Get more out of Claude,
taught in the open.
I teach Claude the way I actually use it. The whole method goes on YouTube, free, nothing held back. The finished artifact from each video lands in The AI Kitchen, my community. Not just for engineers.
I show the parts that break.
Most people film the demo that worked on the third try, which is the least useful version of it. So you get the run that failed and the reason it failed.
Two shorts are up: making Claude vote against itself with three independent agents, and why you can’t be scared of production. That’s the whole catalogue. I’d rather say that than dress up two videos as a library. Long-form starts now, with a CLAUDE.md that actually works.
@alextongme
You keep the files,
not just the lesson.
What lands in The AI Kitchen is the finished artifact from each video: the demo repo, the fill-in-the-blank template, the config you’d otherwise pause the video and retype.
It’s new and it’s small. That’s the honest pitch, you’re early.
In the classroom today
- A skill that audits your CLAUDE.md, scores it, and tells you which lines to cut.
- A script that puts cost and context in your status line.
Not an engineers-only room. If you use Claude for writing, research, or planning a business, the same thinking applies.
Honest numbers, not vendor numbers.
These are the numbers I can point at, from work I actually did. Small numbers I can prove beat big numbers I can’t.
Where the teaching comes from.
Software engineer for more than half a decade, previously at 
and 
. At The New York Times I was named one of the company’s AI Champions: I shipped Claude Code plugins the whole company adopted as the second-most-prolific contributor to the internal marketplace, delivered three presentations including a CLAUDE.md talk at the internal EngX conference, and taught 500+ engineers how to get more out of Claude. On my own team I wrote the AI-assisted migration script that turned a year-plus migration into a few months of work.
That work is what I teach now. Most of the last year went into finding out what Claude is actually good at, and the parts I got wrong go in the videos too.
# CLAUDE.md## How agents work in this repo- Agents draft. Engineers own every merge.- AI-authored diffs meet the same reviewbar as human code. No rubber stamps.- Architecture decisions live in /docs.Agents read them before writing code.- Secrets, migrations, and prod configstay human-owned. Always.$ claude commit
What the engineers I worked with say.
Verbatim recommendations. Every name links to a real profile; nothing here is anonymized or paraphrased.
“One of the key technical drivers behind our organization's adoption of AI-assisted engineering practices.”
“Consistently impressed by his ability to provide creative solutions for complex problems.”
“Alex's code is always well-documented, easily maintainable, and performance-optimized.”
The writing is still there.
Posts on Claude Code and AI engineering, from real production work. It’s not on a schedule right now. The videos are.
I still take a small amount of direct work.
Claude coaching for engineers and for people who don’t write code, workflow builds for the repetitive thing you’d rather stop doing by hand, and fixed-scope engagements for engineering teams. I’m not selling these hard right now, the videos are the focus, but the pages are real and so is the pricing.
After dark
My side projects, the games, and the free Eye Break app all live in Count Tongula’s Castle: the original vampire-themed cut of this site. Wander in whenever you like.
Enter Count Tongula’s CastleTell me what keeps annoying you.
Whatever you use Claude for, there’s one thing about it that consistently gets on your nerves. Post it in The AI Kitchen. That’s genuinely what decides what I film next, and I’d rather build the next video off your answer than off my own guess.
alextongme@gmail.com→I reply within one business day