My background is working in quality assurance in medical device companies, and I've always worked with software teams. I remember always walking past the software department and seeing their big black screens with all this code. It gave me a little bit of anxiety, but it also looked so cool because they looked like these hackers, just like in the movies. They were all sitting there so nonchalantly, but they're literal geniuses. One day I was sitting in a meeting with one of my colleagues, and I had to deal with an Excel spreadsheet that had like a thousand lines of information in it that I had to parse. I was doing this manually because the data hadn't been entered uniformly. One of our engineers was like, "Give me 30 minutes. I'll write a script." I paused, looked him dead in the eye, and said, "Hey man, I need to know how you do that. I want to watch you," and so I did. That was the day that I found out you can literally just control your own machine with scripts. The moment I found that out, paired with GPT-3 at the time, I started making scripts for everything. I even made a little desktop pet that, if you clicked it, would spawn a motivational message in the command prompt. I've also recently learned about bit flips (video in the classroom course) literal cosmic rays from SPACE can hit your RAM and make pretty big changes?????? Anyway, the thing is, I think you all are magicians. ˙⋆✮⋆˚࿔ Every single person who works with software and code and tech ...just makes things, you are all creators, and everyone's being so nonchalant about it. It's crazy. ✧˖°.

@Jake Van Clief We are producing a sales video for one of our Kelowna projects (Siren St. Paul). Generated in Higgsfield, finished in Canva, directed from an ICM workspace. I want to show what the structure did, because the structure did the heavy lifting. The setup. Five numbered stages: source, spec, boards, scenes, assembly. A _shared folder holds the brand facts and the method. Every stage has a CONTEXT.md contract (inputs, process, outputs, what I check before anything moves). Status is just what files exist. Nothing advances until I have read the last output. The core problem with AI video is drift. Same prompt in, different look out. The fix is a three board system (component board, frame library, motion board). The boards are the factory. Stage 3 builds them once, stage 4 references them in every scene. Nothing gets re-explained to the model, ever. Here is the part worth posting. Every stage boundary in the workspace is a gate. Cheap, boring, just a human reading the output before the next stage is allowed to touch it. Those gates caught three real failures before they could reach the film: 1. The curation gate caught a style hijack. One phrase in a shared suffix ("layered mountain haze") turned close up tiles into wilderness scenes. The prompt asked for an urban window glow and the model delivered a rustic mountain chalet. Four tiles rejected, regenerated, and the fix written into the run book (two suffix variants, one for landscapes, one for details). 2. The QA gate caught name contamination. The word "Siren" in the style preset description made the model paint a literal mermaid statue into a test image of a coffee cup. The gate exists precisely for this (five test generations on subjects the boards never contained). New rule: the project name never appears in generation side text. It is composite layer vocabulary only. 3. The review gate caught invented geography. The map in our "center of it all" scene had streets that do not exist. The gate put that frame in front of a human before anything downstream could use it, and the human happened to know Kelowna. The fix came from our own asset library (the amenities map from our showroom signs, used as an image reference). New rule: anything showing real geography generates from a real reference. Same rule that already protected the building (the hero shots are image to image from the actual rendering, so the architecture in the film is our architecture).

I automated my LinkedIn pipeline so the posts are basically a byproduct of work we were already doing. Here is how it works: - Everything gets captured raw → Meeting transcripts, voice notes, messy half-thoughts: they all get dropped into one inbox folder. Zero formatting, zero cleanup. Capture first, structure later. - An AI agent parses the inbox → It reads every drop and sorts what it finds into people (CRM update) , tasks, decisions, ideas, and content angles. Every single meeting gets asked " is there a LinkedIn idea in here?" ( I have flows for what is considered "content" , aligned with the brand) - I approve every item → The agent proposes, I decide. - Approved ideas land as drafts from a template, each one linked back to the meeting it came from. - The words come from a voice fingerprint → I fed my own transcripts into a profile of how I actually talk, so the drafts sound like me. - Once a month I sit down with the agent for a proper conversation → It brings what the industry published that month and what our own meetings taught us, and together we draft the company posts. - I edit, pick the visuals from a small set of templates, and schedule. The end goal: capture once, decide once, and let structure do the rest. If you want to build this yourself, start with the inbox. One folder where everything raw lands. Everything else grows from there.

A FEW WEEKS AGO... I posted that I had made a business, spun-off of one of my competition entries. Full disclosure, it's not exactly any of them (if you happen to go look). I just took what I had learned while building and applied it to a business strategy I wanted to try. And you know what!? I got a client. It's the first client I've gotten for another venture, outside of the main business I've been running. Considering I have spent 10's of thousands of dollars on other "tests," over the years, this was a HUGE win to actually make money from something. AND NOW I BUILD... Now I'm using ICM to automate the dumb stuff and enrich the good stuff. I'm constantly asking myself if what I'm building make sense, or if I'm falling into the trap that Jake talks about, where folks overbuild or build something that doesn't matter. I am pretty sure I'm doing it right because newer AI models won't affect what I've made, except to streamline the dumb stuff more... Anyway, I'm looking at the unit-economics of the business I'm building now, based off this one client that I'm trying to give the most amazing experience to, and I am blown away by the potential. It feels like it can't be real. I'm probably putting the cart before the house, so to speak, but I can't help to be optimistic looking at the numbers, of which, I will admit, I'm making a lot of assumptions about. SUCCESS LEAVES CLUES... I have been studying business for a long time now, to try to understand how riches are truly made beyond what the typical small-business owner does (average is like $50K a year or something). I feel I will lose my mind if it takes me 20 years to build wealth because I had tunnel vision and chose a very slow moving vehicle to get me there. And look, I know it's *all* hard, the grass is always greener, yadda yadda yadda (Seinfeld, anyone?), but it's still true that there ARE easier opportunities to make wealth. There just are. I've come back to this fact time and time again while searching for my next business model.... In simple business terms, if your CAC (cost to acquire a customer) is extremely low, 0, or even negative, and your LTGP (lifetime gross profit -- what the client pays you in total, for the entire duration they use your business) is high, then you're printing money. And if the profit margins are 80% or more? Then you're swimming in gold.

Context: My company is hell-bent in still making excel work for everything. 10-15 worksheets inside one file, THOUSANDS of rows and columns per sheet. Curious question, do agents follow the same ICM principles when you point them at an excel file? Ex. Routing logic in first worksheet so the agent knows where to go instead of searching the entire workbook for something.