The Leverage and the Loss

I've been walking again. Ten thousand steps a day this past week, through the hills in Seattle. I leave with my head full of the usual noise. Token limits, model benchmarks, another company announcing layoffs and pointing vaguely at AI as the reason. Meta cut 8,000 people recently. The industry is moving fast and the ground underneath it is shifting.

But somewhere around the third hill, the noise fades. The trees are just trees. The clouds are just clouds. The incline makes my legs burn and my breathing shorten and I welcome it because it's real and it's mine. No model generated the view from the top. No tool optimized the route. I walked it.

By the time I'm home, the takes and the panic and the predictions have all gone quiet, and what's left is something closer to clarity.

I am not writing this from the outside. I use AI tools every day, and they have made me better at what I do.

I built Pura Letra, a reading app, almost entirely with Claude Code. I'm an experienced engineer. I know how to architect software, how to make decisions about databases and infrastructure and tradeoffs. But Claude Code handled the implementation at a speed that would have taken me months on my own. What used to be a six-month side project became a working product in weeks.

AI also helped me build the financial model I wrote about earlier in this series. My specific numbers, my specific debt, my specific cost of living, all structured into a spreadsheet that finally made the exit math visible. I could have done it myself. It would have taken me a full weekend. Instead it took an evening.

This is the leverage. Twenty-five years of engineering experience, amplified by tools that can move as fast as I can think. For someone like me, building toward independence, that leverage is everything. It means I can ship products, test ideas, and iterate without a team. It means the economics of being a solo builder have changed in my favor for the first time in my career.

I wrote about that in an earlier essay. AI is the fourth technology shift I've lived through, and it's the first one that favors the individual over the institution. I still believe that.

But I've started noticing something else.

At work, and across the industry, there's a growing pressure to use AI for everything. Executives want to see adoption metrics. Managers want to know their teams are "leveraging AI." I've heard that Amazon tracks developer token usage to measure productivity. The message is clear: if you're not using the tools, you're falling behind.

And people are using them. Engineers, writers, designers, analysts. The tools are good enough now that you can produce work that looks competent without fully understanding what you produced. You can generate code that runs, copy that reads well, analyses that seem thorough. The output passes inspection. The question is whether the person behind it is learning anything in the process.

I think about this more than I probably should. When I was coming up as an engineer, there was no shortcut through the fundamentals. I sat with Visual Studio 6 and broke things and fixed them and broke them again until the patterns lived in my hands, not just my head. The understanding was physical, earned through repetition and failure. That foundation is what makes me useful to AI tools now. I know what to ask for because I spent decades learning what matters.

What happens to the person who never builds that foundation?

This is the part that concerns me most, and it's not about engineers specifically. It's about everyone.

When the premium model is available and the tokens are flowing, people perform at a high level. The work gets done. The output is polished. But what happens when you run out of requests for the day? What happens when the model is down, or the subscription lapses, or the company decides the cost is too high?

Because the cost is real. These models are expensive to run. The good ones, the ones that produce the results people are building their workflows around, are not cheap. Companies are pushing adoption while also watching the bill climb. There's a tension there that nobody is talking about openly. And on the individual level, people are developing a dependency on tools they don't control and can't afford indefinitely.

I've seen it in small ways already. The moment someone hits a token limit and their productivity drops. The frustration isn't "I have to do this the slow way now." The frustration is closer to "I don't know how to do this without the tool." That's a different thing entirely. That's not inconvenience. That's fragility.

There's another side to this that I think about on my walks. Everyone can build now. That sounds like good news, and in many ways it is. It's the reason I was able to ship Pura Letra as a solo developer. It's the reason someone with a good idea and basic technical sense can prototype something in a weekend that would have required a team a few years ago.

But more creation doesn't mean better creation. We've seen this pattern before. Streaming services made it easy to produce and distribute content at scale. The result wasn't more great television. It was an ocean of mediocre content that made the great stuff harder to find. The same thing is starting to happen with software, with writing, with design. The barrier to making something dropped, but the barrier to making something good didn't move.

I notice it already. The apps that clearly came from a weekend prompt session. The blog posts that read like they were generated in one pass and published without a second thought. The pitch decks that are polished on the surface and hollow underneath. There's a sameness creeping in, and it's the sameness of output that was never shaped by a human who cared enough to revise it.

I don't have a resolution for any of this. I'm not going to tell you AI is good or bad, that it will save us or destroy us. I use it every day and I'll use it again tomorrow. It has changed what's possible for me, personally, in ways I couldn't have imagined five years ago.

But I also know that the things I value most about my work, the judgment, the instinct for what matters, the ability to look at a system and know where it's fragile, none of that came from a model. It came from twenty-five years of building things the slow way. And I worry about a version of the future where fewer people get the chance to build that foundation because the tools made it feel unnecessary.

The younger generation concerns me the most. Not because they're less capable. Because they're entering a world where the shortcut is the default, where the tool is the first option instead of the last resort, and where the patience required to truly learn something might feel like a waste of time when the machine can do it for you in seconds. They might be right. I hope they're not.

The hills near my house don't get easier. That's the point. Every walk is the same incline, the same burn, the same moment where I want to slow down and don't. The clouds at the top don't care about my token usage or my productivity metrics. The air is just air. My legs are just tired.

I think the reason I keep walking is because it's the one part of my day where I'm not augmented. Not optimized. Not leveraged. Just a person moving through the world, thinking at the speed of thought, arriving at whatever I arrive at on my own. It's slower. It's harder. And something about that still matters to me.

I'm not ready to let that go.

This is essay twelve of Building My Way Out, a weekly series about one engineer's attempt to build a life beyond employment. New essays every Friday. If you're not subscribed, you can sign up here.