Behind the behind, episode 2: The AI checks its own art The second episode of the mini-series about what I learned building Behind the paper plane, made by the same ICM workspace. 4 minutes 45 seconds. Watch it on LinkedIn: https://lnkd.in/p/eJwQzvzd Episode 1, Thirteen messages: https://youtu.be/C2POk9nGRxE Behind the paper plane on LinkedIn: https://lnkd.in/p/eaYD2G-x Why the LinkedIn links: I hope you will like the posts and share them, to help me spread my name out there. You never know who sees it. THE IDEA In the opening of Behind the paper plane, a paper plane was hidden in a chrome pool. A check found it too loud: striped, pulling the eye. So it was turned down. Then a fresh reviewer looked at the pictures and could not find it at all. That is the whole problem in one picture. The helper that drew a scene knows what it meant to draw, so it sees it. A viewer only sees what is there. So in my system, the builder never judges its own art. HOW THE REVIEW WORKS - Every finished part goes to a new helper that has never seen the build. No chat history, no code in front of it. - It gets the pictures: twelve stills to a sheet, each stamped with its time. One for every moment in the plan, and more where one style turns into the next (at a quarter, a half and three quarters of the change). For one part, more than 300 pictures. - It gets a list of what to look for: can a person read it, is anything cut at the edge, is each style true to itself, is the hidden plane there without stealing focus. - It writes back a short list of what a viewer would notice. The fixes go in, and a new reviewer looks again, until the part passes. Tests still run, and they matter. But a test can say a line is there. It cannot say if a person can read it. WHAT THE REVIEWS CAUGHT - The opening's first list had 8 things on it. The plane nobody could find was one. - The camera cut words off at the edge. Again in a close-up of The paper plane's script: a slow pan was tried, and the reviewer still saw "sketchboo" and "and threw" cut off. Now long lines wrap, the way VS Code wraps them.

Behind the behind, episode 1: Thirteen messages In my post about Behind the paper plane I promised the "behind" Behind the paper plane: the things I learned while building it. Here is the first one. It became a mini-series, made by the same ICM workspace, and this is episode 1. 4 minutes 45 seconds. Watch it on LinkedIn: https://lnkd.in/p/eYnb2vUU Behind the paper plane on LinkedIn: https://lnkd.in/p/eaYD2G-x Why the LinkedIn links: I hope you will like the posts and share them, to help me spread my name out there. You never know who sees it. THE IDEA After Behind the paper plane I wrote "Before setting it loose, I wrote Claude 13 messages." This episode checks that claim against the record, and tells what those messages were. It holds up. 13 messages about the film, 884 words in all. The finished film's narration is 1,314 words, so I wrote fewer words than the narrator speaks. One message, the picture in my head (the blueprint, the glass, the liquid chrome, the rolodex, the hidden paper planes), was 365 of them. And there were 2 corrections, not 1: - Before the film was built: the script said I make videos for clients. I do not. The record had it wrong, and one line changed. - After it was done: I watched it and thought a time on screen was missing a digit ("7:20" should be "17:20"). Looking closer, it was not one digit. The line mixed clock times from two different days with a moment from inside the film. The checks had measured every frame and found nothing broken. It still made no sense to a person. My decision: keep only what the voice says. Seven marks became three. The point of the episode: 13 messages sounds like very little. It was not. The work was done before the first message, in the folders, the rules and the record. The messages only had to point the way. HOW I KEEP MY WORDS ON SCREEN HONEST The first draft showed every word of my messages on screen. I said no: reading every word costs the viewer time for little value. So my messages are shown as bubbles sized by their length, the text blurred, with only a few short fragments readable. Those fragments are exactly what I typed, typos kept ("serius", "tribut"). Claude's answers are shown only as frosted blocks, never as words. Nothing is put in anyone's mouth.

Most of us are out here grinding agents, wiring tools, fighting for clean data sources. Meanwhile this giant, boring, open list has been sitting on GitHub the whole time. https://github.com/public-apis/public-apis 492k stars. MIT license. A community-curated dump of free public APIs, sorted by category: finance, weather, geocoding, government data, machine learning, health, transportation, and dozens more. Each entry is a plain table row. Name, description, auth type, HTTPS, CORS. No signup wall. No "request access." Just the raw list you can fork, search, or parse. 📁 Own the stack. Drop the markdown or the JSON into your folder system. Your memory, your rules. Audit it, filter it, keep the ones that stay free. 🔍 Inspectable over magical. You see the auth, the HTTPS status, and the CORS flag before you wire anything into a harness. No black-box surprises. 💰 Value on a budget. Exchange rates, weather, public datasets, geocoding, open ML endpoints. The stuff agents actually need for real work, not demo theater. 🔁 Transparent and repeatable. When an API dies or changes terms, the community updates the list. You're not locked into one company's roadmap. Closed APIs and SaaS platforms want you renting everything forever. That's the same leak I used to leave on client CMS installs: nothing persists except what lives in someone else's admin. This list flips that. Stop duct-taping proprietary endpoints you can't inspect. Start from a public, auditable catalog and only pay, or self-host, when the free one isn't enough. Bookmark the repo. Skim the categories that match what you're building. Then tell the rest of us what you wired up. That's the part I read the comments for.

One the right is my autobots skill that run headless sessions creating 6 episode of the "behind" behind the paper plane. 5 minut videos where I only have approved the scripts and the "autobots, roll out". On the left is my dashboard to see where they are at and what is completed. Behind is a mini series with issues and solutions in creating Behind The Paper Plane. This week: The Paper Plane - 9 minutes https://lnkd.in/p/ecYUXUyg Behind The Paper Plane - 12 minutes https://lnkd.in/p/eaYD2G-x Behind series - 7 episodes each 5 minutes. Almost an hour of video in a week!

Hi everyone! I had to share my latest project. I am so proud of it and hope you enjoy it! Thanks to @Jake Van Clief and his crew for creating the community. I feel like nothing is impossible to create. I had Claude code write the next part: I gave Claude one sentence. Two days later it was a 9.5 minute animated film. My idea was one line: "a child's room where different art styles are used." What came out is "The paper plane": a 9 minute 31 second animated film with 25 art styles, 8 narrator voices and an original soundtrack. Not one image or video generator was used. Every frame is drawn in code. Here is how it came together, and why ICM is the reason it worked. THE IDEA I had a folder of 84 reference art styles (sumi-e ink, claymation, film noir, Win95, synthwave and so on) and one picture in my head: a child's room where every object is a door into another style. The camera dives into the TV and you are inside a video game. The poster on the wall turns out to be a whole world. One style draws an arcade machine, and inside it the next style is already playing. THE BRAINSTORM Claude ran a structured brainstorm with me (the superpowers brainstorming skill): one question at a time, and a quick mockup of the room so I could see the camera route. Out of that came four "hand-off moves" that join one style to the next: - Dive in: through a screen into its world. - Drawn doorway: my arcade idea. - Pull back out: the world was the poster on the wall all along. - Carried thread: something crosses the seam from one style to the next. Then Claude pitched me three stories: 1. The paper plane: a plane that flies from style to style. 2. The bedtime question: a child asks something, and the room answers. 3. How pictures learned to move: from shadow puppets to 3D games. I loved all three and could not pick one. So I asked Claude to weave them into one story, and to drop the 2 to 3 minute limit we had planned, because a great story needs time to play out.