FREE MONDAY

FREEMONDAY

It started with an email.

Months earlier I'd had a long conversation with an AI about AI taking our jobs. That chat ended up being the script for a video I never got to make: The Identity Paradox. It sat in my inbox as a 52-page PDF.

The idea at its core: «The real problem of AI might not be jobs. It might be identity.» For decades people have built who they are around what they do: I'm a lawyer, I'm a developer, I'm a waiter. So when the machines do more of the work… «Who are you when you don't have to work anymore?»

On 22 October, at 3:51 in the morning, I went looking for that email. «I want to make a song about this.»

The idea.

The angle came from that question: AI won't take your job, it'll take the identity you put into it. Anticipatory anxiety, suffering today for a future that hasn't happened yet. And an uncomfortable question: if you hate your 9 to 5, why are you so scared of losing it? For the chorus: who are you on a free Monday?

The frame came in two words: AGI TIMES. A made-up news show, with scary headlines read with a straight face. More to make you think than to criticise.

The video, before a single photo existed: «a girl walking through the City of London among executives. She starts as one of them, takes off the uniform little by little, and people join her as she walks.»

The lyrics, in one night.

One hour and eight minutes, talking with Claude and mostly cutting. The first draft was «weak», so I wrote the thesis myself: «It's not the job. It's your identity what AI will take. Your worth is not what you do.»

The verses became headlines. The choir went from 40 shouts to 4, one word each: WHO? · YOU! · STILL HERE! The chorus is a question the choir answers, and the last line answers the whole song: «Get ready to find out who you are
on a free Monday.»

A silly trick: Suno read «AGI» as a word, so it's written the way it sounds, «Ay Gee Eye».

Six takes and the street test.

«I don't like that the beginning sounds like every other song»: out went the stock synth intro, in came the a cappella hook at second 0 and a whisper: «Breaking news. You got Mondays off. Forever.»

Six takes. Take 6 had the catchier hook; take 4 had clearer lyrics and a chorus that hit harder. After so many listens they all sounded the same, so a new rule was born: listen to the finalists walking down the street, away from the screen. Take 4 won.

Take 6 had a slow bridge I loved («it sounds human»), but it was in another key. So we described it to Suno and rebuilt take 4's bridge with an Extend.

The look: Kate Moss.

She's the girl from my other videos, described with the same words in every prompt: long leather coat, sunglasses that hide her eyes (eyes are what AI gets worst), a tight bun.

What she wears under the coat comes from a photo of Kate Moss: a light blue striped shirt with a white collar and a black leather skirt (Pinterest · Elle UK). I asked for the skirt shorter, and high boots.

Other people's photos never go into the machine: the clothes are described in words, and every test is shot on the same street with the same light, so you only compare the clothes. Ten looks before choosing. For dancing, nothing that flies.

The real City.

We tried a red double-decker bus. I dropped it: «no bus, let's go to locations that don't look like AI; a real place in the City, better.»

A place you recognise, with flat, still ground, is what deforms the least when it moves.

The photos: Midjourney.

Every shot starts as a photo. Claude writes the prompt, I run it in Midjourney, pick one, and it goes into its row of the board. Below, every photo we downloaded: the lit ones became clips.

The photo rules the clip. If people are already looking at the camera, you can't ask them to turn their heads. And the Editor redoes one photo on top of another keeping the exact framing: the crowd, the empty street and the little girl, ready to cut on a beat.

The opening: only she moves.

Each clip is a photo + the exact piece of the song + one line of what happens. The first clip came back with her mouth shut: the prompt said «the solo voice sings», and the machine thought someone else was singing. One word fixed it: «she sings».

Then the idea «for more impact»: the whole City frozen mid-step, coffee leaping out of a cup, papers floating. Only she sings. On the kick drum, everything starts moving again.

The news, in the glass.

Night falls at once and a glass screen turns and lands in front of her. The screen is in the photo; the clip moves it; the headlines are drawn afterwards inside the glass, following its edges, each one on its word.

The first try was «what a circus»: tiny face, glass across it. Redone in a medium shot, with one rule: she doesn't speak until the glass is in front of her. And one gesture on time beats six late.

The chorus: people join in.

«10 seconds walking and singing, raising the arm on WHO… from second 5, more people from the street walk behind.» That's the idea of the video.

Then the same, but serious and silent, only shouting «WHO?». The machine moves mouths with whatever voice it hears, so it got its own audio: the music, with the voice only on the two «WHO?». In the video, the real song plays.

The office.

My idea: robots walk in and take people's seats; the people leave with their cardboard boxes, and in the street they throw them in the air and dance with her. The lift is seen through a security camera: it's what looks least like AI.

A mistake that teaches: the first office photo came out in the street. Her reference photos had the street behind them, and Midjourney copied the background.

London Bridge: what do you do?

The bridge the City's office workers cross every morning. She stops, they surround her and shout at her together: WHAT DO YOU DO? On screen, everyone's job title gets crossed out and turns into «WHO AM I?».

Then a glitch: the office workers become kids with balloons and back. Both clips were made with the same prompt and the same audio, so everyone lands in the same spot of the frame.

The slow part: her silhouette, in red.

The bridge stays slow on purpose («it's powerful because it's emotional; the shots are long so nothing distracts»). On «When nobody's counting», the crowd disappears and she's in the street where she grew up. It shouldn't feel like a cut: it should feel like the world changing around her.

So before making the clip, we traced her silhouette from the shot before and laid it over every candidate photo. The one that matched won, and it was moved and scaled until she stood in exactly the same spot, at exactly the same size.

Even the little girl who comes next was chosen by her face: the computer compared four girls with the one already in the video and picked the closest.

The drop: the crowd that jumps.

Arms crossed: «Free Monday». Behind her, office workers and robots get into the song until they all shout at the camera. My direction: «no random dancing, watch the beat… like when a song comes on.»

In the first clip the crowd was packed tight and barely jumped. The second was made from a photo where they're already in the air, and the clip kept that energy. If you want them to jump, the photo has to catch them jumping.

The ending.

The news glass comes back, turns and lands in front of her: «Today's headline: you're still here.» On London Bridge, with the office workers standing still like statues, she says the last line: «Get ready to find out who you are… on a free Monday.»

Then one second of black, to let it breathe, and the front page of The AGI Times.

The edit: my own studio.

The video is put together in my own dashboard, connected to Claude: every piece of the song with its second, its lyrics, its prompts, its photo and its clip, and a bar under each clip showing the part that made the cut. Before paying for a clip we watch the whole video as a mock-up, and the final is rendered from the originals, every singing mouth on its exact second.

CapCut is only for the small details: a cut a few frames earlier, a zoom, a split screen. When I save, Claude reads the CapCut project and the final picks it up.

Soon, for anyone who wants it. The studio on your own Mac (it needs a setup), or maybe online, with a small subscription. Join the list →

The graphics, at the end.

Graphics go last, on top of the edit, with a style made for this song: lyrics in little boxes, the choir's shouts huge and red, job titles that become questions, and the AGI TIMES only inside the glass.

Every option is tried for free and kept, so we can always go back: 34 options before choosing.