Back to blog
18 November 2026Nicolas Dabène — Développeur Full Stack & Orchestrateur IA chez Profileo , 77-24 e-commerce hosting et Zentria6 min

With AI, I Develop Faster. I Also Had to Learn Not to Burn Out Faster

Intelligence artificielleAgents IAAPILLM & modelesPédagogieSecuriteSouveraineté numérique
Contents
Developer reflecting on sustainable workflows while integrating AI tools into daily coding tasks

Less than two years ago, a day of development was already a mix of coding, problem-solving, and technical decisions. Some architectural questions kept me busy for several days. Today, with AI, I can explore multiple topics and make several important decisions in the same day.

RSS

RSS sends new articles to the reader of your choice, without an algorithm or newsletter.

Open RSS feed

I move faster. I can build projects that would have been much harder to complete alone. And I still enjoy developing just as much. But I’ve also discovered a less-discussed consequence: by accelerating the production pace, I sometimes forget that my attention has its limits.

What I share here is my personal experience. This is neither a study on developers nor a critique of AI. It’s the observation I’ve made from using it intensively on a daily basis, and how I’m learning to adapt my way of working.

Development Still Demands Decisions

When talking about AI and development, the focus is often on the time saved from writing code. That’s true in my practice too. I spend less time on repetitive tasks and can delve deeper into topics that interest me.

But the work doesn’t disappear just because I’m not writing as many lines of code myself. It shifts toward framing, technical choices, verification, and understanding what AI agents produce.

Before, I could spend a significant portion of my day implementing a solution. This phase also left me time to think about the problem incrementally. With AI, implementation can move much faster. In return, I’m forced to chain decisions more quickly: Does this approach hold up? Is it consistent with the existing architecture? What are the security risks? What needs to be verified before moving forward?

These questions already existed. What changed for me is their frequency and the speed at which they come. A single day can now concentrate several technical topics that, in the past, would have been spread over a longer period.

This isn’t necessarily negative. I enjoy solving problems and making architectural choices. But even when the work is exciting, chaining decisions requires energy.

When Waiting Time Disappears

I’ve noticed another habit: as soon as the AI is working, I start something else.

On an important topic, I might request an application security analysis or take time to adjust an architecture. These steps matter to me: the time saved thanks to AI also allows me to delve deeper into points I don’t want to handle lightly.

But while an analysis runs, I usually don’t wait. I open another topic, start another task, or return to an ongoing problem. Then I go back to the first analysis’s results before switching to something else again.

At the moment, this organization gives an impression of efficiency—and it can be. Multiple topics advance. But each return requires me to recontextualize, remember the decisions made, and understand what deserves my attention. With experience, I keep track better than before. The effort of switching contexts, however, remains very real.

Waiting time could sometimes have been a break. In my case, it almost always turns into additional work. I haven’t just sped up certain tasks—I’ve also filled the gaps between them.

The Vicious Circle of Prompts and Fatigue

Fatigue didn’t hit me all at once. I noticed it gradually, especially in how I interact with AI agents.

When I’m tired, I become less patient. My instructions can become shorter, less precise, or less clear. The AI then has a weaker framework, and its responses require more corrections. I end up going back and forth to clarify what I could have explained from the start.

And the longer these exchanges go on, the more frustration can build. I keep insisting even when I no longer have the necessary perspective to properly frame the problem. The result is that I end up spending even more energy on a task I thought I was speeding up.

With experience, I’ve started recognizing these signals: corrections that chain without real progress, growing impatience, increasingly vague instructions, or difficulty calmly evaluating a proposal. Today, when I spot them, I know that a pause and a fresh reformulation are often more useful than another prompt.

I’m still trying to intervene earlier. Recognizing the problem once you’re already in it is helpful, but learning not to wait for that moment remains a work in progress.

Verifying Without Treating Everything the Same

One might think this fatigue mainly comes from the fear of missing an error. In my case, that’s not the biggest concern. I’m comfortable with the idea that validation is part of my job, and I use tests, tools, or other agents to help.

However, I don’t verify everything at the same level. A prototype meant to explore an idea doesn’t carry the same stakes as a sensitive feature or a choice that impacts a project’s architecture. On these topics, I dedicate more time to analysis and adjustments.

AI can help me produce and examine solutions, but I must always understand what I’m trying to validate and adapt my attention to the risks. Even with good tools, this requires staying mentally available to make decisions and keep an overview.

Fatigue Doesn’t Always End at the End of the Day

This pace also has consequences outside of development. Some days, I end with less mental availability for my loved ones or hobbies. Other times, when I still have energy, I want to keep exploring ideas or working on personal projects.

This also makes boundaries harder to see: I don’t continue just because I feel obligated. Sometimes, I continue because the topic excites me and I want to see how far I can go.

For the past few weeks, I’ve made more of an effort to disconnect to preserve my well-being. It’s not a perfectly ingrained habit or a magic solution. It’s an awareness: if I want to keep enjoying work and have energy for the rest of my life, I can’t fill every available moment with a new task.

What I’m Trying to Change

I don’t have a universal method to offer. However, I’ve identified several things I want to work toward in my practice: taking a break when the first signs of fatigue appear, limiting the number of parallel topics, accepting not to start anything during certain waiting times, and setting a real end to my workday.

Some of these habits are starting to come naturally with experience. Others still require effort. I still sometimes want to use an ongoing analysis to move forward on something else, or extend a workday because an idea sparks my interest.

The goal isn’t to slow down systematically or give up on the possibilities offered by AI agents. It’s to choose when parallel work truly adds value and when it consumes attention I’ll need later.

Continuing to Build Without Forgetting the Human Pace

I’m still convinced of AI’s value for development. It helps me learn, experiment, reduce repetitive tasks, and turn ideas into reality faster than I could alone. It gives me more autonomy and allows me to tackle more ambitious projects. These are gains I hold dear.

But my experience has taught me to look beyond the number of completed tasks. A highly productive day can also be one where I’ve chained so many decisions, analyses, and context switches that I have little left once I close my laptop.

AI can speed up part of our work. It doesn’t make our attention infinite. To keep benefiting from what it offers, I’m learning to consider breaks, limits, and the right to have free time as part of my development process.

Nicolas Dabène

Author

Nicolas Dabène

Développeur Full Stack & Orchestrateur IA chez Profileo , 77-24 e-commerce hosting et Zentria

Senior PHP/Laravel developer with 12+ years of experience in e-commerce. Specialised in PrestaShop architecture, AI agents and automation.

RSS

Follow this blog

RSS sends new articles to the reader of your choice, without an algorithm or newsletter.

I want a simple setup

Choose a reader, add the feed, then every new article appears there automatically.

  1. 1. Pick one of the readers below.
  2. 2. Open it and add this feed URL.
  3. 3. Read the next articles from one place.