← Home

The Edgecraft Thesis

Most knowledge lives in unstructured form — books, podcasts, hard-won experience. Experts can't always articulate what makes them elite. The things that actually matter are buried, implicit, and never taught directly.

Edgecraft is the practice of systematically extracting, structuring, and operationalizing that knowledge. Not notes. Not a wiki. A decision-grade knowledge graph.


The Pipeline

Four steps, applied to any domain. The output is always the same: a structured graph of skills, prerequisites, diagnostic trees, and non-obvious edges.

01

Find Experts

Identify the best sources: books, podcasts, coaches, practitioners with proven track records. Not influencers — people who have actually done the thing at an elite level and can articulate why.

02

Ingest Work

Extract structured knowledge: skills, prerequisites, diagnostic trees (symptom → root cause → fix), progression levels, coaching cues. A 500-page book becomes a structured knowledge graph, not a summary.

03

Extract Edges

Surface the non-obvious: where conventional wisdom is wrong, hidden causal levers, elite-only behaviors that nobody teaches explicitly. The convergence of multiple expert frameworks IS the insight.

04

Build & Test

Productionize the signal: build apps that operationalize discovered edges, test against real outcomes, measure decision quality. The user walks away knowing what to DO, not just seeing data.


Why This Exists

Flat notes don't capture prerequisite chains or diagnostic relationships. You can read ten books on a subject and still not know what to practice next or why you're stuck. The problem isn't access to information — it's the structure of information.

Inspired by Math Academy's prerequisite-gated knowledge graphs and Alpha School's mastery-based progression. Both systems share a core insight: learning is a graph, not a list. You can't learn calculus without algebra. You can't fix a flinch without understanding trigger control. Prerequisites matter.

The diagnostic tree is where this diverges from traditional knowledge management. Every skill has failure modes. Each failure mode has a root cause. Each root cause has a specific fix and a coaching cue — the exact words a coach would say in the moment. This is what separates structured knowledge from notes: it tells you not just what to do, but what's wrong and how to fix it.


Before → After

What happens when raw expert knowledge goes through the pipeline.

Raw Transcript

"...the big thing with grip is, you want to hold it as loose as you can force yourself to with your dominant hand. Most people squeeze way too hard. Your support hand does the clamping. When the support hand disconnects, that's when the dominant hand starts compensating and everything goes sideways. The thumbs should just float — I can always do nothing more consistently than I can do something..."

Structured Knowledge

Skill: Grip

Category: Marksmanship | Level: 1

Prerequisites: none (foundation skill)

Diagnostic:

Symptom: Groups open under speed

→ Root cause: Support hand disconnecting

→ Fix: Asymmetric pressure drill

→ Cue: "Loose dominant, clamp support"


The Edge Philosophy

Decision quality is the metric

The user walks away knowing what to do, not just seeing data. Every feature is evaluated against: does this make the decision surface clearer or noisier?

Leading indicators over lagging outcomes

Submission timing patterns over final grades. xG over final scores. Engagement drift over dropout. Optimize for the signals that predict outcomes.

Revealed preferences over stated preferences

What people actually do — click, use, buy, cancel, cobble together — matters more than what they say. Compensating behaviors reveal the real job.

Multiple frameworks, converged

Apply multiple analytical lenses to the same problem. The convergence of independent frameworks IS the insight. Single-framework analysis is shallow.

Help me see what others miss

Every product serves the same job. Synthesize data into a clear decision. The edge is in the extraction and structure, not the raw information.