Marty Sampson.aiAI, applied
Explainer · source-backed

Private notes are not a publishing queue.

Capture preserves context. Publication requires a different chain of decisions.

An AI assistant can organize Markdown, connect notes and bring old material back into view. None of those capabilities answers the public questions: Is this fact accurate? Is it mine to share? Does it expose someone else? What original contribution makes it worth publishing?

An original M-brand explainer

A note crosses five gates before it becomes public

  1. PrivateCaptureKeep the source, date and context.
  2. PrivateClassifyMark sensitivity, ownership and intended use.
  3. ReviewVerifyCheck claims against current primary sources.
  4. ReviewTransformWrite an original explanation with safe examples.
  5. PublicReleaseApprove exact bytes, publish and verify the result.
No automatic promotion.A note, tag, agent summary or generated draft does not authorize the next gate. Revisions after review need a fresh release decision.
This is an editorial control model, not a diagram of an installed automatic publisher. The public artifact should retain evidence and correction paths without exposing the private source record.

One invented note, three possible outcomes

Synthetic example · no real vault content

A fictional daily note says: “Try smaller instruction files. The assistant missed the testing rule after a long setup note. Screenshot attached.”

The note is useful privately because it preserves a question and the moment it arose. It is not yet a public claim. The screenshot may contain unrelated tabs or account details. “Missed” could describe one prompt, one product version or a mistaken expectation.

Keep private

The note contains another person’s message, confidential project context or a screenshot that cannot be safely cleared.

Research next

The idea is promising, but the product behavior, version and causal explanation have not been verified.

Publish a new piece

Reframe the question, cite primary documentation, create a synthetic test and report the setup, miss and limitations.

The publishable article is not a cleaned-up copy of the note. It is new work. The source record can remain private while the public piece explains the method and evidence.

Keep four decisions separate

Retention
Should this item remain available privately, and for how long?
Retrieval
When may an assistant bring it into working context?
Editorial use
May its idea inform a new draft without copying protected or sensitive material?
Release
Are the exact public bytes, sources and assets safe and ready now?

Combining those decisions creates a dangerous shortcut: “the assistant found it, therefore it can publish it.” Retrieval establishes relevance, not ownership, accuracy, consent or approval.

A compact publication record

For each public piece, keep a record that can answer:

  • Which stable private source IDs inspired the question?
  • Which claims were independently checked, against which primary sources, and when?
  • What was excluded for privacy, rights, uncertainty or lack of relevance?
  • Which text, examples and diagrams are original?
  • Which exact artifact was released, and how was the live result verified?
  • How can a correction replace or retire the public version without rewriting the private history?

This record need not be public in full. Its job is to make the public artifact explainable and reversible.

What the sources establish

Open files make agent work possible

Kepano’s Obsidian skills repository documents agent skills for Markdown, Bases, JSON Canvas and Obsidian’s CLI. It shows that an agent can work directly with open vault formats; it does not make vault contents public or publication-ready.

Automation raises the review stakes

The obsidian-second-brain repository describes capture, retrieval, synthesis and scheduled maintenance over a Markdown vault. Those are the project’s documented mechanisms and claims. The gate model above is an independent editorial design, not a claim that this repository supplies publication approval.

Different projects share the same private-note premise

Claudesidian, second-brain-skills and claude-obsidian offer neighboring approaches to agent-assisted Markdown knowledge work. They are references for the design space, not evidence that one workflow fits every vault.

Primary repositories checked. No repository was installed or tested for this article. No private note or captured screenshot was reproduced.

Related: Context is not memory →

Back to field notes