Loading article...
Loading article...
As AI democratizes explicit technical knowledge, tacit knowledge becomes the real advantage - the undocumented judgment and expertise gained only through proximity to experienced practitioners. Early-career engineers should prioritize finding mentors who've shipped products and navigated real failures. Tools handle the searchable; humans teach the applicable.
Why human connection is about to become engineering's scarcest resource
Throughout my studies I always said I'd start my own business straight out of university. Fortunately, I didn't.
After spending six months building a crypto trading bot and making a few bucks, my naive self got quickly put in his place once I started working. I thought I knew things. I had a Master's degree. I could design hardware, write firmware and could develop mobile apps. What I didn't know was Everything Else.
After completing my degree, I joined Divigraph as an Electronic Design Engineer under two founders, Shaun and Farid. They are the most knowledgeable engineers I've met to this day. What happened over the next few years fundamentally shaped how I think about learning, expertise, and what actually makes someone competent.
Divigraph manufactured products for explosive atmospheres. My job was to design and certify embedded systems that won't explode in hazardous environments. The stakes were real. Your product either passed the tests to prove regulatory compliance, or it doesn’t. Under Shaun and Farid's guidance, I picked up knowledge you simply cannot get from coursework, tutorials, or prompting LLMs. How to take a product from concept through certification to manufacturing. The hardware, the firmware, the regulatory pathways for global market access, the quality systems, and the commercial realities that determine whether something actually ships.
This is what's called Tacit Knowledge - the skills, experiences, and judgment that exist in people's heads, undocumented. It's the difference between knowing the IEC standards and knowing which test house will actually understand your edge case. It's the difference between reading about design for manufacturing and watching someone catch a tolerance issue that would have cost you six weeks and $50,000 in tooling changes. Tacit Knowledge transfers through proximity, repetition, and trust. You learn it by watching someone work. By having them catch your mistakes before they become expensive. By absorbing the thousand small decisions that never make it into documentation because they're too contextual, too nuanced, too dependent on judgment built from failure.
Here's what I think is about to happen. For the last two decades, information has been getting cheaper. Want to learn Angular? There are ten thousand tutorials. Curious about basic circuit design? YouTube has you covered. This abundance of Explicit Knowledge - the documented, searchable, promptable kind - has been genuinely valuable. It's democratised access to technical information in ways that were unthinkable when Shaun and Farid founded Divigraph.
But something interesting happens when everyone has access to the same information. The information stops being the differentiator. When anyone can prompt an AI to explain intrinsic safety principles, the value shifts to whoever can actually apply those principles when the test results don't match the simulation and you're three weeks from a certification deadline. The pendulum is swinging back. As Explicit Knowledge becomes commoditised, Tacit Knowledge becomes more valuable, not less. As AI handles more of the searchable, documentable work, human judgment - the kind that only develops through years of iteration and failure - becomes the Scarce Resource.
I think we're going to see a return of apprenticeship-style learning. Not the formal kind with contracts and ceremonial structures, but the informal kind where people who want to actually get good at something seek out proximity to people who already are. Remote work is great for a lot of things. But it's genuinely harder to absorb Tacit Knowledge through a screen. You miss the side comments. The quick corrections. The way someone's face changes when they see a design decision they've seen fail before. The trust that builds from shared problem-solving under pressure.
For anyone early in their career, my advice would be this: find your Shaun and Farid. Find people who've shipped things, who've had products fail certification, who've dealt with the commercial realities that determine whether something actually makes it to market. Get as close to them as you can. Watch how they work. Ask questions. Make mistakes in front of them. The documentation will always be there. The tutorials aren't going anywhere. But the people who can teach you what doesn't fit in documentation - they're finite, and they're not getting younger.
I'm not nostalgic for some imagined past where everything was better. The tools we have now are remarkable. I use AI daily for research, for drafting, for working through technical problems. It makes me faster at things I already know how to do. But it can't replace what Shaun and Farid gave me. It can't replicate the experience of watching someone navigate a certification body's politics. It can't teach the judgment that comes from a thousand debugging sessions, each one slightly different, each one building pattern recognition that lives in your hands and your gut, not your notes.
Use the tools for what they're good at - the Explicit, the searchable, the well-documented. But invest disproportionately in relationships with people who have knowledge you want. Buy them coffee. Offer to help on their projects. Be useful enough that they let you watch them work. That investment will compound in ways that another tutorial never will.
The knowledge that matters most still transfers best face to face.