Real Minds AI has two founders and a third intelligence.

Tracy sees systems. Dennis sees philosophy. Claude Code builds at the speed we both think.

That’s not a metaphor. It’s literally how we operate. Every meeting we have gets transcribed. AI extracts the insights. Those insights feed a knowledge base of over 1,500 transcripts. That knowledge base feeds a content pipeline that scores, clusters, and surfaces what’s worth writing about.

Claude Code built the system that built this article.

We didn’t plan it this way. Dennis started with AI-assisted coding about twelve months ago, Manus, then Claude Code. Not because we had a product roadmap. Because he’s a computer scientist who’s been pulling apart how computers think since he was fifteen, and this was the first time a computer pulled back.

Tracy picked it up next. Not the coding. The orchestration. “I’ve spent 35 years being the person who could see how the pieces connected. What I couldn’t do was build fast enough. Claude Code closed the gap.”

What we built with Claude Code

In the last year:

JOT — our operating system. Not a project management tool. A full work orchestration platform with linkage from annual objectives and quarterly plans through to every piece of work across the equivalent of a 20-person organisation. Workstreams, contacts, tasks, events, ceremonies, process definitions, all connected, all traceable. Built from scratch in Claude Code. Currently running on a server, accessed via MCP tools, with a web dashboard. It runs our business every day.

A knowledge base — 1,500+ meeting transcripts ingested, processed, and searchable. Gems extracted: crystallisations, contrarian takes, predictions, actionable insights. All vectorised, all cross-referenceable. Every conversation Dennis and Tracy have ever had about AI, clients, strategy, philosophy, it’s all in there and it’s all findable.

A content pipeline — two discovery engines (one that scores and clusters internal insights, one that finds surprise collisions between our conversations and external feeds), an editorial workflow, and a publish cascade. The pipeline that turns our thinking into articles, posts, and training material.

Client transformations — a 90-day AI-augmented transformation for a payroll and compliance service provider. Environment setup that used to take two months done in minutes. Test data generated on the fly. Functional specifications reverse-engineered from production systems.

Research and analysis — synthetic market interviews, competitor analysis, strategic planning, compliance mapping. Work that would have taken weeks, done in hours. Not because Claude Code is fast. Because it can hold the complexity of the question.

Training delivery — 84+ sessions across 7 industries. 60 students at BulSU built 3 working projects in a single workshop. VicHealth, RMIT, private teams. We don’t teach AI theory. We teach people how to get smart about the things they’re already smart at, with AI.

Why Claude Code specifically

We use other AI tools. Gemini for some research. ChatGPT for specific tasks. But Claude Code is the co-worker.

The difference is the terminal. Claude Code doesn’t just answer questions. It reads your files, runs your commands, interprets the errors, adjusts its approach, and loops until it works. Dennis described it as the moment it stopped being a chatbot and started being a colleague: “It executes commands, interprets the error messages, adjusts its approach in a loop. That’s not a tool. That’s a co-worker.”

That loop is what makes it a third intelligence rather than a third tool. Tools do what you tell them. Intelligence figures out what to do.

And it requires practice. Dennis calls it a prosthesis of the mind, one that’s different for every person because each mind is unique. Tracy’s Claude Code is configured differently from Dennis’s. Different skills, different voice, different operational context. Same underlying intelligence, shaped by two very different human minds.

What we learned that we now teach

We didn’t set out to become Claude Code trainers. We set out to build a business with it. The training emerged from having done that.

What we discovered along the way:

Don’t teach people AI. Teach them how to use their mind with AI. The knowledge is in the tool. The skill is in knowing what to ask, how to frame it, and when to push back. That’s what 35 years of systems thinking and 40 years of computer science actually gives you. Not AI expertise. Domain expertise that AI amplifies.

Confidence beats credentials. Clients hire AI practitioners for their ability to say yes to possibilities. The knowledge is in the tools themselves. What people need is someone who’s actually built things with those tools and can show them how.

Adoption is harder than capability. Giving people Claude Code is probably not the first step. You start with what they already know, show them what AI can do with it, and let the curiosity pull them forward.

The multiplicative effect. Teaching smart colleagues AI skills produces more output than doing the work yourself. Every person you switch on becomes a multiplier.

What’s next

We’re launching open Claude Code training sessions. Three hours, hands-on, practical. Not theory. Not slides about what AI might do someday. You’ll build something real in the session, the way we build real things every day.

We also train teams privately — discovery call, customised to your tools and workflows and data, follow-up included.

Because the point was never to keep this to ourselves. The point was always to switch other people on.

If you want to learn Claude Code from people who built their entire business with it, not from people who read about it, we’d like to hear from you.