When will the Ideas Guy be enough?

There's been a lot of chatter this week about "clarity of thought" in AI-assisted development. It resonates deeply with me. The more I work with AI tools in day-to-day software development, the more I find myself producing specifications, implementation plans, and testing strategies at a scale I never have before.

Notice I said producing, not writing. That distinction matters. The ideas, the process, the steps—those are all mine. The mundane sludge of getting it all to paper? That's all the machine.

The more I refine this art, the more I notice how velocity shifts with clarity. Whether I'm building a small feature, fixing a bug, or developing a new interface, the clarity I bring to specifications and plans directly determines the implementation path. Break things down precisely, and the AI has a clean roadmap. More importantly, I have a refined context to return to when problems arise or adjustments need to be made.

My wife reminds me that my success with AI is rooted in my background in the craft. I can tell the AI when it's wrong. I can critique code performance and discuss different architectural approaches. There's truth to this—it's analogous to a seasoned mechanic diagnosing car problems by sound or feel, then communicating fixes to junior shop mechanics.

But I wonder: for how long does that experience advantage matter?

Watching my 13-year-old son sit in front of Claude and build browser games gives me pause. He's coded practically nothing in his life. He knows how to paste errors back to the AI and describe what he wants, but he brings zero critical theory to the table. And yet... the games work.

His success isn't an anomaly—it's a signal. Five years ago, what he's doing would have required months of learning. Today it requires an idea and patience to iterate with an AI.

We're at a fascinating point on the AI curve. As we continue learning and evolving our craft, the tools are changing underneath us at an unprecedented pace. This tight loop—from learning to production and back—is unlike anything we've seen in the history of software development.

So I find myself asking: how long before process and critique are naturally embedded in these tools? How long before my son with his truck game idea can produce something polished without understanding tech stacks, performance considerations, or architectural patterns?

I don't think he'll be building aircraft control systems anytime soon. But the next generation of arcade games? That might already be within his reach.

And where does that leave me? In a fascinating position—one that breeds creativity and space. My decades of experience still matter, but differently than they used to. I'm not writing less code; I'm writing less obvious code. The AI handles the boilerplate and the repetitive patterns. I handle the interesting parts: the "wait, that approach won't scale" moments, the architectural decisions that don't reveal themselves until you've been burned a few times.

So when will just being the ideas guy be enough? I think we're already splitting into two paths: those who bring critical thinking to AI collaboration, and those who bring only requests. My son represents one future. I represent another. Both are producing working software.

The question isn't whether being the ideas guy is enough. It's whether you're bringing ideas worth building.

Full Disclosure: This article is based on daily experiences with Claude Code (Pro) and Cursor Enterprise.