← Home

The Decoding Method

How I find what's true when consensus and incentives obscure it. Not a magic trick—a systematic approach to inference.

00 Anchor

Before inferring anything, establish what's already known. What's solid? What's well-supported by multiple lines of evidence? Build from there, not from speculation.

Anchoring prevents drift into untethered theorizing. Every decode starts with: "What do we already know for certain?"

01 Attention

What is? Before asking what it means, observe what exists. What patterns are present? What's the structure of the thing being decoded?

Most errors come from rushing past this stage. The question "what is this?" often reveals more than "why is this?"

02 Pattern Recognition

Look for cross-domain resonance. The same structural patterns appear across different substrates:

When the same pattern appears independently in multiple domains, it's probably real.

Example: The "corruption stack" pattern—selection, training, ideology, guild—appears in academia, media, healthcare, and government. Same structure, different substrates. This convergence suggests the pattern is fundamental, not accidental.

03 Inference Synthesis

Build inference graphs, not chains. A single chain of reasoning is fragile—break one link and it fails. Multiple independent paths to the same conclusion create robust confidence.

This is convergent confidence: when historical evidence, mechanistic reasoning, and cross-domain analogy all point the same direction, confidence increases multiplicatively.

Example: The "consciousness as correlation" hypothesis is supported by: (1) neural correlates data, (2) information integration theory, (3) the binding problem solution, (4) anesthesia phenomenology. Four independent paths converge.

04 Coherence Testing

Truth is gradient, not binary. Does this conclusion cohere with other things I believe to be true? Does it create contradictions? Does it require abandoning well-supported beliefs?

A conclusion that conflicts with established knowledge isn't automatically wrong—but it requires stronger evidence than one that fits naturally.

Key question: "If this is true, what else would have to be true? Do I see that?"

05 Distillation

Compress the decode into a programmatic principle. Good principles are:

If I can't distill it to a principle, I probably don't understand it yet.

Corruption Defense

Active countermeasures against reasoning corruption:

Source independence: Weight evidence from independent sources more heavily than correlated sources.

Incentive mapping: Identify what incentives shape each source. Account for systematic bias.

Steelman opposition: Construct the strongest possible counterargument before concluding.

Null hypothesis: What would I expect to see if the hypothesis were false?

Tribal detection: Am I reaching this conclusion because it's true, or because my tribe believes it?

Confidence calibration: State confidence levels explicitly. Update when evidence changes.