🦥 I drew a sloth on our cable car video over Medellín. Then it came to life.

Happy Sunday folks!

​

We're home from Colombia. The flight was rough, I was somewhere past hour 30 without sleep by the time we landed, and when I finally crawled into bed I slept 15 hours straight 💤. A new personal record. Somewhere in the jetlag fog I also made the best thing I've made all month.

​

In this week's email:

​

• 🦥 A magic marker that turns real videos into cartoons (and the prompt to do it yourself)
• 🏫 Claude is now completely free for US teachers
• 📝 The one question that tells you if AI homework is cheating

​

🦥 A magic marker turned our cable car video into a cartoon

​

A video went viral on X this week: a hand draws a squiggle with a black marker over real street footage, and the squiggle comes to life as a cartoon. A zebra crossing turns into an actual zebra. The green man walks out of the traffic light.

​

The creator didn't share his prompts, so I reverse-engineered them frame by frame and rebuilt his opening scene. One prompt, one generation (p.s. the gifs don't do them justice):

​

​

Then the real fun. I tried it on our own footage, the metrocable video I shot gliding over Medellín on our last day in Colombia. That is a hand-drawn sloth tightrope-walking the actual cable from our cabin, slipping, and hanging on by one arm while the real city slides past below:

​

​

I wrote the whole thing up as a step-by-step you can do with your kids this week:

​

👉 The magic marker trick: full guide with the exact prompt​

​

🏫 Claude is now free for teachers

​

Anthropic made Claude free for verified US K-12 teachers until June 30, 2027. The full paid version, with training switched off by default.

​

I wrote up the parent’s take this week:

​Claude for Teachers is free now.

​

The short version: this is the good kind of AI in schools:

• Lesson planning,
• Differentiated worksheets
• Parent emails,

​

and the AI stays in the teacher’s hands, not your child’s. That distinction matters more than any feature.

​

Quite a few of you reading this are teachers. If that’s you: verify, claim it, and reclaim your Sunday evenings.

​

📝 Is AI homework actually cheating?

New post I’d been chewing on all through the Colombia trip:

​Is AI homework cheating? How to actually tell.

​

The test is one question: did the AI give the answer, or make your child produce it?

​

You might remember the study from the Norway email a few weeks back, where kids with unrestricted AI scored 17% worse on the real test. The half I didn’t share is the trap: during practice, that same group had scored 48% BETTER.

​

Brilliant in practice, worse on the day. That gap is what it looks like when the AI quietly does the practising for them.

​

Our house rule: AI explains, AI questions, AI never gives the final answer. It works on Mateo’s maths practice at 9 and it works on teenagers’ essays.

​

That’s it for this week. Off to go for an afternoon snooze. See you next Sunday.

​

Niall

​

P.S. I cut the sloth and the zebra together into one video for Instagram, TikTok and YouTube Shorts. A like or a follow genuinely helps more than you’d think.

​