A year ago I was the guy rolling his eyes at the AI demos.

I tried to use it back then to score sales leads, and I couldn’t get it to reliably do something as basic as flag every company with more than 25 employees. A plain “x > 25” was beyond it.

That wasn’t the 10x revolution everyone kept promising. Not yet.

Things have changed fast.

A year later, I just spent two weeks and about $4,000 in AI tokens rebuilding our entire website. I did it with what I’ve been calling our content engine: an AI agent I put together for the job, wired up with an MCP (Model Context Protocol) server, a Claude plugin, and a few other parts. It produces new content at a pace that still feels a little insane.

The tools genuinely got good.

But the pitch that they turn everyone into a 10x developer is still broken in a couple of ways, and once you see how, you spend your tooling budget very differently.

Writing code was never the bottleneck

Here’s the flaw under all the hype. How fast you can type code was never the thing slowing your team down.

The slow part is everything around the typing: knowing what to build, having the authority and ownership to go build it, and the skill and the time to do it well. AI is great at the typing, the easy and smaller part of the job. It does almost nothing for the rest of that list.

The numbers say the same thing. Atlassian’s 2025 survey of 3,500 developers and managers found that half of them lose ten or more hours a week to non-coding work: unclear requirements, meetings, waiting on a decision or an unblock. (atlassian.com) Most of the week was never spent writing code in the first place, so making the writing faster can only ever move a slice of it.

A senior developer at the research firm DX put the real gain plainly:

The easy tasks get a little easier, and a four-day task might take three.

Across 400 companies, DX saw AI usage jump about 65% while output rose only around 10%. The typing got faster, the shipping barely moved.

On hard work it can even reverse. METR ran a randomized trial on experienced developers in mature codebases. They predicted AI would make them 24% faster, and afterward they felt it had. They were actually 19% slower.

The tools feel fast while quietly costing you time.

Google’s DORA team put the cleanest frame on it: AI is an amplifier. Give it to your above-average developers and they go a little faster. Give it to the bad ones and it just turns them into AI slop machines. It magnifies what’s already there, it doesn’t manufacture it.

AI will write the code for you. It still can’t tell you what to build.

If everyone’s a 10x developer, nobody is

Here’s the second flaw, and it’s the one that should end the hype on its own.

If AI really did turn every developer into a 10x developer, then 10x would just be the new normal.

Everybody fast means nobody’s ahead.

The label only means something because it describes a gap between one person and the rest. A tool everyone has doesn’t create a gap. It lifts the baseline and leaves the spread exactly where it was.

Even in the world where the vendors are right, the 10x developer is still whoever was going to be the 10x developer anyway.

I’ve always been a 10x developer

I’ll say the thing the myth-busters won’t. The 10x developer is real, and I’ve been one. I’ve founded and sold 3 SaaS companies because of it.

The reason why is simple: I know what needs to be done. I know how to do it. And I can just go do it.

That is what creates a 10x developer.

There’s no organizational friction in my way. I don’t have to answer to anyone, ask for permission, wait on a ticket, or schedule a meeting before I start.

Take the content engine. Nobody handed me that project. I saw we needed content faster, decided how to build it, and built it. That’s the unlock, and it has almost nothing to do with how fast I type. It’s knowing what to build plus the authority to go build it without anyone slowing me down.

That is also why 10x shows up in startups and founders far more than in big companies. It’s not that the people are better.

It’s that the friction is gone.

There’s a second thing, and it’s less comfortable to admit. I worked around the clock on that content engine for two weeks and let almost everything else on my plate slide, because I was so infatuated with building it that I couldn’t focus on anything else.

It also takes passion for the problem.

That obsession is the magic in the bottle for 10x developers. When a problem grabs me, I disappear into it. A lot of that is ADHD, and the hyperfocus that comes with it is as much an edge as it is a problem.

If you want the longer version of that ADHD story, I wrote it up here.

Why didn’t anyone on my team build it?

That question has nagged me since I finished the content engine. My team had the same tools I did. Any of them could have built something like it. Nobody did.

Part of the answer is uncomfortable, and it lands on me. If the people around me are waiting to be told what to do, it’s partly because somewhere along the way they were trained to wait. They got handed tickets instead of goals, and nobody made it safe to go build something nobody asked for.

That’s a leadership failure before it’s a talent one.

My job stopped being the person who builds the thing. It became the person who builds the conditions where someone else does, and that shift is harder than building it myself ever was.

So the leadership work is concrete: hand people real goals instead of tickets, give them the authority to chase those goals however they see fit, and make it safe to build something that isn’t on the roadmap. Then get out of the way.

The problem is that can also be a slower at the beginning. It takes time and a different work culture for that to grow.

But here is the real key: ownership is the lever AI will never be.

I’d be lying if I said better leadership fixes all of it.

The ones who turn into 10x developers are the ones a problem can grab the way it grabs me, and you can’t install that with a goal-setting framework or an AI license. Plenty of strong engineers treat the job as a job, do it well, and go home, and that’s completely fine. You just can’t count on them to go build the thing nobody asked for.

That’s the real gap with AI right now. The gains don’t come from the tool; they come from the thinking, strategy, and judgment it takes to aim the tool at the right problem, plus the drive to do it unprompted. AI will execute almost anything you can specify. It still can’t decide what’s worth building, and deciding what’s worth building is most of the job.

At Full Scale, we changed how we interview developers a long time ago to focus much more on how they think: problem solving and product thinking. That’s the real unlock, and it’s what you need if you want to unlock real gains from AI.

Finding a true 10x developer is still a lottery ticket. But it starts with building the right work culture around them. They will never exist in a company full of red tape and bureaucracy.

So stop buying 10x

If you’re an engineering leader deciding how much to spend on AI this year, spend it.

The speed on the routine work is real and it adds up. Just don’t buy it expecting the 10x the hype promised, and don’t let it distract you from the work that actually moves the needle.

The developers who get the most out of AI are the ones who were already good at the parts it can’t do. If you want a 10x team, hire those people and give them room to run.

AI is leverage on judgment and ownership you already have. It was never a substitute for it.