🎬 This wasn't possible 6 months ago...

Happy Sunday folks!

​

Quick one this week — three things from the past seven days that are worth your time.

​

In this week’s email:

• 🎬 How I turned my 8-year-old’s daydream into a 15-second AI film (any sport, hobby, or scenario)
• 📚 We turned a school topic on Africa into a 10-minute podcast (with the infographic-loop trick I is very handy)
• 📊 Three new studies just changed what “AI for studying” actually means

​

​

🎬 I turned my 8-year-old's daydream into a 15-second AI film

​

Every kid has a daydream they replay in their head. Scoring the winning goal. Representing their country. Finding the Hogwarts letter on the kitchen table.

​

As regular readers of this newsletter will know, my 8 year old son's daydream is football. When he is not out the back garden playing football with his friends, he is playing by himself narrating his own commentary under his breath: “He drives forward, plays it wide, ball comes back into the box, Mateo shoots… GOAAAAALLL…”

​

This week I tried to put him inside the version of the match that lives in his head. A 15-second animated short where he scores the winning goal in a packed stadium, then wakes up in his bedroom and realises it was a daydream.

​

​

​

The workflow:

1. One photo of Mateo → ChatGPT Images 2.0 generated a 10-frame storyboard
2. Storyboard → Seedance 2.0 turned it into a cinematic video
3. Total time: 20 minutes. Total credits spent: about €4.50.

​

One thing that surprised me: I asked for a poster of his favourite player to be on his bedroom wall (Saka) in the storyboard but the Higgsfield’s eligibility check blocked the video generation immediately. Real public-figure names by name on signs or posters are not allowed. Genuinely useful guardrail. I regenerated with generic posters and it passed.

​

When I played the result for Mateo, he genuinely loved it. On the second watch, with a big smile on his face, he threw his arms in the air to match the celebration shot frame-for-frame, then asked me to play it again. He came close to re-creating the moment in his u8 5-a-side match this morning!

​

​

​Read the full guide with exact prompt →

​

This same workflow works for any daydream your kid has on repeat. Gymnastics medal. NBA buzzer beater. Skateboard kickflip. Pop star on a stadium tour. Same 15-frame structure: setup → action peak → match-cut → soft bedroom landing.

📚 We turned a school topic on Africa into a 10-minute podcast

​

Mateo's class was learning about Africa last week. The school worksheet covered the basics. He wanted to go deeper.

​

So I built him a NotebookLM for the topic. Three YouTube videos. Two vetted reference websites. A handful of kid-friendly geography PDFs. And the secret weapon: the infographic from last week's newsletter I generated using ChatGPT Images 2.0, fed BACK into NotebookLM as another source.

​

The Audio Overview the AI hosts produced explicitly referenced the visual summary. The infographic stayed on the fridge for a week. Mateo came back from the 8-minute podcast naming three African capitals he had not known that morning.

​

The infographic-as-source loop is the workflow trick I would not have figured out without testing both tools the same week.

​

I wrote up the full setup, the safety guardrails (you control the inputs entirely), age recommendations by feature, and the Irish Famine story we did at New Year's:

​

​Read the full NotebookLM for students guide →​

​

Prefer a quick-reference version? The NotebookLM tool review condenses our 4 months of family testing into one scannable page — age recommendations by feature, pricing, safety, and where it falls down.

📊 Three new studies just changed what "AI for studying" actually means

Ethan Mollick (Wharton) shared a study on X this week with a finding that captured something I have been seeing at home for months:

​

Three new randomised trials in 2025-2026 all point the same way:

• Harvard, Scientific Reports (June 2025). AI tutor beat in-class active learning by 0.73 to 1.3 standard deviations. Students learned more in 49 minutes than in the standard 60-minute class.
• UK secondary RCT (165 students, ages 13-15). Supervised AI tutor outperformed online human tutors at follow-up problem-solving, 66.2% vs 60.7%.
• AI assisting human tutors (math). Students were 4 percentage points more likely to progress when their tutor used AI assistance. The biggest lift was for less-experienced tutors. This is the at-home parallel: you are the tutor, AI is your assistant.

​

​

​Read the full breakdown →​

​

That's it for this week. Have a great rest of your weekend!

​

Niall (AI Dad Lab)

​

P.S. If you try the daydream animation with your kid this weekend, reply and let me know what they daydreamed about and how they reacted.

​