owned and risk-tiered, and an independent adversary attacks the corpus's own findings before anything is declared closed.

FIG. 04 — In focus / The schema gate

Nothing enters the brain unless it passes

Verifiable at the door · reproducible on re-run

INTAKE CANONICAL SCHEMA GATE EXCLUDED

Conforming documents pass into the canonical store. Malformed data is deflected at the gate and excluded — never quietly mixed in with good data.

RUN 01 a3f9 4d31 c7e1 RUN 02 a3f9 4d31 c7e1 = IDENTICAL CHECKSUMS

Run the same corpus twice and the pipeline returns identical checksums. The brain gives the same answer twice — not a chatbot improvising a new one each time.

FIG. 05 — Part 03 / What's real

Proven now — and an honest roadmap

A working prototype, single-tenant, run end-to-end on a real corpus

Proof artifacts · proven now
  • Canonical instrument with pinned weights and version lineage, enforced by a preflight guard
  • JSON-Schema contract and the validator that gates against it
  • Idempotent orchestrator — one command runs the full loop
  • AI governance register — every system owned, risk-tiered, zero orphans
  • Self-audit / teardown pass that blocks closure until limitations are documented
  • A multi-document corpus processed end-to-end through every stage
Honest roadmap · not yet built
  • Per-vertical instrument profiles — same core, swappable domain rules per industry
  • Multi-tenant isolation — per-client data and findings separation
  • A thin, read-only client-facing retrieval skin — the "ask your brain" front door
  • Scheduled per-client runs tied to a retainer, plus billing and SLA
"A system that documents its own limitations is the product thesis, applied to itself."
The prototype is real and single-tenant today — the productization path is scoped and sequenced, not speculative.
IS

Put a verifiable brain behind your documents

A scoped pilot points Verity at one real corpus and returns source-traceable findings you can check, line by line.

Discuss a pilot Book a 20-min call
Read the technical brief →
Infinitive Storytelling  ·  storiesfordays.com
0
Skip to Content
Home: Do the Work Modules
Do the Work Module 1
Do the Work M2
Do the Work M3
Do the Work M4
Do the Work M5
Do the Work M6
Interactive Canvas
Do the Work Download
Infinitive Storytelling
Home: Do the Work Modules
Do the Work Module 1
Do the Work M2
Do the Work M3
Do the Work M4
Do the Work M5
Do the Work M6
Interactive Canvas
Do the Work Download
Infinitive Storytelling
Folder: Do the Work Modules
Back
Home: Do the Work Modules
Do the Work Module 1
Do the Work M2
Do the Work M3
Do the Work M4
Do the Work M5
Do the Work M6
Interactive Canvas
Do the Work Download