I have a rule I keep relearning. The first time I do something, I do it by hand. The second time, I stop and turn it into a system. This week was mostly second times: podcast editing, follow-up emails, the welcome sequence. None of it was glamorous. All of it now runs with a lot less of me in the loop, which is the whole point of a one-person shop.
Episode 02: What Teachers Actually Need from AI
Episode 02 is live: What Teachers Actually Need from AI. I grew up in a construction household, and nobody ever reached for the cordless nail gun first. Charge it, load it, set it up, and you have burned twenty minutes before driving a single nail. You grab the hammer and start framing. That is the test I run every AI tool through, and most of what gets marketed to education fails it. Districts are signing six-figure subscriptions for features nobody asked for, while the thing teachers actually need costs almost nothing: one intelligent model everyone can access, plus a prompt that works the same way every time.
Three years of building AI tools for educators keeps teaching me the same lesson. The tools that get used twice, then a hundred times, are the simple ones: upload a document, ask the question, get the result. If you lead a team, here is the fifteen-minute version. Get your people in a room and answer one question: where would saving time matter most? Start there. Not with the tool, with the time. Give a teacher back an hour of planning a day and you get a teacher who is not grading at the kitchen table every night. Those are the tools that win.
Anthropic Education Report: How educators use Claude by Anthropic. Anthropic read through about 74,000 anonymized conversations from higher-ed faculty. The top use was building curriculum, at 57 percent. I keep this handy for the meeting where someone asks "what are people even doing with this stuff." Now you can answer with data instead of a vibe.
Stop Letting the Professor Be the First Person to Find the Flaws by Doan Winkel. The idea: build a "red team" reviewer and a rubric reviewer into the workflow so the flaws get caught before the work is submitted, not after. He writes it for students, but swap "professor" for "the board" or "the parent" and it is a staff workflow. Catch your own problems while there is still time to fix them.
Anthropic and Teach For All launch global AI training for educators by Anthropic. Training headed to more than 100,000 teachers across 63 countries. The part worth noticing is not the size, it is the bet: the bottleneck is fluency and support, not access to the tool. That matches what I see in districts. The license is the easy part.
Here is the exact move I used this week to make podcast editing repeatable, and it works for any task you do more than once.
Step one: the next time you do a recurring task, narrate it. Keep a doc open and write down every decision as you make it, in plain language. "Cut the first 20 seconds of dead air. Run the audio cleanup. Strip filler words. Add captions. Export at these settings." You are not building anything yet, you are just refusing to let the steps stay invisible.
Step two: turn that narration into a numbered checklist with the inputs and the exact settings filled in. The test of a good checklist is that someone who is not you could follow it and get the same result. If a step says "clean it up," that is not a step yet. Make it specific: which tool, which setting, what "done" looks like.
Step three: name the failure points. Where does this usually go wrong? Write the fix next to the step. For my edit, the failure point is starting from a blank timeline and forgetting the intro, so the checklist now starts with "import the intro and outro first."
Step four: do it once more, following only the checklist, and fix anything the checklist missed. After that pass, you have a system. Mine became a skill my tools can run, but you do not need code for this to pay off. A clean checklist in a shared doc saves the same hour, every time, and it means the work does not live only in your head. That last part matters more than the time saved, because a system you can hand off is the thing that lets a one-person practice grow past one person.
That's it for this one. If something here landed, share it with someone who'd get something out of it. If it didn't, that's part of the deal too. You can find more of what I'm working on at evalveconsulting.com, or book a call if you want to talk through what you're dealing with.
Talk soon, Chris