Do I Need to Be a Hundred Times More Efficient?

Lately, I have been thinking about a question: if I do not have a software project that I am rapidly iterating on, do I still need to pursue tenfold, hundredfold, or even thousandfold gains in efficiency, as other engineers do?

When I see people share how they use AI to write code, it is easy to feel that I should learn their methods too. How to manage context, how to get several agents—AI assistants capable of carrying out tasks—to work together, how to connect testing with code review: every one of these practices makes sense. But once I have learned them, do I actually have anything at hand that calls for them?

For people who maintain projects over time, these methods often grow out of specific problems. As the codebase expands, changes become more frequent, and some part of the process begins to slow them down, they invest time in fixing it. An improvement can be used repeatedly in the work that follows, making it easier to recoup the investment.

Writing about what they have learned also helps them clarify their own thinking. Later projects continue to test those lessons, while writing preserves them. This kind of accumulation can indeed compound over time. But behind it is an ongoing project, with real problems and feedback.

The people reading those articles may not be in the same position. I can understand a method and configure it for myself, but if I have no opportunity to use it afterward, it is hard to know what makes it valuable, when it is worth using, or where it needs adjustment. A few days later, a new tool appears, bringing a new learning task with it. I may have spent a great deal of time without being able to say what any of it helped me accomplish.

Of course, the need is not always clear before we try something. Experimenting with a new tool sometimes reveals that something once impossible is now possible, and a project emerges from that discovery. There is also nothing wrong with exploring out of curiosity. But that possibility does not mean every update needs to be followed.

Learning can begin with two questions: What matters most to me right now? How well do I want to do it?

Even without a rapidly evolving software project, I can still keep using AI. Writing an article, understanding a technical issue, or preparing for a client conversation all offer concrete uses. When the goal is clearer, it becomes easier to tell what is worth learning and what can wait.

If I need to write an article now, I can start by discussing the ideas with AI, organizing the material, and producing a first draft. Wherever I find that I cannot explain something clearly, I can research it, fill in the gaps in my knowledge, and revise the wording. If a particular step repeatedly holds me up, then I can look into better tools and methods.

This way, I can see whether what I have learned is useful: can I make what was once vague clear, and did I finish the article? As for the AI coding workflows other people are using, I can wait to study them until I have a project that calls for them.