PowerUp AI Mastermind — April 2, 2026
“If it’s stupid and it works, it’s not stupid.” — Don Back (military proverb, applied to vibe coding)
This Week in 30 Seconds
- Lou returns from Mexico — nice to be back; session opens with prep for a 20-minute presentation for Amy’s group
- Eight eras in five years — Lou’s AI evolution framework sparks a rich debate about where most audiences actually are vs. where the frontier has moved
- The curse of the expert — Dirk and Bally name it plainly: you’re in a bubble of super-experts; most people still aren’t on Claude
- Codifying judgment — Lou’s deepest current insight: the real leverage is extracting expertise, perspective, and values into AI workflows, not just prompting better
- Ambient intelligence — skills in every folder, every piece of content becoming live and accessible — a vision for what a truly AI-integrated knowledge base looks like
- Dirk’s investor story — pitch deck built with Claude, investor meeting appeared 12 hours later, CTO search prompted Claude to offer: “I can build you the app instead”
- Micro-apps and tiny tools — Lou’s 100-utility library: things that would have cost $25-35/month as SaaS, now built on demand and available as skills
Lou’s 20-Minute Presentation Dilemma
Lou returned from Cozumel with a presentation problem: 20 minutes, a new audience, and too much to say. The prep discussion that followed became the session’s organizing thread — and surfaced several of the most valuable insights of the quarter.
The context: Lou had been invited to speak at Amy’s event, presenting to coaches and consultants who are “fairly newbie in terms of tech.” His initial instinct was to show the frontier — paradigm collision, ambient intelligence, skills as judgment codification. The group pushed back, and the pushback was instructive.
Dirk cut to it: “You’re in a bubble of super-experts. You always think every week in terms of AI speed. 99% of humans are behind you. Pick one thing.” Don Back refined it: start with the 8 eras as context, then position the audience where they likely are (eras 4-5), and look over the horizon just enough to excite them without overwhelming them.
Jamie pointed out the practical reality: she’s been on calls where people still don’t know what skills are. “From my impression, people are really not beyond trying to figure out how to prompt well.” Bally ran a session on skills last week and discovered that none of her participants were on Claude yet. “So that was actually a lesson last week.”
The resolution Lou landed on: one focused slide on the 8 eras as context, show where they probably are, give them one actionable thing (skills — here’s what they are, here’s how simple the text file is, here’s how Claude creates it for you), and be ready to step up to agents or Claude Code if the room is ahead of him.
💡 What This Means for You
If you teach AI, this conversation is your calibration. When you’re deep in the work, the average audience’s starting point feels further back than it actually is — but also further back than you realize. Show the arc, meet them where they are, offer one step forward.
Deep Dive: Insight - Teach One Era Ahead of Your Audience, Not Eight — why teaching from the bleeding edge creates overwhelm, and how to calibrate for maximum impact.
Eight Eras in Five Years — The AI Evolution Framework
Lou shared a visual framework he’d built with Claude that maps the evolution of how knowledge entrepreneurs have interacted with AI from roughly 2021 to now. The eras progress through observable shifts in how people use AI — not in what the technology can do, but in the mental model and workflow of the user.
Roughly: Observer (willing to experiment, Q&A mode) → Prompt Crafter (prompt engineering, prompt packs) → Framework Builder (mega-prompts, meta-prompting) → Platform Builder (custom GPTs, Claude Projects, RAG pipelines) → Context Curator (high-judgment advisory, strategic analysis) → Strategic Delegator (skills, agents, workflow orchestration) → and beyond.
Don Back’s reaction: “This is an eye-opener for me. I’m living in the moment and not aware of this stepwise evolution that’s been happening.” Donald confirmed: “Everyone is really overwhelmed. How to keep up — everyone feels that.”
Kasimir made the practical observation: people are either chasing every new model, or so overwhelmed by the pace that they’ve checked out entirely. His advice: pick two or three core platforms, learn how they work, stay with them. Consistency compounds.
The group consensus on where most business users are: somewhere between eras 3-5. The emerging competitive moat is in era 7 — not better prompting, but transferring expertise and judgment into reusable skills and micro-applications.
“Everything you thought you had to learn, you don’t need to anymore. Good news is, it’s gonna all be done for you. All you have to do is learn how to create this text file.” — Lou, describing the skills pitch for Amy’s audience
Codifying Judgment — The Real Leverage Point
This was Lou’s most personally significant disclosure of the session. When asked what he’d teach if he could only teach one thing, he kept circling back to the same answer: not skills, not prompting, not agents — but the extraction and codification of expertise, perspective, and values into AI workflows.
His current experiment: have Claude Opus 4.6 watch him work in a natural, uncontrived conversation, and then ask it to articulate what it noticed — not what was said, but why it was said, what implicit frameworks governed the choices, what the expert was doing below the surface. Then codify that into a skill so the AI can apply it strategically going forward.
“Look at what I said, but more importantly, imagine why I said it. It’s pretty darn good at figuring that out. And then what I want to do is just codify that.” — Lou
This is the deepest version of skills-as-judgment: not just “here’s a process I want you to follow,” but “here’s how I think about problems — now apply that to everything you do for me.” Lou described it as the NLP modelling pattern applied to AI: watch the expert do their thing, notice what they’re not conscious of, extract the micro-decisions and implicit reasoning, package it for replication.
