Master what machines can't replicate

If you've arrived here, I suspect you already sense the problem — even if you haven't yet had the language to name it precisely.

You've watched a junior colleague lean on an AI tool in a way that made you uneasy. Not because the output was wrong, exactly, but because you recognized that the process by which it was produced had bypassed something important. Some hard-won judgment they should have been developing. Some error they needed to make. Some friction that was, in fact, the whole point. You are not a technophobe. You are not opposed to AI. You are, however, deeply aware that expertise is not a database — that what you know in your hands, your gut, your clinical eye, your engineering instinct — was formed by years of immersed, embodied, often difficult practice, and that it cannot simply be downloaded.

That awareness is the starting point for everything this course does.

Tacit & Machine is my attempt to build a rigorous intellectual framework around a problem that is urgent but under-theorized. We draw heavily on Michael Polanyi's philosophy of the tacit dimension — the idea that "we can know more than we can tell" — and pair it with cognitive science on skill acquisition, honest analysis of what modern AI systems actually do and don't do, and cross-disciplinary case studies drawn from radiology, surgical training, chess, software engineering, education, and design. The goal is not to give you comfort or alarm in equal measure. The goal is to give you precision: the analytical vocabulary to evaluate any claim about AI and expertise in your own domain, and the design thinking to act on that evaluation.

I want to be direct about what this course is not. It is not a productivity playbook. It is not a list of recommended tools. It is not a reassuring narrative that says human expertise will automatically survive. It is a serious, cross-disciplinary seminar that asks genuinely hard questions — about deskilling, about the conditions under which automation helps rather than harms expert development, about what "human-in-the-loop" really requires structurally if it's going to mean anything — and trusts you to sit with the complexity long enough to think your way through it.

What you will walk away with is a framework that is yours: a conceptual architecture for distinguishing tacit from explicit knowledge with philosophical precision, a method for diagnosing hollowing-out risk in your domain, a set of design principles for AI-human workflows that actually preserve and develop deep skill, and a capstone strategy tailored to your field and your professional responsibilities. That last piece matters to me especially. The point of serious intellectual work is that it changes how you act — as a practitioner, as an institutional designer, as a mentor to the next generation of experts who will inherit this landscape.

I built this course because I believe the professionals who understand this problem deeply are the ones who will shape how AI is deployed in their fields — not the technologists who built the systems, and not the administrators who purchased them. That means you. I'd be glad to think through it together.