[DigitalToday reporter Chi-gyu Hwang] "Do not start another coding loop while AI is coding. You can burn out. Instead, it is better to read, propose pair work to a colleague, and try writing yourself."

Scott Basson (스캇 배슨), a staff software engineer at media platform Medium, shared those thoughts on Medium about whether developers should use productivity gained from AI for self-development. He said his experience shows that using AI to assign more work while it is working may lift productivity in the short term but can later lead to being overwhelmed by work.

He said productivity rose sharply after he introduced AI tools into his work. He finished in a few hours what used to take days. But being able to work fast was not unconditionally good. He compared himself to the cartoon character Flash. He said it was like getting superpowers and running around excitedly, but without direction.

Basson said, "As my productivity increased fivefold, I took on five times more work. I stepped up and said I would handle everything, from bug reports to a new settings page to admin tools, but the result was burnout." He added, "I reached burnout 10 times faster than before. Code reviews got sloppy, pull requests were left for weeks. Context kept disappearing."

In that situation, Basson changed direction. Instead of doing more work, he took another look at how he worked. The conclusion was simple. He said the moments he felt most satisfied were when he learned something or helped others learn.

He said, "While an AI agent creates sub-tasks on its own and stacks up pull requests, I decided to make different choices instead of starting yet another work loop. Reading, proposing pair work to colleagues, and writing myself."

Basson said he had spent more than 15 years in web development and was used to skimming technical writing. He said he would just find the code he needed, read only the conclusion and click library links. Before AI, writing code itself took a long time, leaving no time to calmly read others' writing. He said, "What I missed in that process was the human side of the industry, like people who write in an engaging way, overlapping experiences, and different journeys."

These days, Basson said he reads documentation much more carefully. Rather than grabbing only code snippets on how React View Transition works, he said he is trying to understand the principles first. That reading also changed how he gives instructions to AI agents. He said he can give an agent instructions only after properly figuring out how far a feature is needed. The activity he said was most valuable was writing. He said, "I write a work journal almost every day for the company's internal media. It was something I could not even consider before because I thought 15 minutes was too much."

The response was unexpected. He said, "Colleagues read the posts and left comments, and it led to real conversations. Deliverables like a content calendar and a new editor page all started from ideas that came out that way." He added, "I could transfer the productivity gained from AI into something useful for actual users. It is worth the token cost."

He also said, "I realised as I slowed down and put my thoughts into writing. The more carefully you work, the more productivity rises in the long term." He added, "Now I do not make 10 times more than before. But what I make now is 10 times more meaningful."