Today's email is different.

​
When I saw Marco Van Der Vlag was vibe coding an After Effects rival and it could render 1000s of layers in real time in the browser – I knew it was a story worth investigating. And it was. As with every podcast, I came away with one big shift that changed the way I think (more on that later). Watch it, and I´m sure you´d have your own takeaway.

​
​
Before I tell you mine, we need to go back a few months. I am ashamed to say I recorded a podcast 6 months ago, and I only published it last week. Why?
​
Two reasons:
​
One, the audio fucked up; two, I hate editing. You could call that two tiny pieces of friction. Retiming the audio, a couple of hours extra work = a 6-month delay. Two hours became 6 months of anxiety. The actions we don't take become tiny amounts of mental debt, unclosed loops that weigh on the mind and cause anxiety.
​
For creatives, anxiety can be the gap between idea and execution. I tell this story to highlight: The smallest amount of friction derails most humans.
​
This isn´t just a Luke problem.
​
My friend used to be a personal trainer; once a week, he did a group class. On rainy days, attendance dropped 90%. This wasn't a city centre gym, almost no one went by foot. So that 1 min of getting wet between your front door and car, that short dash from the car to the gym entrance derailed 90% of people.
​
Don't be most humans. The Cheat Code: Close the idea → action gap quickly, push through minimal friction.

When I finally edited the podcast, the theme was (thankfully) more relevant than ever. AI can help with production, but without human input; an initial idea and a final human pass – it's just AI slop.
​
Too many people are aiming for for 100% AI workflows, but it just doesn't work; it's not interesting if the idea came from AI or it's not refined enough if no one edited it through their human experience. The first and final 10% of your process must be human.
​

​
So, as all good humans do, I immediately put this learning into action. I built a content creation system to write this email. I curated the content over a week (10% human), then I asked AI to connect the dots, create a theme, and draft an email. I´d edit it (final human 10%). It was glorious, friction removed, email written in 1 hr, based on my own POV.

​
​
It worked well until the second podcast episode, and a critical new perspective. Don´t underestimate this.
​
​Cheat Code 2: If AI isn´t making you smarter, it's making you dumber.
​
I realised my system was making me dumber, I systematised the writing to save time, but that is the part where ideas are formed. I wasn't living by my values, I wasn't doing the hard thing, I wasn't doing the thinking. The interesting part.
​
When everyone else is trying “to be someone” on social media, its more important “to be someone who does interesting things” - not a crowd pleaser, not a content machine, but someone who can make sense of the world and hold a POV.
​
​Cheat Code 3: Don't outsource your thinking to AI. Stay aligned with your values (or risk burnout).

My personal filter, “Do the hard thing”. If everyone else is doing AI research and AI writing, with an automated workflow, do it the hard way. Write and think for yourself, no AI pass, no AI improvements.
​
That's why today's email is different. It's proper human bad writing.

One thing I saw: Friend of the podcast Erin Sarofsky built the Skrimps using an AI pipeline, now it's got popular, she physically made the puppets - use AI to test, filter for attention, then do the hard thing, make it physical.
​
​One thing I heard: Codex is better than Claude now.
​
​What I´m doing: Learning to work with agentic AI Claude (and now Codex), rebuilding a new content workflow. 70% human, 30% ai collation.
​
​What I´m thinking: How would you build an AI-first fashion company? Mock up and test clothes with AI quickly, then make real products people voted for.