ContinuousLogic
Ask about the belief audit

The anchor artifact

The operating map

Your AI assistant is confidently telling people things that are no longer true. You have no way to know how often, no way to stop it, and no record of who approved what it believes.

The map sets out where the answers come from, the six stages every signal passes through, the operating rules, and the six questions to ask before trusting an AI output.

Ask about the belief audit Read the working papers

Designed as a printable canvas at 1200px. Scroll sideways on a narrow screen, or print this page for the poster.

The operating map

Your AI believes things nobody approved. This is change control for what it believes.

ContinuousLogic
Context flowRaw signal in
Trust flowVerified, usable
Permission flowWhat can be used
Feedback flowCorrections loop back
The operating layer, six stages
01
Quarantine
Captured. Not trusted.
  • Every artifact fingerprinted on capture
  • Screened for prompt injection before anything downstream reads it
  • Nothing enters active use by default
  • Origin and boundary recorded for every artifact
"Capture is suspicious until proven otherwise."
02
Classify
Weigh the source.
  • Every source assigned a weight; not all sources carry equal say
  • Injection risk labeled: none, low, medium, high
  • Reliability tracked over time, not assumed
  • Signed sources separated from unsigned
"Not all sources deserve the same say."
03
Retrieve
Context with provenance.
  • Evidence assembled around the question, not loose fragments
  • Every excerpt carries its source and its weight
  • Contradicting evidence travels with supporting evidence
  • The model reads context; it cannot rewrite it
"The model sees evidence, not opinions."
04
Challenge
Changes are contested before they take effect.
  • Triggered by staleness, contradiction, or impact
  • The case for keeping and the case for changing are both argued
  • Supporting and opposing evidence weighed together
  • Ruling recorded: confirmed, revised, retracted, or held for review
"A change must survive challenge before it is accepted."
05
Enforce
Rules in code, not prompts.
  • Every state change checked against thresholds before it lands
  • Changes require corroboration from independent sources that meet the threshold
  • Caps on how much a single update is allowed to change
  • Human approval required on high-impact changes
"Prompts can be talked around. Rules in code cannot."
06
Commit
Every change on the record.
  • A permanent record of every change, with justification and sources
  • Current state can be rebuilt from the record and checked against it
  • Every belief carries a review-by date; nothing is permanent by default
  • Every answer traces back to who approved what, and when
"Every change is replayable. Every belief expires."

What compounds once beliefs are governed

Audit-ready by default.
Every decision traceable to evidence.
Drift detected, not discovered.
Staleness surfaces before damage.
Injection does not survive.
Screened on capture, tested on change.
Nothing changes silently.
Every change lands on the record.
Confidence is earned.
Source weight plus corroboration.
One governed answer.
The same source serves every surface.

Broken model versus working model

Broken Prompt Assistant Confident answer Wrong belief spreads Damage discovered later
Working Quarantine Classify Retrieve Challenge Enforce Commit

Beliefs are governed objects, not loose text·Evidence is weighed, not equal·Changes are contested, not assumed·Rules are code, not prompts·Changes are recorded, not edited away·Change control is the moat