Independent thinking. Documented progress.Field notes on working with AI

The field journal

Every edition.
Every lesson.

A growing record of work with AI: what moved forward, what the evidence says and what remains unproven.

New editions and resources are published when ready. Nightly website publishing is not enabled. The RSS feed changes when an edition or resource is published.

Published editions

Public web publication began October 13, 2026. This archive lists the editions currently available.

Development notebook · October 13, 2026

Development note: from a launch to a learning library

A same-day development checkpoint, separate from the historical launch record.

Published October 13, 2026. This same-day development note describes the public site's version 8 library checkpoint, before the present editorial expansion. The original launch record above keeps its own scope; its historical check counts are not totals for this expansion.

What changed

The site grew into three full learning collections: 10 Educator workflows, 10 Agentic Systems guides and 10 Workspace Development guides. The five journal companions remain a separate, smaller collection. Each of the 30 counted learning artifacts has a complete method, a worked example, checks, public source notes and a Markdown download. Hubs, the educator repeat log and alternate file formats are not extra artifacts.

The change is useful because it creates a route from a question to a concrete output. A teacher can move from a learning goal to a coherent packet and source check. Someone considering an agent can compare it with a simpler workflow, define permissions and rehearse a bounded loop. A person assembling several documents can establish a source set, coordinate drafts and audit consistency.

What the checks actually establish

For version 8, a local inventory counted the 30 artifact records and their routes. Its local release checks covered content and download consistency, internal routes, feed structure and selected interface behavior. A separate anonymous browser spot check opened the homepage and all three collection hubs at a 1180 × 757 desktop viewport. Those four pages rendered readably without horizontal overflow; each hub displayed 10 artifact cards, and the Learning menu opened.

On one Workspace Development guide, the copy control displayed its success message and its Markdown download contained the expected title. The clipboard contents were not independently compared byte for byte. This is a narrow interface observation, not an end-to-end check of every copy control or download.

These observations do not verify the revised pages accompanying this note. No successful live 404 verification was established in the version 8 browser pass. Mobile layout, physical printing and assistive-technology behavior also remain unverified here. A local check and a desktop spot check should not be relabeled as a comprehensive accessibility audit.

Three lessons from the expansion

Count useful units, not packaging. One complete method may have a page, download, example and source notes. Counting each representation separately would make the collection sound larger without giving a reader another method.

Make the first decision easier. A question such as “How do I check this lesson?” provides a more useful entrance than a long list of unfamiliar terms. The resource routes connect neighboring methods while letting readers stop after one reviewed result.

Keep demonstration and evidence separate. A synthetic example can show how to inspect a decision. It cannot establish a learning gain, time saving or reliable deployment. The next evidence must come from an actual attempt, with its conditions and limitations recorded.

The next question

Can an unfamiliar reader find the right method, understand its limits and complete the first exercise without extra explanation? A proposed reader study would record the route taken, the output produced, points of confusion and the reader's own account of the evidence limits. It has not run. Until it does, the claim is that the materials and navigation exist, not that the collection has proved its learning impact.