Your agent is building things you'll never use

I built more in two months with agents than in the previous year. I used almost none of it.

That's the confession nobody in the AI-productivity crowd wants to make. Building feels like progress. The agent churns, the code appears, and you get a hit of satisfaction. But satisfaction isn't value. And building isn't shipping.

The problem is treating agents like strategy engines when they're execution engines. Its really why we are getting a ton of push back in open source, people writing and reading about agents in code, and even use.

Point an agent at a vague goal—"build me a tool that helps with X"—and you'll get something that looks impressive and rots in a folder. Point an agent at a specific task—"rewrite these 200 API calls to use the new authentication pattern"—and you'll save a week.

One is generative theatre. The other is actual leverage.

The difference is tactical versus strategic deployment.

Tactical deployment means applying agents to problems you already understand. The data transformation you've done manually a dozen times. The test coverage you know exactly how to write. The refactor where the pattern is clear but the labor is mind-numbing.

Strategic deployment means pointing agents at fuzzy goals and hoping they produce something useful. This almost never works. You get slop that demos well and decays fast.

The fundamental lesson: agents amplify clarity. If you know exactly what you want, they're a force multiplier. If you don't, they're an expensive way to generate garbage you'll never use.

I call this the Build-Use Gap. Agents have collapsed the cost of building while the cost of using remains unchanged. Integration, polish, workflow fit—these still require human judgment and sustained attention. When you build faster than you can use, you accumulate digital debt that never gets paid.

The tactical move: reserve agents for tasks you already know how to do that consume disproportionate time. The migration you've mentally mapped but dread executing. The boilerplate you could write in your sleep. The transformation that's clear but tedious.

Stop using agents to explore what's possible. Start using them to execute what's necessary.

The productivity gain isn't in what you build. It's in what you actually use.