Published a new article! 😺
I wrote about why CHANGELOG.md still matters in the AI era, and how changelog-bot helps automate it.
It supports Git history, PRs, GitHub Release Notes, OpenAI, Anthropic, Gemini, and fallback generation without AI.
https://t.co/9qjl8ii7a2
Published a new article! 😺
I wrote about why CHANGELOG.md still matters in the AI era, and how changelog-bot helps automate it.
It supports Git history, PRs, GitHub Release Notes, OpenAI, Anthropic, Gemini, and fallback generation without AI.
https://t.co/9qjl8ii7a2
CHANGELOG.md is boring.
But it's surprisingly important.
* users read it
* maintainers read it
* AI agents read it
When I'm triaging Dependabot updates with AI, I often feed CHANGELOG.md into the context.
The problem?
Most projects stop maintaining it.
I don’t think the real risk of AI in software is “humans vs AI.”
The real risk is humans using AI to avoid responsibility.
Avoiding design decisions.
Avoiding trade-off discussions.
Avoiding long-term thinking.
AI is powerful, but it amplifies our habits.
@barkin_cat I think so too.
If AI costs increase over time, blindly relying on it may become harder.
So it’ll probably become even more important to have solid fundamentals and use AI intentionally.
In the end, I don’t think the basics ever disappear.
I’m not against using AI for software development.
But I’m worried that AI can make old anti-patterns look new.
- Shipping fast without clear design.
- Accepting bugs as a cost of speed.
- Assuming future tools will clean up today’s mess.
We have seen this pattern before.
The scary part of AI-generated code is not only that it can be wrong.
The scarier part is it look clean while hiding poor decisions.
Good names.
Nice comments.
Reasonable structure.
Maybe even types.
Broken code used to look broken.
Now it can look polished.
AI is not magic.
If we skip design decisions and use AI only to move faster, we may just create technical debt at a much higher speed.
Before, it was:
“We’ll fix it later because deadlines are tight.”
Now, it is:
“We’ll fix it later because future models will be better.”
Thanks for featuring my article! 😺
I compared openapi-typescript, hey-api, Orval, and Kubb from a frontend TypeScript perspective.
https://t.co/3MN2K0W45l
AI can speed up implementation.
But it does not automatically take responsibility for design, maintainability, or long-term thinking.
We should delegate tasks to AI, not design responsibility.