In theory, AI allows all of us to do “everything, all the time.”1
Non-coders can now “code.”
Non-designers can now “design.”
Non-writers can now “write.”
And so on.
Reality is more nuanced.
AI does raise the floor. Anyone can now produce something reasonably good.
Without AI, I couldn’t create the images I make with Google’s Nano Banana or the software I build with Claude Code. But to rise above this elevated floor still takes another ingredient: expertise.
I’ve written before that AI mirrors the quality of its input. If you’re an expert at something, you can describe what you want with precision non-experts can’t match, and AI rewards that precision.
Take an accomplished director friend of mine. He has feature films to his name and top-notch film school credentials. He’s a master at spotting stories, turning them into scripts with sharp dialogue, and knowing exactly how to describe the scenes and shots he needs.
Those skills translate perfectly to AI video tools. He can now do at a computer what used to take a crew.2 Few others can match what he can achieve with those tools, simply because they can’t describe what it is they need in the same kind of jargon and detail.
I’m another example.
I’ve done product management, worked with lots of freelance developers and designers, and done web design in the 90s. The only blocker to developing digital products myself has always been my inability to code.
AI has removed that barrier. I can now describe in English what I want, and, because of my previous experience, do so in a way most others can’t.3
So if you’ve spent years developing genuine expertise, don’t fear that AI is giving it away for free. Yes, it raises the floor for everyone, but it also removes the ceiling for experts.
One of my favorite quotes, from the song Life in the Fast Lane by The Eagles.
This, of course, is also an ethical dilemma. For exactly this reason, my friend hasn’t done much AI video making yet; he’s conflicted about the IP theft and replacement of many traditional filmmaking jobs.
Real programmers using AI coding tools will still put me to shame, and whatever I built is currently not ready for large-scale production. I can neither judge nor fix security and scalability issues, though these challenges are shrinking as the models get more capable.
this echoes my thinking and was articulated in a very concise way. thank you!