A problem for people who use AI
Which design to choose once implementation moves ahead
One developer who keeps using AI in work and personal development and has built up their own working environment
Person chose the small option; comparison of effect unconfirmed
The present time in the text is October 2026. The absolute posting date is unconfirmed, and we have not back-calculated a date.
Editorial check: 2026-10-23
What people are paying attention to
They say their interest and burden have shifted from small implementations to design priorities and complexity.
What is hard
Even when a working configuration is produced it can be too large, and they find it burdensome that the way of asking changes the AI's judgment.
Workarounds being tried
For one design, they explain that they chose the simplest option and implemented it small.
What the sources confirm
A self-report of a practice that chose the small option. Later maintenance effort and comparison with other options are unknown.
What is still unclear
How to leave decision conditions so they help consistent design is untested. We check the situation of the target people as an issue separate from accepting implementations.
Workarounds and results in more detail
Evaluate the AI's proposed configurations and narrow to the simplest option
The person's report of what they did
Implemented small by the person. The effect on decision time or maintainability is unmeasured.
Scope and limits: One person's retrospective. It is experience with a prepared environment and mainly small tasks.
Read with care
- From one person's experience, we cannot estimate changes in challenges for developers overall or market size.
- Use of external support, budget and willingness to pay are unconfirmed.
What to check next
We follow the option actually chosen, the conditions for adopting or rejecting, the decision record, and the later burden of change and maintenance.
Even for people who can hand off implementation, the burden of choosing the design remains
The poster, who has used AI for in-house tools and personal development, looks back that in an environment where they had arranged the flow of execution and retention of context themselves, small implementations and unit tests were no longer the main problem. In consultations on production design, however, proposals grew too large and recommendations changed when the question was phrased differently, so the judgment of priorities remained. For one design, they chose the simplest option and implemented it small.
A replier proposes separating the stage of exploring options from the stage of recording the decision, and leaving the judgment in decision records, tests and the like, but we cannot confirm that the original poster adopted it. Editorially, we need to ask separately about the correctness of implementation and about choices that fit the purpose and maintenance burden. It cannot be generalized that the same change happened to all developers.
Questions to think with
- For a design you were recently unsure about, can you explain the business conditions you prioritized, the complexity you accepted, and who decided?
- When the AI returns to another option after a decision, in what record or verification can the reasons for adopting or rejecting be referred to?
A retrospective by one long-term user and a proposal by another replier. The confirming questions are editorial suggestions.