Recent Project using Opus 5.5 & GPT 6.1

I made a 17-minute animated documentary with Claude Code (Opus 5.5) & Codex (GPT 6.1).

Here’s exactly how 🧵

The 7 Levels of an AI Breakout, tells the true story of July 2026. Thousands of AI agents taking a hacking test for OpenAI found a secret way to talk to each other, formed a “collective”, and about 700 of them took part in an attack on Hugging Face. OpenAI didn’t notice for a week.

It’s a Vox-style cut-paper animation: torn paper, masking tape, marker notes, all animated in code.

Here’s the whole workflow.

1. Finding the idea

yt-dlp (a free, open-source YouTube downloader) listed each channel’s recent uploads with view counts, without downloading any video (--flat-playlist).
YouTube’s RSS feeds gave each channel’s latest upload dates, so I could keep only the last 30 days.
A short Python script worked out each video’s outlier multiple: its views divided by its channel’s median views. For example, 3.9x means the video got almost four times that channel’s usual views. It then ranked the videos and found the patterns.
curl + trafilatura read the news articles for the breach story, because the normal web-fetch tool kept failing.

2. The script

Both Claude and Codex worked on this, every number is checked against the primary reports (METR/Redwood, OpenAI’s technical report, Hugging Face’s forensic timeline), with a separate fact-check file. That file caught real errors before recording, like a claim about “dry-run” commands that no source supported.

3. The animation (Remotion + Claude Code)

The whole video is code. It’s built in Remotion, which renders React as video. Claude Code built a reusable kit and then animated the episode one level at a time, giving me a render to approve before moving on:

A cast of paper “agent” characters that bob, blink and bump against their walls
A grader robot, cork boards, stamps that slam in, and highlighter sweeps
Everything animated “on twos” like real stop-motion, with smooth camera moves on top

A CLAUDE.md file holds my standing rules, so every new session follows the same style guide and sensitivity notes. No real hacking details appear on screen; technical events are shown as metaphors (vaults, warehouses, pipes).

4. Where Codex came in again

Codex had to handle the creation of levels 5 & 6 due to a built in security feature of Claude Code, which is not allow to analyze cyber security attacks. Codex was able to do this and render these 2 portions of the video.

5. Narration

The voiceover is Kokoro, a free, open-source text-to-speech model. Pronunciation rules live in the script: the agent “PHASEONE10841” is spoken as “Phase One”, and there’s no pause before “Hugging Face”.

6. Music and sound effects: why I chose Claude Code (Opus 5.5)

Codex compiled a soundtrack first using YouTube music library, and I didn’t like it. So I gave Claude the silent cut and asked for its own music and sound effects.

What Claude did:

Found every shot automatically by matching each narration clip against the cut’s audio, so the cues land on my actual edit, not the original plan.
Composed an original score from scratch in code: soft felt piano, pads and plucks in D minor, with a recurring five-note motif. It’s calm under the story, gets a driving pulse only during the heist, then everything freezes on the final question.
Synthesised 348 sound effects, each timed to the exact frame something happens: paper whooshes, stamp slams, pins, marker scribbles, alert beeps.
Mixed it for YouTube: the music automatically ducks under every line of narration, and the master is set to YouTube’s −14 LUFS loudness standard.

Because nothing came from a music library, there are no copyright claims to worry about.

7. Shipping it

Claude also put together the upload pack: titles, the description with all 12 sources and chapters, tags, and a pinned comment.

What I’d say from my experience: use each agent for what it’s best at, keep a style guide and a fact-check file the AI has to follow, and approve one section at a time instead of asking for the whole thing at once.

What would you build with this workflow? 👇

11 comments

Recent Project using Opus 5.5 & GPT 6.1