Issue 006

Sage Mode
Not slower. Stiller. And TGIF y’all.
WHAT’S INSIDE
  • TOPIC: If AI can write the code, what’s left to learn? A former computer science teacher on the one rule he’d give students now.

  • PRACTICE: Keep your date with your work, even when the answer refuses to show up.

  • SIGNALS: Fresh perspectives from around the industry.

Topic

Tony Morinello left teaching in 2018, after seven years and seven courses of his own design. Now, as Sageworx builds an academy, he’s thinking about what those courses missed and what he’d do if he taught them again.

Paper-collage illustration of a handheld game console split open beside its intact shell, with a hand working on the circuit board with a precision screwdriver

If You Can’t Explain It, It Isn’t Yours

AI can do the typing. The understanding still has to be yours.

For seven years I taught teenagers to write code. They built games, and the hardest thing I asked of them was to sit with a broken one until it made sense.

I ask the same thing of peers, who have a site an AI built for their business, with their name on it. Here at Sageworx, we’re all teaching one another, and next we’re building an academy to share our collective knowledge.

I left in 2018 to build software, when this year’s high school freshmen were in kindergarten. ChatGPT came out when they were in fifth grade. Upper administration wore me out, but the kids never did.

I’m not going back, but I think about what I’d do differently, especially now that AI writes most of my code, more of it every year, and my job is mostly reviewing what it wrote.

The surprise, eight years in, is how little of the work was ever syntax. Every real project tracks its changes in version control and runs on a database. None of it counts until it’s published. The seven courses I wrote spent most of their time on syntax, and most of my students never touched the rest.

They were great at emailing each other code, though. When I asked one of them to walk me through “their” program, they couldn’t, so we rebuilt it together, line by line, until they could. AI just makes the copying instant. Today, I’d have one rule: if you can’t explain a line to me, it isn’t yours yet.

A fifteen-year-old would have a fair question: why learn loops when a chatbot writes them in a second? Because you can’t fix code you don’t understand, and you shouldn’t put your name on it. The basics would still come first, with no AI: variable types, conditionals, loops, functions. Reading a loop isn’t the same as writing one, and I haven’t worked out how I’d check that a student can do it on their own.

What excites me most is teaching how software gets built now. Every student would make a portfolio website about themselves, and everything they built in class would live there, games included. It would sit in version control from the first day, the way real projects do, and it would get published. AI could help build it. But it’s a site with their name on it, so every line has to be one they can explain.

Two things worry me -- the kid who turns in a working project without grasping any of it, and the kid who can’t use these tools at home. Then there’s which ones to pick, and how to pay for a whole class using them all period.

The hardest thing I’d ask of them now is what I always asked. Sit with it until it makes sense, even when the code shows up finished and looking right.

If you can’t explain it, it isn’t yours yet.

Tony Morinello
This issue's contributor
Tony Morinello
I taught high school computer science for seven years, writing every course myself, then moved into software full time in 2018. Today I build software for clients and for fun. AI tools like Claude, Cursor and Codex do a lot of the typing; I keep the judgment calls. Find out more at tonyt908.com

Practice

Focus on what you’re avoiding.

A project can be close to finished and still have one part you keep working around. In The War of Art, Steven Pressfield calls the pull away from the work “resistance.” Here’s a way to use his approach:

  1. Identify the resistance. Name the part you keep avoiding. It might be a paragraph that won’t say what you mean, or a piece of the project you don’t understand well enough yet. Start there.

  2. Be a professional. Show up at the time you said you would and work on it, including on days when doing something else might feel better. The habit gives you another chance at the problem.

  3. Lean on the muses. Keep going even when you can’t see the solution. You can’t predict when an idea will finally take shape.

What’s one thing you’d teach a beginner in your field before handing them an AI tool?

Now you’re leaving the quiet part.

Signals

Fresh perspectives from around the industry

Two links worth arguing about, from people who’d know.

01 / AI + DESIGN FAST COMPANY →
AI-pilled bosses and clients are turning designers into ‘pixel pushers’ Fast Company illustration

KELLI UPSHUR SAYS

“This encompasses what I’ve been saying for a while now. Sometimes an AI-generated design ≠ cheaper. Sometimes it takes me longer to untangle and fix.”

READ THE PIECE →

Pass this signal along →

Quick poll

When AI-generated design needs untangling, who pays for it?

The client
The designer
Nobody’s tracking it

Tap one to send your vote. It opens a quick email to us.

02 / CINEMA LINKEDIN →
The 94th Letter, Claudia’s love letter to cinema Still from The 94th Letter: a woman in a dark coat runs through a circus lot holding a letter

CLAUDIA LALAU SAYS

“I wanted it to feel less like a history lesson and more like running through the many lives cinema has let us live; carrying the same love letter through every era, then delivering it at the Eden Théâtre in La Ciotat (the birth of Cinema) as a simple thank you.”

WATCH THE 94TH LETTER →

Pass this signal along →

That’s Sage Mode for this round.

Back soon. Until then, sit with something unfinished.

Did someone forward this to you?

Get the next one here →

One idea at a time.

From a field of 100+ senior specialists. No pitches.