Selected AI builds / 2026

AI systems that take the busywork off your week.

Not pitch decks. Real systems that do the repetitive work, so you can focus on the decisions.

Four builds. One pattern: use AI where it helps, keep the human decision, leave proof.

01 AI system / Public example

A publishing system for a small publisher

A small publishing business was losing a week to each article. We built a system that finds the topic, checks what other people have published, gathers sources, drafts in their voice, makes images, and formats the post.

The team reads the finished draft and decides. Every article is handled separately, so one bad piece can be deleted without affecting the rest. The public example is on GitHub.

Public example 22 automatic checks 1 decision per post

Read the full case study

Client outcome: 1 month of work in 1–2 days (client estimate)

  1. DemandTopics, seasonality, search results
  2. DraftThe publisher's voice, sourced
  3. CheckStructure, claims, layout
  4. ImagesGenerate, resize, WebP
  5. SubstackFormatted draft, ready to read
  6. Final callThe team decides
02 AI readability / 2026

A blog that AI tools can actually read

Most websites give an AI tool the same noisy page a person sees: menus, pop-ups, and code. We made this blog return a clean text version of each article when an AI tool asks for one. The normal website is unchanged for people.

All 48 posts can be read this way. The site passed an external agent-readiness scan at Level 3, and one plugin handles every post.

Level 3 48 posts 1 plugin

See the blog Read the AI index

GET /transmissions/strategy/ai-tokens/
Accept: text/markdown

200 OK
Content-Type: text/markdown
Vary: Accept
X-Markdown-Tokens: 1965
text/markdown Vary: Accept noindex

HTML stays the default. The alternate is advertised, not forced.

03 Build / In progress

One source of truth for books, curriculum, and shop

A 26-book alphabet series, a homeschool curriculum, and an Etsy shop can easily become three separate businesses. We keep them in one place: the same facts, the same voice, and the same rules for every book, lesson, and product.

The system also creates the art, builds printable PDFs, and turns finished units into shop products. The publication is the front door; the books and curriculum are the deeper offer. Same owner, same business, one source of truth.

26 books Print-ready PDFs One source of truth

Ask for a curriculum sample

A to Q drafted. 26 planned. One source of truth.

04 Under the hood / Shared

An AI system built to keep costs down

Most AI systems get expensive because every step re-reads the whole project. Ours saves its work in files and passes the file name to the next step. The AI is used for the thinking; ordinary code handles formatting, resizing, and tests.

The same setup protects quality. Facts are checked, the voice is compared to a baseline, the layout is checked, and a person makes the final call. It costs less to run and is harder to break as it grows. Cheap is not the point. Durable is.

4 cost rules Code does the chores 28 formatting tests

See the public example Architecture notes

Where the work goes
Saved work
FilesThe to-do list, archive, and progress.
Chores
CodeFormatting, resizing, tests. No AI cost.
Thinking
AITopic, argument, review.
Final call
PersonOne decision per piece.

The pattern

Let code do the chores. Let people make the calls.

The tools change. The rule does not.

01

The boring parts are code.

Research pulls, image resizing, formatting, and checks run as code. AI is used for the open-ended work, not for what code does better.

02

People make the calls.

Topic, argument, voice, publish. One decision per piece, not a dashboard to babysit.

03

Proof over promises.

Tests, scans, and live artifacts. If it cannot be checked, it does not go on this page.

Start here

What is taking up your week?

Tell us where the work piles up. We will tell you what can be automated, what should not be, and what it would take to build.

Start a conversation

Some builds are internal or partner work. Names are withheld where the work is private. The systems are real. Need the service? See AI automation.