No AI Problems: Avoiding Becoming a 10x Dependent Developer

No AI Problems: Avoiding Becoming a 10x Dependent Developer
Photo by Pawel Nolbert / Unsplash

Calling this a reaction post wouldn't be fair because I don't necessarily disagree with the original post. Instead, consider it an expanded perspective.

If you haven't yet, read this post by Namanyay: https://nmn.gl/blog/ai-illiterate-programmers?

The key quote is:

We’re not becoming 10x developers with AI.

We’re becoming 10x dependent on AI. There’s a difference.

I would definitely agree with that. Being a 10x or whatever developer isn't about the quantity of output; it is about quality. Good solutions are simple.

The Problem and a Solution

Don't get me wrong. AI tools are very useful and I use them myself to help get through tedious and boring tasks quickly. If you're not using AI, I can only assume you enjoy typing and the RSIs that come with it.

Namanyay takes a similar position, pointing out how those learning to solve problems are not learning to solve problems if they can only do so with AI assistance. AI tools are useful and can help everyone work faster, but everyone needs to know the fundamentals.

Namanyay suggests a solution could be regularly fasting AI to make you and only you regularly work on problems to stay current. This isn't necessarily a bad idea and is something I've done over the past few years several times, each providing a renewed love for actual problem solving rather than pressing tab. It can be refreshing.

However, I'm not sure that such an approach is necessarily the best. Maybe rather than "No AI Days," we should have "No AI Problems."

I don't know the last time I manually did a long division problem, and while I could probably grok it, I don't really remember how from memory. Is that bad? I'm not sure. But I'll probably continue using a calculator.

Likewise, as AI continues to get better and do things that used to take a lot of time, it's becoming more clear that delegating such tasks can provide the most impact, letting you focus on high-value tasks.

Effective Application of Effort

I read a book a few years back called Run Studio Run by Eli Altman. It's about how to run effective agencies. Chapter seven talks about human resource allocation. The key idea in that chapter is "Keep your high-value employees focused on high-value activities." You'd think that would be obvious advice, but in more cases than not, within an organization, you'll find highly paid and highly skilled team members working on projects that are below their capabilities. Everyone wants an expert to do everything. As a result, you end up with overqualified employees doing tasks that anyone could either quickly learn to do or intuitively do.

This is an efficiency problem that's caused by a higher supply of expertise than the market really needs, pushing down the cost of inefficiency.

When experts can be had for not much more than an average Joe, firms are willing to put the expert on problems that aren't fully utilizing their potential because it may result in a slightly lower defect/problem rate than if using a less qualified person.

This economic problem has severely impacted tech for the past few years. Industry leaders say there is a talent shortage but often only have open positions for seniors with less than competitive comp. Why not get someone with 10 years of experience to do easy entry-level work if they can be hired for about the same or a little more than a good junior? It makes sense financially in many cases.

The industry application of this is that maybe rather than a talent shortage, we have a hard problem shortage. On a personal level, rather than toiling away at boring problems, use AI to do those tasks and focus deeply on hard, impactful issues. With access to tools like AI agents, everyone is a manager, and you must effectively utilize your AI team in a way that's effective.

No AI Problems

When first learning, every problem should be a "No AI Problem" where you think through and learn the ins and outs of what you're doing. Kids in elementary (hopefully) won't be given a calculator to help them solve 1+1. But maybe in middle school, they will get to use a "dumb" calculator that can help with addition, subtracing, multiplication, and division. Later on, they'll get a scientific calculator that can solve for x in equations. If they continue into college, they might use a computer and use a whole programming language to do highly advanced calculations. That's okay. We learn the fundamentals then we abstract that work using tools.

Learning to code and working as a professional coder must take a similar approach.

When I need to use a complicated math formula, I make sure I understand it. I might get AI to help explain it, but I don't just assume it is doing what I need it to. This differs from division, where I'm pretty confident that / will divide and modulus % will give me the remainder.

Rather than having mathematicians spend all day doing trivial math problems, let them focus on the big problems and use machines to automate the rest. Same thing with programmers. Use AI agents and tools to quickly solve small, tedious problems and focus your effort on creating highly efficient and effective solutions where it matters without relying too much on AI.

I challenge you to think through your most important problems yourself. These "No AI Problems" should be problems where small improvements will result in big gains and where architecture really really matters. By following this approach, you can delegate and abstract everyday tasks while putting your focus on problems that matter, ensuring you don't outsource all of your thinking and become dependent on AI.