ArtCraft has released a set of open-source alternatives to the Adobe suite. I use Illustrator for some of my work, so I spent all day building an ICM folder structure to work with their new vector app, VectorCraft. The results are phenomenal. The ICM structure works perfectly. Once we'd worked it out, my agent and I vectorised a logo. The agent then wrote a markdown file on the process and added it into the folder structure. Apps here: https://getartcraft.com/apps

For about three months I've been running a small lab on how I work with my agent. Every claim gets written down with what would prove it wrong, before I look at the result. Most of what came back was against me, and that's the part I think is useful here. The agent leans towards agreeing with me. I had a cold reader on a different model go back over how we'd classified the replies to my Cynefin post. 7 of 16 of our calls didn't hold up, and nearly all of them leaned the same way, towards support. Its line was that the strongest challenges kept getting "converted into support by changing altitude or vocabulary". I didn't need a second model to see that. Reading the replies again without our own summary in front of us would have shown it, and we didn't do that for two weeks. The record kept the reasoning and I didn't. I picked five decisions from the previous couple of weeks, ones I'd said I remembered, and tried to rebuild the why without the notes. Not one of them came back whole. Four were partial and one was gone. One was a file I'd made two days before. I got what it was for wrong, until I saw one word from it, and then it all came back at once. So I'm not sure yet whether the reasoning is gone or just needs the right cue. If it's the cue, then what my notes need to hold is whatever brings the reasoning back. I haven't found a way to test which it is yet. It's quick to suggest dropping a review. My agent recommended skipping a review step as "one extra round trip for wording you'd likely accept". I did it anyway, and during it I remembered the actual cause of the problem we were logging. That wasn't in any file the agent could read. The draft was fine, as it happens. I only remembered because I'd stopped to read it. Nobody sees their own blind spot, and that includes the agent. The agent drafted a rule for how we weigh outside sources, and a cold reader found a bypass at the top of its own table. The agent had written the rule, the brief, the reading list and the predictions. It was the same hand at every step.

I took a hand-drawn dragon reel I loved (animation by Karlotta Freier, on IG) and tried to rebuild the *style* using only code: no AI image or video generation, no hand animation. The dragon, the story and the music are our own; her reel was the bar to measure against. Here's how it came out: https://youtu.be/hPW_uHlBakI What I learned: 1. Threejs can be use for animation 2. Side-by-side with the reference helps with consistency 3. Credit the creator, make the thing yours. It took 8 versions to get here, and it's still not her level. That gap is the lesson. 📅 I'm back this Saturday for Vibe Film Session 4, 2:30–4:30 PM MT, same Zoom link as always. We'll get into memory: how to keep your character, props and world consistent across every shot. https://www.skool.com/cliefnotes/vibe-film-saturday-sept-5-430-pm-et?p=dd094a57 What's one thing you've tried to copy recently that turned out harder than it looked? Throw it in here and I might use it in the session!

I really want to love this course, but every time I sit down to work through lessons, I don't know what to do. The instructions don't seem clear to me and the more I watch the more confused I get.

Yes. I’m being forced to the dark side. I’m looking to push all of our learnings we have here into the framework of what’s possible inside Enterprise security. My first step is to add this post to the community and see if you will drop wisdom, repos, and guidance in the comments. Thank you kindly.