If you haven’t heard from us in a while or at all, it’s because of the thing I wrote about recently. An old automation that pushes °FAI to subscribers quietly broke. We kept publishing. You just stopped seeing it.

The fix is in. The quarterly audit is on the calendar. While we’re here, let me re-introduce what °FAI actually is.

What °Fahrenheit AI (°FAI) Is

Fahrenheit AI is a three-year old experiment in understanding generative AI from inside the work. It started with a series on YouTube, Diving In and Learning AI. 

It evolved to blog posts of the custom projects that I develop with AI for screen print shops. These posts are drafted by AI. Specifically Claude. They’re written in my voice, a voice I’ve been training, analyzing, and refining for three years. I bring the topic, the technical work I’ve actually done, and the editorial judgment. The model drafts faster than I can type. I edit until it sounds like me. Sometimes that’s a comma. Sometimes it’s a full rewrite. It also lets me focus on the project more than the writing.

That feedback loop, building the workflow, watching it fail, fixing it, documenting it, is the whole experiment.

What We Cover

AI tools and automations that actually work in a print shop. We documented three automations running my computer without me Downloads cleanup, screenshot sorting through a multimodal LLM, TimeTracker exports auto-formatted into Excel. We covered building slide decks with Claude by dumping context first instead of asking for “a 15-slide deck on screen printing safety.” There’s a Claude Vision spell-check workflow for artwork, an Ideogram drop-folder pipeline that turns raw generations into production-ready files, and Google Sheets production reports that grade themselves. Same pattern in all of it: find the repetitive thing, write explicit rules, audit the output.

The Side-Eye Stuff

Where industry hype doesn’t survive contact with the data. Klarna and Salesforce dressing up regular layoffs as AI restructuring while in 2025 Oxford Economics found 4.5% of 2025 U.S. job cuts actually cited AI. Meta installing keystroke and screenshot tracking on employees to train its agent models. Sycophancy as a design flaw, the reason a chatbot will confidently tell you exactly what you came in believing. Phony orders getting smarter and harder to spot. The TAKE IT DOWN Act as the entire federal AI law on the books. Skepticism, with sources.

The Fair-Is-Fair Stuff

Where AI is doing measurable good right now. AlphaFold solving a 50-year protein folding problem and winning the 2024 Nobel in Chemistry. A Lancet Oncology study finding AI-assisted mammography catches 20% more breast cancers without raising false positives. BlueDot flagging the Wuhan pneumonia cluster nine days before the WHO. The places where the upside is real and the math is on the table.

The Workflow Is the Point

The real thing °FAI tests: can you scale your own shops with AI without losing what makes it yours?

Three years in, the answer is absolutely yes, with discipline.

If you missed us, we missed you too. The automation is fixed. Welcome back.