Everyone wants the tool list. Which models, which prompts, which stack. I keep wanting to walk people back to the part that has nothing to do with tools, because that’s the part that built any of this.<\/p>\n\n\n\n
The order matters, so let me reflect back on it. It started with twenty years of judgment. That became written IP. The IP became GNYS “Genius”. And GNYS built the products. Expertise, then a brain, then the things the brain makes. Nobody believes the sequence until I walk them through it, so here is the walk.<\/p>\n\n\n\n
The judgment came first, and one piece of it crystallized years before I built with AI. At Deloitte’s national sales office, I was the creative strategist, helping with UX and experience design for an internal enablement platform serving 16,000 employees. The firm’s knowledge lived in pockets, some in the cloud, some on someone’s laptop; all of it needed in the middle of live deals, the pursuits the firm refused to lose. Working with the firm’s leadership, we landed on a constraint that has governed everything I’ve built since: surface the right thing at the right moment without handing anyone an article to read, a lecture to watch, or a course to take. Sixteen thousand people means every level of seniority asking the same question from a different altitude. So the system had to first determine who was asking, then answer at the altitude they actually work at. No model existed to do that then. The principle did.<\/p>\n\n\n\n
When I started writing down the methodology, that altitude principle went in alongside everything else. I wrote it as a system, with rules strict enough that a machine could run them and a junior could not bluff past them. The frameworks I’d built in client rooms across categories and company sizes went in too: the noun a company owns in the customer’s mind, the 4-Level Positioning Canvas, Gravity versus Glitter, the conviction that a company proves its positioning through costly decisions, the kind words alone can’t fake. I built the judgment into infrastructure before I built anything that ran on it. You write the playbook before you field the team.<\/p>\n\n\n\n
That corpus became GNYS. Genius, literally; the name is what went into it. I stood it up on Lovable and Supabase, and it works the way the Deloitte platform was supposed to. It reads who it’s talking to and meets them there. A founder gets one altitude, an enterprise strategist another, from the same brain. It runs where people already are, inside a product, over WhatsApp, in Slack. And it delegates. GNYS hands research tasks to a separate worker agent, and I gave that agent a budget and a daily cap, the way you hand a new hire a card with a limit. I built the manager first, and then I let the manager hire.<\/p>\n\n\n\n
The corpus stays the single source. When client work teaches me something, I revise the methodology in one place, and everything downstream inherits the revision. The products send lessons back the other way. Nothing in the stack sits frozen.<\/p>\n\n\n\n
Kit came first of the products. One hour, self-guided, built to find the single noun a founder owns. An AI CMO and an AI CTO sit inside it, so the work runs without me in the room. It stands in for the kind of positioning engagement that normally runs into five and six figures. The brain, packaged small enough for one person to hold.<\/p>\n\n\n\n
Analyzer came second, and it’s free on purpose. It points the brain at a company from the outside and reads how the market actually sees it, sorted into four plain questions: what you think you sell, what you actually sell, what customers say they buy, and what they actually buy. The gap between those is the product.<\/p>\n\n\n\n
Building the engine under that is where the management instinct stopped being a metaphor. Early on, a single run on CalDigit put twenty-seven candidate customer voices into the pipeline and saved zero at the end, and I couldn’t tell you why. A query mistake, a fetch failure, an over-strict checker, or simply no real voices out there to find. The logs couldn’t tell those apart. So I rebuilt it the way you run an accountable team. Every step became its own role with one job. The hunter that finds a source can’t be the fetcher that reads it. The extractor that pulls a quote can’t be the writer that summarizes it, because a model allowed to do both will eventually write prose where a real quote belongs. An independent checker verifies every quote against the original bytes, never against another model’s summary. Nobody marks their own homework. After that, a run that ends in zero became a five-minute question with an answer, instead of a mystery.<\/p>\n\n\n\n
Monopoly came third, the same brain at enterprise scale, for companies where a wrong self-image gets expensive. It runs on more than seventy-five sources, each a specialist with a narrow brief: regulatory filings, court records, employee reviews, financial disclosures, competitive signals. One model writes the analysis. A second model, paid separately, exists only to attack what the first one wrote. Another filters the noise before anything reaches a person. Read that list back, and it’s an org chart, which is the whole point. <\/p>\n\n\n\n
The reports land in minutes, and enterprise buyers have checked them against research that runs fifty to two hundred and fifty thousand dollars and found they hold up.<\/p>\n\n\n\n
Then the part I didn’t plan for. GNYS built Monopoly, and now GNYS lives inside Monopoly, Kit and Analyzer \u2014 coaching people through their own reports. The manager became part of the product it built.<\/p>\n\n\n\n
Read the whole build back, and there’s no AI story in it. There’s a value chain, a set of outcomes, a workflow, and a person at the end of each one I was trying to take work and confusion away from. The model is the last decision I make. The first is the person. By the time a model shows up, it’s a specialist with a brief, the same as the fifty data sources, the same as the worker agent, the same as the analysts and designers and strategists I managed years earlier in rooms with billions on the table. <\/p>\n\n\n\n
The principle is old enough that the most quoted version came with a price tag. WWDC, 1997. A developer told Steve Jobs to his face that he didn’t know what he was talking about. Jobs granted the man was right in places, then said the line people still clip: you have to start with the customer experience and work backwards to the technology.<\/em><\/strong> The part they cut is the cost. He’d made the opposite mistake more than anyone in the room and had the scar tissue to show for it. In the same answer, he let a technology die that he admitted could do things nothing else could, because it didn’t fit where the customer was going. A named casualty. That’s the bill the principal runs when someone actually pays it.<\/p>\n\n\n\n I paid a version of it. Years writing the methodology with nothing to ship, while people wrapped an API over a weekend, called it a company, and raised on it. I could have done that. The lens I’d spent twenty years building wouldn’t let me field a team with no orders.<\/p>\n\n\n\n Which is the only durable thing here. Anyone can rebuild my org chart in a weekend. The providers are public, the models rent by the token, the wiring is a tutorial. The briefs took twenty years, and there is no prompt for that. Most of the slop in your feed is an army with no orders, specialists wired together by someone who never asked what the person on the other end was trying to do.<\/p>\n\n\n\n Build with AI, by all means. Just notice which word in that sentence was ever doing the work.<\/p>\n","protected":false},"excerpt":{"rendered":" Everyone wants the tool list. Which models, which prompts, which stack. I keep wanting to walk people back to the part that has nothing to do with tools, because that’s the part that built any of this. The order matters, so let me reflect back on it. It started with twenty years of judgment. That […]<\/p>\n","protected":false},"author":1,"featured_media":4502,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_coblocks_attr":"","_coblocks_dimensions":"","_coblocks_responsive_height":"","_coblocks_accordion_ie_support":"","footnotes":""},"categories":[84],"tags":[],"class_list":["post-4487","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4487","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/comments?post=4487"}],"version-history":[{"count":3,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4487\/revisions"}],"predecessor-version":[{"id":4503,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/posts\/4487\/revisions\/4503"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/media\/4502"}],"wp:attachment":[{"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/media?parent=4487"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/categories?post=4487"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/paulsyng.com\/blog\/wp-json\/wp\/v2\/tags?post=4487"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}