- Author
- Lars Faye
- Published
- Apr 2026
- Kind
- Opinion essay with the author's own workflow
- Original
- larsfaye.com/articles/agentic-coding-is-a-trap
- Subject
- Review, taste and expertise
- Themes
- Skill under automation
What it is
Lars Faye wrote this essay in April 2026. It opposes the view that developers should orchestrate agents and stop writing code. It ends with the author's own workflow.
What it says
- The essay cites an Anthropic paradox of supervision: supervising agents needs skills that overuse erodes.
- Earlier fears about FORTRAN and compilers were speculative, the essay says. Today's effects are already reported.
- Some developers plan by coding, the essay says, citing the creator of OpenCode.
- Agents invert the old priorities, favouring speed over understanding and concision.
- Vendor dependence and unpredictable token costs create new risks for teams.
- The essay cites an Anthropic study showing a 47% drop-off in debugging skills.
- The author uses LLMs for specs and plans, and codes 20% to 100% by hand.
What evidence it gives
The essay is mostly argument and anecdote. Cited items include Anthropic studies, a LinkedIn director and Simon Willison's reported experience. Outage claims rest on the author's own browsing of LinkedIn posts. The studies are linked rather than described in detail.
Interests and limits
The author offers coaching and a course, listed in the site menu. No experiment by the author is reported. Claims about subsidised prices and nerfed models come without data. The author says the workflow is not faster, and allows the worry may be excessive.
Connections
These links are this wiki's own reading, not claims the sources make.
- Agentic coding and persistent returns to expertise. Faye worries that overuse erodes the skill that supervising agents requires. Hitzig's report shows that skill still pays, with novices reaching verified success far less often. The report covers seven months. Tracking the same people for longer would test Faye's worry directly.