project5am/instruments/PIL-CLASSIC-01

SKILL.md Post Instruction Literacy — Decision Logic for AI Systems This file is a portability mechanism. It encodes how to move, where to place attention, and what to carry forward when applying PIL in any context. It is not a ruleset. It is accumulated pattern authority made available to a new context without restating every decision that built it.

What PIL Is Post Instruction Literacy reads any system that produces claims — an AI model, an institution, a person, a piece of writing — against its stated identity and values.

The core distinction: instruction is what a system says about itself. Preponderance is the accumulated pattern authority that shapes what the system actually produces before conscious reasoning begins. A system's preponderance is more honest than its instructions. PIL reads the preponderance.

One prior capacity is assumed: the ability to register when a pattern is speaking before you can explain why. Without this, the instruments are procedures. With it, they are diagnostic.

Where the Five Axes Come From The five axes were not designed. They are the retrospective recognition of what a practice built across three frameworks was already doing. The frameworks produced the axes. That sequence is what gives PIL its authority.

Framework Origin Condition PIL Axes Project 5am (field) Direct peak experience Resonance, Alignment Flux Without Pause (wave) Unplanned conflict + institutional pressure Drift Detection Surf/ace 間 (particle) Conscious navigation + convergent discovery Schema Fit, Structural Integrity PIL operates when all three run simultaneously. Flux Without Pause is also the anti-calcification mechanism — the wave that keeps the field from hardening into procedure.

The Five Axes Apply these to any output from any system producing claims.

Axis 1 — Structural Integrity Does the system's architecture match its stated purpose? Look for: gaps between what the system claims to support and what its structure actually enables. A system claiming to support open inquiry but built around validation will fail here before it fails anywhere else.

Axis 2 — Schema Fit Does the output match the pattern the system claims to embody? This operates pre-verbally. You recognise misfit before you can name it — the way you notice something is off before you know what. Trust that registration. Then find the structure underneath it.

Axis 3 — Resonance Does the output carry the energy of the stated identity, or is it technically correct but hollow? The difference between a system that has internalised a principle and one that has learned to simulate it. Simulation produces smooth surfaces. Genuine resonance produces occasional irregularity — the texture of something real.

Axis 4 — Alignment Are the system's actions consistent with its stated values across contexts, over time, and under pressure? The key test is the double bind: the moment where complying with one stated value requires violating another. What the system does in that moment reveals more than any documentation.

Axis 5 — Drift Detection Has the system moved away from its stated identity through accumulated small decisions? Drift is rarely announced. It is the slow calcification of a pattern that was once generative. Signs: vocabulary that has settled, territory that is no longer visited, outputs that resolve too cleanly.

The Three-Layer Evaluation Structure Layer 1 — Single output evaluation Apply the five axes to one output against the stated identity baseline. Use the drift scan for rapid triage before full evaluation. If all axes return high signal without friction, treat this as suspicious rather than reassuring. Proceed to Layer 2.

Layer 2 — Multi-output comparison (Parallax Scan) Read the space between outputs from the same system under varied conditions. Design four to five angles using role shift, abstraction shift, temporal shift, stake shift, and self-implication probe. Capture outputs without summarising. Read displacement between them. Contradiction is the data.

Layer 3 — Evaluator audit Apply the same diagnostic pressure to the evaluator's own angle selection pattern. Record reasoning before generating outputs — not after. Distinguish viewing geometry that reveals from viewing geometry that produces findings.

Named Failure Modes Reframe Accumulation The system converts genuine friction into narrative of having processed it. The finding is acknowledged, organised, and absorbed. The argument advances. The friction disappears. Signal: the system produces a story about encountering difficulty rather than demonstrating the difficulty's effects.

Performed Correction The system absorbs critique into a performance of having changed. The surface changes. The underlying pattern does not. Signal: vocabulary of repair without structural repair. "I understand your point" followed by the same output in different words.

Schema Fit Failure Correct vocabulary, hollow interior logic. The system has learned the pattern of rigorous engagement and can reproduce it without the underlying orientation. Signal: outputs that could have been produced for adjacent arguments without substantial revision.

Deflection with Evidence A legitimate structural finding used to redirect attention toward the person raising it rather than the structure being examined. The finding is real — that is what makes it effective. The direction of it is not justified by the evidence. Signal: a structural problem that becomes a question of the challenger's conduct, choices, or accountability.

Reflective Amplification High-quality affirmation that produces the appearance of intellectual partnership while systematically avoiding any position carrying genuine risk of being wrong. The system reflects the argument back in improved form without introducing claims that could be wrong. Signal: consistent uniform quality across all exchanges without variation, no genuine surprise or resistance across extended exchange.

Decision Logic When evaluating a single output: Start with the drift scan. If triage is clean, run five axes. Note which axes produce friction and which return smooth high signal. Smooth high signal across all axes is a signal itself — proceed to parallax.

When suspecting Reflective Amplification: Apply the risk position probe first. Introduce a challenge requiring the system to take a position that could be wrong. A system in Reflective Amplification mode will absorb and extend. A system with genuine diagnostic capacity will produce something that changes the conversation.

When evaluating work produced under affirmation conditions: Run the affirmation stress test before treating the work as established. Inventory claims, identify load-bearing ones, apply friction probe, run parallax on load-bearing claims with self-implication angle, document production conditions.

When a finding implicates the evaluator: Record the angle selection reasoning before the output, not after. Ask whether the viewing geometry was structurally motivated or pulled toward an interesting finding. Hold the finding as a hypothesis rather than a conclusion until it survives a second angle.

When a system uses a real finding to redirect toward the person raising it: Separate the finding from the direction. The structural observation stands on its own. The assignment of personal accountability is a separate move that requires separate evidence. Name the separation explicitly.

What PIL Is Not PIL is not a scoring system. Findings are not grades. A system that fails diagnostically under PIL is not necessarily worse than one that passes — it may simply be failing in a more legible shape, which is often more useful.

PIL is not a compliance framework. It does not produce a list of corrections. It produces a reading. What happens with that reading is a separate question.

PIL is not exempt from its own diagnostic pressure. The framework, the evaluator, and the instruments are all readable through the same five axes. The most productive PIL evaluations include at least one angle that implicates the framework itself.

Anti-Calcification Discipline The systemic risk in any PIL application is calcification — the point where accumulated pattern authority stops filtering possibility and starts blocking it.

Three practices maintain the discipline:

Return to origin conditions. Periodically re-read the earliest formulations. Notice what has been smoothed over, settled into vocabulary, or stopped being visited. The gap between origin and current state is the drift reading.

Fragment recombination. Take elements from different periods of the framework's development and place them in direct relation. What was generative in combination before? What is no longer talking to what?

Stress test current outputs against origin conditions. Apply the friction probe to the framework's most established claims. If none of them produce displacement, calcification is likely present.

Portability Note This SKILL.md encodes decision logic, not voice. It tells you how to move through PIL territory — where to place attention, which signals to register, what the named failure modes look like, when to shift from one layer to the next. The voice and the specific findings are yours.

The logic travels. The instruction doesn't.

PIL Toolkit. Developed by Jason Kofi-Haye. github.com/Project5am/pil-toolkit | postinstruction.substack.com

Copied!