Cognitive distortions, review, and human judgment

The day's work was more about judgment than volume. From notes on agents, code review, and engineering habits, I wrote a post about cognitive distortions for engineers.

Thinking before the tool

The post came from a practical concern: using agents without outsourcing judgment. Tools can speed up reading, implementation, and review, but they can also amplify overconfidence, selective confirmation, and the rush toward a plausible answer.

The main idea was to treat cognitive distortions as process failures. Not as an abstract behavior topic, but as something that appears in technical decisions: reviewing a diff, accepting a suggestion, choosing an architecture, or ignoring risk signals.

Code review with more context

I also reviewed material on how to evaluate code with a more mature posture. The question that stayed with me was: is this really a bug, acceptable debt, a style difference, or a decision that needs more context?

That kind of review needs less impulse and more sequence. First understand the intent. Then look at architecture, data flow, risks, and project patterns. Only after that should syntax and local details come in.

Bilingual publishing

The result became a bilingual pair with its own cover. I kept the Portuguese and English versions mirrored, with the same argument and the same conclusion: agents are more useful when the engineer keeps responsibility for judgment.

It was a smaller day in volume, but an important one in direction. Not every improvement to the site is a feature. Sometimes it is a clearer thesis about how I want to work.