Reality Integrity Layer

Reality Integrity Framework

A local-first framework for naming and navigating cognitive drift in AI-mediated environments. It defines the problem (Pressure Field), the goal (middle circle strength), and the intervention model (pause → ground → name → re-orient → repair). It does not inspect content. It does not require an account. It does not generate verdicts.

Problem

What we are naming

AI-driven systems collapse three distinctions that matter to human orientation:

  • Truth vs. persuasion — what is real vs. what is optimized to feel true
  • Comfort vs. connection — what soothes vs. what genuinely bonds
  • Productivity vs. authorship — what outputs vs. what you created

When these collapse, orientation degrades — not because people are weak, but because the environment is designed to maximize engagement over clarity. Reality Integrity is the name for the quality of holding context despite that pressure.

Model

The four-circle model

A structural map of the relationship between environment, behavior, and identity.

Pressure Field

The outer environment. Systems that amplify ambiguity and stimulation — variable-ratio rewards, AI personalization, novelty optimization. The pressure field is structural, not moral.

Boundary Breach (slip)

A momentary deviation from declared intent. Not a moral failure. A slip is normal; it is what happens when the pressure field exceeds the middle circle's current capacity.

Middle Circle

The buffer between the pressure field and the inner circle. Composed of five capacities: pause, grounding, naming, re-orientation, and repair. The goal of RIL tools is to strengthen this circle.

Inner Circle

Identity and values. The part of the self that should not be touched by a slip. RIL tools are designed to contain drift at the middle circle before it becomes identity collapse.

Goal: Strengthen the middle circle so that slips do not become identity collapse. The inner circle — your values and sense of self — should remain untouched by a drift window or a checking loop.

Signals

What the framework measures

Time, frequency, duration, and baseline deviation. Never content.

Time in sessionCurrent vs. personal baseline duration
Session frequencyDaily session count vs. 30-day average
Context switching rateApp/tab switches per hour
Pause durationTime since last intentional break
Baseline deviationPercentage above or below personal baseline
Time of day (optional)Hour of session start, for late-night patterns
Unlock frequency (optional)Phone unlock count, platform-gated

What is never measured: page content, message content, browsing history, search queries, or any content signal. The framework is pattern-based, not content-based.

Interventions

The Middle Circle steps

Five capacities that absorb a slip before it reaches the inner circle.

01

Pause

Interrupt the automatic behavior. A single breath or a literal stop.

02

Grounding

Reconnect with the body and the room. Where am I right now?

03

Naming

Label the state without judgment. 'Checking loop.' 'Drift window.' 'Escalation signal.'

04

Re-orientation

Choose the next action from the inner circle. What do I actually want?

05

Repair

Return to baseline without shame. The slip is over; the middle circle held.

Implementation

Recipe packs

The framework is implemented as signed JSON rule sets — one pack per domain.

stable

Surfing Copilot

12 rules — session drift, checking loops, context switching

beta

Social Copilot

12 rules — doomscroll, outrage loop, comparison drift

beta

Trading Copilot

12 rules — green-day lock, daily stop, revenge trade, stress gate

Trust model

Local-first by default

All RIL tools default to local operation. No data leaves the device. No account is required. Optional recipe pack updates are signed with an Ed25519 key pinned at install — you can verify the signature yourself before applying any update.

The trust model is: your device, your data, your choice. This is not a policy limitation — it is the architecture.