Hey Chief Notes fam! 👋 I'm David and I'm stoked to be here. 🙋 A little about me: I used to be a head chef and restaurant owner. But life happens, and I had to pivot. So I came back to an old friend: the computer. And with the way tech moves today, it really makes you wonder what you're missing with each passing day. 🎯 My current goal: Truly understand how AI works under the hood — mastering its configurations, architecture, and the deep "how and why" behind it. 💪 What I'm currently building/working on: I'm juggling two main projects right now: 1. The dream project: Launching a video game studio built around giving power back to the players. 2. The tech & business project: Leveling up my AI skills to help under-capitalized businesses access turnkey IT solutions — network maintenance, hardware/software setup, cloud management, and practical automation for professionals. 🤔 My biggest struggle or question right now: Getting AI to truly align with my quality standards. We definitely don't always share the same definitions or expectations just yet! Let's get it! 🚀

You've probably done this. The AI helps you write something, a proposal or a price for a job, and before you send it you ask it to check for mistakes. It reads it back, maybe fixes a comma, and tells you it's solid. Then, to be safe, you switch the same session to a more powerful AI and ask again. It agrees too. Two yeses might feel like a second opinion, but it isn't. Here's how it happens in an ICM project. Your client folder holds client-brief.md, and the deadline written in it is the 21st. Early in a session you tell the AI the deadline is the 14th. You're working from memory, and you're wrong. The AI believes you and writes proposal.md with the 14th in it. Then you ask it to check the proposal. It finds nothing wrong, because it isn't comparing the proposal with the brief. It's comparing the proposal with what you told it, and you told it the 14th. It's checking its work against itself. The more powerful AI does no better. It reads the same session, so it has heard the 14th too. Your two asks still got one opinion. I did exactly this when I switched to a more powerful AI partway through, hoping for fresh eyes, and got the same conclusions back. In a reply on an earlier post here, @Gabriel Azoulay had already named it. The same AI asked three times is, in his words, "one reader in three hats", one opinion wearing three different hats. Some of the fix below comes from how he guards against it. The fix is a reader that starts fresh. A fresh session starts with no memory of the last one, so it has never heard you say the 14th. Give it the two files, proposal.md and client-brief.md, and one job, check every date, name and figure in the proposal against the brief and list anything that doesn't match. The proposal says the 14th. The brief says the 21st. It has nothing else to go on, so it puts the two side by side and the mismatch shows up. Telling the same session to forget what you said doesn't work. It has still heard the 14th.

I only started making AI videos around April. I was playing with Higgsfield, figuring things out one video at a time, with no real idea where it would lead. Then I found this community, and honestly, as it grew, I grew with it. Jake gave me my first real shot with Vibe Film (host on community every Saturday), and then brought me on as one of the coaches in the Lyceum. I've learned so much from teaching those classes, and I've met so many great people here and in class along the way. Now here's my second win. This October 31st, my sister Hannah and I are the AI Visual Directors for KimLoi's Hello America Concert at the Colosseum at Caesars Palace, one of the most iconic stages in Las Vegas. Celine Dion, Elton John, and Adele have all played that room. It's also the same stage where BLACKPINK's LISA starts her residency a couple of weeks later. It's about 4,300 seats, mostly Vietnamese-American families, plus a worldwide livestream, all in front of a 110 × 43 ft LED wall. One of the singers, Đức Phúc, won The Voice Vietnam and Intervision 2025. One of his song alone has 170M views. None of this happens without this community, and a special thank you to @Jake Van Clief @Sonija Quinn @Matthew Creamer and team for the endless supports 🙏. It's even more meaningful that the 1st official check I got for Reelix Studio LLC (my new startup) is from this very community. Here's the honest situation. The show grew from 1 hour of visuals to 3.5 hours, with about 28 songs, and most of them are slower Vietnamese ballads. Right now it's just the two of us, and we have under three weeks. So we're building the visual loop system first, with clean loops that carry most of the songs, and we'll circle back for custom moments if time allows. Where I need help: 🫰Funding. Sponsorships start at $1K and go straight into generation credits and editors. You get your logo on the LED wall, an MC thank-you and an outro credit, and the higher tiers add tickets, a sponsor video and an on-stage moment. The proposal and reel are below. The fund will goes straight to the stage visual production.

I've had a couple of my content creator clients reach out using ICM as our default staging architecture and process. Does anybody have an existing working/functional pipeline and which tools are you using? I'd love to know what others are doing before I try to build this myself. Mind you we don't want to be building video content out of nothing. We want to be editing video so that's the clarification. I see a lot of posts and a lot of content around building AI videos and I really don't want to do that necessarily in this use case. I want to manipulate our existing video files. We started looking at connecting to CapCut via a CapCut extension/MCP. That's not really supported by CapCut but is community-supported. Anyway just wondering whatever others are doing before I go headfirst into trying to solve this problem myself. I appreciate any and all feedback, thoughts, and considerations.

This is the website I created for them — pikolhaus.com. Small start, but officially my first real client project. More to come.