The graveyard in the corner
Guitar you stopped playing. Website you never finished. Baking phase, climbing phase, the two months you were obsessed with electrical wiring. If you have a collection like this, you probably have a name for yourself: dilettante. Unfocused. Can't just pick a lane.
I had a version of this. Marketing, then engineering, then product, then design systems, then AI infrastructure. People would ask what I did and I'd struggle to give a clean answer. "Software and strategy, kind of."
I've been thinking about this differently lately — because I think the people who built the messiest skill graveyards are about to have a very good decade.
What AI actually commoditizes
Here's what AI is good at: depth without breadth. Ask it to write code in a single language. Analyze a dataset. Draft a legal clause. Summarize a research paper. It does these things faster than most humans, at a fraction of the cost.
What it's genuinely bad at is synthesis across domains. The judgment call that requires you to understand how the technical constraint, the business model, the user psychology, and the team dynamics all interact — at once. That's not a deep skill. That's a wide one.
The specialist who spent 10,000 hours perfecting a single craft is now competing with a tool that has processed billions of examples of that craft and never sleeps. The person who has a working understanding of five different domains — who can see the engineering problem as a product problem as a communication problem — isn't competing with that tool. They're the one directing it.
The M-shaped person
There's a concept I keep coming back to: the M-shaped mind. Not T-shaped (broad knowledge, one deep specialty) but M-shaped — multiple pillars of depth, connected by curiosity.
The web developer who used to farm. The product manager who has a theater background. The engineer who spent three years doing sales. On paper, these look like careers that couldn't make up their mind. In practice, these people can do something almost nobody else can: far transfer.
Far transfer is when you take a concept from one completely unrelated field and apply it to another. The farmer who codes doesn't just understand systems in the abstract — they understand organic growth, emergent behavior, what happens when you optimize one variable without understanding the ecosystem it lives in. That's a perspective that no number of computer science courses will give you.
This is where AI creates leverage. Not for people who know one thing deeply. For people who know several things — well enough to connect them, well enough to direct a specialized AI that handles the execution.
Serial mastery, not simultaneous mastery
There's a trap here. Reading this and thinking: okay, I need to learn five things at once. That's how you build a graveyard instead of a portfolio.
The people who pull this off — the real polymaths — do it sequentially. Seasons of focus. A few years deep in music, then a few years deep in software. Not abandoning the music, just finishing that phase of construction. Each season builds a pillar. Eventually the pillars connect.
I've watched this play out in my own work. The years I spent doing UX before writing code changed how I think about every API I've built since. The time I spent in marketing before taking on engineering leadership changed how I frame technical decisions for non-technical stakeholders. I wasn't wasting time switching tracks. I was building the M.
The mistake is trying to build all pillars at once. One season at a time. The bridge between them builds itself.
Redundancy as a survival strategy
Elephants rarely get cancer. The reason is genetic redundancy — multiple copies of the same cancer-fighting gene. If one fails, another takes over.
Your scattered interests are redundancy. If AI automates the specific thing you're known for, a specialist is in trouble. A person with five partially overlapping skills has five different pivots available. They're not replaceable in the same way, because there's no clean job description that maps to "the person who connects all these things."
I've thought about this with my own setup. I run 26 AI agents handling tasks across engineering, marketing, finance, and operations. The agents are specialists. I'm the M-shaped person directing them — translating between the business problem and the technical implementation, between the user insight and the product decision. The agents are faster than me at any individual task. But they can't see the whole board the way I can, because seeing the whole board requires having played in every corner of it.
What this means practically
If you're early in your career and you're feeling pressure to specialize: specialize for now, but don't close the other doors. Build one pillar deeply. Then go build another.
If you're mid-career and you feel behind because your path was nonlinear: you're not behind. You've been building something that takes longer to compound but is harder to replicate.
If you're a founder or a team lead: the people you actually want aren't the deepest specialists. They're the people who can hold multiple domains in their head simultaneously and make judgment calls that span all of them. Those people are getting rarer as education and career systems keep pushing everyone toward a single lane.
And if you're looking at what AI does to your work: it handles execution inside a domain. What it needs, and what it can't provide, is judgment across domains. That's the M-shaped skill set. That's the thing worth building.
The graveyard in the corner isn't a graveyard. It's a portfolio you haven't named yet.