Simon Willison: Coding agents are making software engineering harder

Summary

Simon Willison acknowledges in a brief note that coding agents can achieve impressive results. However, according to Willison, reaching their full potential requires extraordinary discipline and deep knowledge, which paradoxically makes software engineering more difficult.

Why it matters

Willison is one of the most followed independent voices in the AI and large language model (LLM) space. Therefore, positioning agents as difficulty multipliers rather than simplifiers presents a stance that contradicts the narrative that 'AI will make coding easier.' This approach highlights that even passionate practitioners hit a ceiling where the tool's power is limited by the operator's expertise. This point is critically important for teams investing heavily in agentic workflows.

Highlights

• According to Willison's growing belief, coding agents do not simplify software engineering; instead, they make it harder.
• Unlocking an agent's full potential requires 'extraordinary discipline and knowledge.'
• The note is framed as a personal observation from practical use, not a systematic study.

Critical take

This 'article' consists of only two sentences. There is no supporting evidence, concrete examples, distinction between agent types, or explanation of *why* the difficulty arises. This makes the text feel more like a headline with a subheading than an argument. A reader seeking actionable insights or the mechanism behind the claim finds nothing beyond a strong opinion.