Coding agents make software engineering harder, not easier, argues blog note
A brief blog note dated 24 September 2026 pushes back against the common narrative that AI coding agents simplify software development. The author writes: "The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder." The note does not name a specific agent, tool, or product, nor does it explain the mechanism behind the claim, it states the conclusion without supporting examples or evidence.
The author does not dismiss coding agents outright. The same note adds a qualifier: "We can do amazing things with them, but unlocking their full potential requires extraordinary discipline and knowledge." In other words, coding agents can produce impressive results, but only for practitioners who already bring rigor and expertise to the work, for everyone else, the tools add difficulty rather than removing it. The note is short and offers no further elaboration, statistics, or named examples beyond these two claims.
Key facts
- A blog note dated 24 September 2026 argues coding agents make software engineering 'even harder' the more one uses them
- The author simultaneously concedes that 'amazing things' are achievable with coding agents
- Realizing that potential, per the note, requires 'extraordinary discipline and knowledge'
- No specific coding agent, tool, or example is cited as evidence for either claim
- The note gives no mechanism or explanation for why coding agents increase difficulty
Why it matters
Coding agents are widely marketed as a way to make software development faster and easier. This note takes the opposite position, that heavy use of coding agents can make the underlying engineering work harder, unless the user already has strong discipline and knowledge. It is a contrarian data point against the prevailing hype around AI coding tools.
Who it affects
Software engineers and teams who are adopting or considering AI coding agents in their daily workflow, as well as anyone evaluating claims about how much coding agents actually reduce engineering effort.
How to use it
The note reads as a caution rather than a guide: it suggests that adopting coding agents should not be expected to lower the bar for expertise. Readers who want to get value from these tools should treat the note as encouragement to build up discipline and domain knowledge alongside agent use, rather than substituting one for the other.
How solid is it
The note is extremely short, two sentences of argument, and offers no named agents, no data, and no explanation of why coding agents make engineering harder. It is a personal opinion asserted without supporting evidence, so it should be read as one practitioner's impression rather than a demonstrated finding.
Risks and caveats
The central claim is stated without a mechanism or examples, so it cannot be independently checked from the note alone. Readers should be wary of generalizing a single, unelaborated opinion into a broader verdict on coding agents as a category.
“The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder.”
— the author of the note