Stepped away from “work” to play a bit with video. I am not a video person but wanted to challenge myself. Took @Danney Tee’s awesome notes and write-up and challenged myself to create a trailer video. I wanted to tackle character and product consistency, play with movement, and just stay curious, have fun, and learn. Here’s a teaser - will update with the video and lessons learned when it’s done. Danney’s original post and inspiration: https://www.skool.com/cliefnotes/i-turned-a-reference-reel-into-an-ai-headphone-ad?p=07166dea

Hayley Cullen-Xiang asked a few days ago whether checking Claude's work with ChatGPT or Codex beats just opening a fresh Claude. Most of us in that thread said yes from experience (me too). Nobody had tested it. So I'm running the test. The full design is here: https://claude.ai/artifact/8awrFEqk4d1pnd8qZGh5CK The question has two parts. Does a second AI catch more real errors? And does the work actually come out better after the fixes, or just different? Most comparisons have a trap in them. They put one Claude review against Claude plus ChatGPT, and two reviews beat one whatever the models are. So this one compares two Claude reviews against Claude plus another company. Same number of looks, different eyes. What's in it: 12 pieces of realistic work for a made-up renovation company (code, an SOP, a client proposal, cost estimates, schedules) with 96 errors planted on purpose. Plus 3 pieces with no errors at all, to see who invents problems. ChatGPT, Gemini, Perplexity, Claude Desktop and Claude Code each review every piece 3 times (270 reviews). Claude then fixes each piece from different mixes of those reviews. Every review it gets has 2 fake findings slipped in, to see if it takes bad advice. Sometimes it's told the review came from "a different company's AI" when it didn't. People who haven't seen any of this rate the fixed versions blind. Grading is blind too. I grade, and so does Jev (TypeSafe), which isn't one of the AIs being tested. The questions, the scoring and the analysis code were locked in before a single result exists. If it comes out against what I expect, I still post it. One thing already happened. I ran the design itself past fresh AI reviewers, 3 rounds against 22 written checks. 8 passed, then 13, then 13. Every round found something new and smaller, which is exactly what Shaun Lawton described in that thread. I stopped at 3 rounds (thanks @Heith Murray ).

I’ve been going between Codex and Claude for checking each other’s work if it’s something really important, because they ALWAYS pick up on issues and intuitively it feels more sound than just using a fresh instance of Claude alone.. It also seems like it makes sense, since the companies are competitors, so I hope they’re more incentivised to poo-poo on each other (despite Claude and Codex/ChatGPT being so delightfully polite about it). In one case I used Fable to start, and some less sophisticated ChatGPT model to check!! I took it back to Fable who was basically like “oh it’s right, I missed those major things, good catch by Codex”. BUT: Maybe I’m just deluded, I’ve never spent the time to do a proper controlled trial, with cloned files/folders and comparing fresh AI from the same thing (e.g. Claude and fresh Claude), vs Claude with Codex. And if I’m honest, I can’t see myself sitting down to do the trial either, I’ve got too much backlog of more urgent things to do. Has anyone already looked into this?

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

Hey Clief Notes fam! 👋 I'm Tyler, and I'm stoked to be here. 💪 What I'm currently building/working on: I'm excited to learn & master structure/infrastructure, AI & agents. Let's get it! 🚀