The AIM writing team skill was built exactly this way — Lou iterated on an article, then asked Claude to reconstruct the process: first you asked for research, then outline, then critique… That pattern became a skill.
Deep Dive: Insight - Codify Your Judgment Into Skills, Not Just Prompts — why the most powerful AI workflow isn’t a better prompt, it’s a captured judgment model.
Ambient Intelligence — Skills in Every Folder
Lou surfaced a vision he’d been developing before the presentation prep took over: ambient intelligence through folder-level skill inheritance. The idea: if you place a skill file in every folder of your working directory, and Claude automatically inherits skills through the project hierarchy, then every piece of content you have becomes a live, AI-accessible asset. Your knowledge base isn’t passive storage — it thinks.
He hadn’t fully fleshed it out yet (“something I was brainstorming just before we met”), but the concept is compelling: instead of building an explicit RAG pipeline or a separate knowledge management system, you build intelligence into the folder structure itself. Every folder has its own micro-context, its own mini-skill that tells Claude how to interpret and use what’s in it.
Deep Dive: Insight - Ambient Intelligence — Build a Skill in Every Folder to Make Your Entire Knowledge Base Alive — how to turn static file storage into a live, AI-navigable intelligence layer.
Dirk’s Investor Story — Claude Builds the CTO Search App
Dirk came to the session with a real story that perfectly illustrated the “build tiny tools” theme. He’s working on a pre-seed funding pitch. He built a pitch deck with Claude — fast, solid. A friend who happened to be a CEO saw it and said: “I have an investor meeting tomorrow at 11, I’ll put it on the table.” Twelve hours’ notice. The investor was interested; the team structure was the issue.
So Dirk started working with Claude on an executive search brief — precise CTO profile, specific criteria. Claude offered: “I can build you an app. They won’t have to type it in — they just type in what they’re searching for, it connects to LinkedIn.” Dirk’s response: “It was too fast. I said, just give me the code.”
The app didn’t end up being the right solution (it required a LinkedIn team account, which the researchers didn’t have), but the point stood: Claude’s first instinct in response to a process problem is increasingly to offer a tool, not just advice.
Lou connected this to his own practice: “Every time I have a question, I ask it to write an app for me. I’ve got about 100 utilities now — things that would have been a pain to do manually, and not enough of a pain to pay $25/month for a SaaS. Now I just have them in my local GitHub repo.”
Don Back’s verdict: “If it’s stupid and it works, it’s not stupid.”
Deep Dive: Insight - Build Tiny Tools That Remove Real Friction — why the era of custom micro-apps means you no longer have to tolerate repeated friction.
The AIMM Writing Team and Skill Genealogy
Lou traced the origin of the AIMM writing team skill — a 3-4 skill pipeline for article generation — as an example of how skills develop naturally from working conversations. He iterated on an article, going through research, outline, multiple rounds of critique, and revision. At the end, he asked Claude to reconstruct the process: what did we do, in what order, and why? That reconstruction became the skill.
He plans to use this story in Amy’s presentation: show them the skill that generates the mastermind recap, then show them the writing team, and say “this is just a codified version of how I write an article.” The leap from ‘I write articles’ to ‘I have a skill that does it’ becomes visible.
This is also a template for any member who wants to build their first skill: do the thing naturally, then ask Claude to watch you do it and describe it back.
GEARS Status Check
Brief but important: Lou checked in on GEARS progress. Don is 15 minutes behind on submitting his ontology corrections. Kasimir needs a follow-up. Lou’s gentle nudge: the next stage — schema generation and content generation — depends on correct ontologies. If yours needs correction, now is the time.
Community Corner
Donald’s note in chat — “Teach the SKILL of all skills” — was a sharp summary of what Lou was trying to do in Amy’s presentation. The skill that teaches you how to make skills is the ultimate entry point.
Bally’s field report was genuinely useful: she ran a session on skills for her clients last week and discovered they weren’t on Claude yet. This is real feedback from the frontline. The gap between where this group operates and where most coaches are is larger than anyone in this room defaults to assuming.
Don Back’s military quote — “If it’s stupid and it works, it’s not stupid” — was so appropriate to the vibe-coding discussion that Lou immediately asked Claude to save it as the recap pull quote. Done.
Links Shared in Chat
No links shared this session.
Try This Before Next Session
Build one tiny tool that removes a recurring friction point from your week.
- Identify something you do manually that takes 5-15 minutes and repeats at least weekly (reformatting something, looking something up, processing a type of document, generating a standard output)
- Open Claude and say: “I have a recurring task: [describe it]. Build me a simple tool that automates this. It should run locally, use simple inputs, and give me the output I need without me having to think about the steps. Keep it minimal — I want something I’ll actually use.”
- If Claude offers to build an app, let it. If it offers a skill file, accept that too.
- Use it once and notice: what does it feel like when a repeated friction point disappears?
Then: tell the group what you built next session.
Open Threads
- How can PowerUp teach advanced AI concepts in a way that creates confidence instead of overwhelm?
- What is the simplest repeatable process a client can convert into a first useful skill?
- Which kinds of founder friction are best solved with tiny custom tools rather than purchased software?
- Lou’s deepest teaching moment — the judgment codification / NLP-modelling-with-AI approach — is it too advanced for a first audience, or is there a simpler entry point that leads to the same insight?
Next session: TBD