Evidence-led playbook / Draft

Make faceless technical content with evidence, not guesses.

A research-led system for original, mostly faceless explainers. We study how experienced creators make decisions, test what transfers to a solo founder working with Claude Code, and adopt only what the evidence supports.

Draft

Nothing here is adopted or validated. No pilot has been run, no content has been published, and the founder-time and cash caps are still unset.

01 / Playbook

The nine steps

Each step ends in a decision and a concrete artifact. We develop one step at a time against real work, starting with Step 0.

  1. 0

    Define the channel

    Proposed

    Audience, editorial promise, platform roles and limits. The charter and platform matrix are drafted; time and cash caps are unset.

  2. 1

    Select and validate the idea

    Not started

    Turn a subject into a specific viewer question and run the cheapest useful demand check.

  3. 2

    Research and substantiate

    Not started

    Build a source pack and claim ledger. Every consequential claim is supported, qualified, or cut.

  4. 3

    Design the story and test the hard part

    Not started

    Script or storyboard, plus the cheapest test of whatever could make the piece fail.

  5. 4

    Produce the episode

    Not started

    Build, edit, caption and export. Log active work, rework and render time separately.

  6. 5

    Review release readiness

    Not started

    Check what the final cut communicates: accuracy, comprehension, captions, rights and disclosure.

  7. 6

    Package and publish

    Not started

    Final title, opening frame, sources and disclosures. Publishing stays a separate founder decision.

  8. 7

    Measure and learn

    Not started

    Compare results with the episode's goal and separate what we observed from what we suspect.

  9. 8

    Standardize and scale

    Not started

    Promote only practices that held up across episodes: reuse, adapt, retest or retire.

02 / Research

What the research says so far

The source-and-practice register records what operators report, what their work shows, and what we still need to test. Every entry names the limits of its evidence.

Register v0.2

11 practices across topic selection, editorial review, production and sustainability. Ten are marked Verify and one is marked Defer. None is adopted.

Local tests: Not run

Pilot candidate

"One order. Two webhook deliveries." compares a plain diagram edit against a Claude Code + Lemo version of the same simulated trace. The question is whether animation helps viewers understand the problem enough to justify the extra work.

Pilot: Not run

How we decide what to adopt

Each practice is traced to the problem it solves, adapted to a solo-founder workflow, and tested against a simpler baseline before it becomes a default. We keep four things separate: what a creator reports, what their published work shows, what we infer, and what our own pilots demonstrate.

03 / Status

Open before the first pilot

Published pages will be added here as they are reviewed. Working notes stay private until they are explicitly prepared for publication.