Agents are now an audience you must design for and defend against

A Brazilian labor court fined two lawyers for hiding a prompt-injection command in a filing aimed at the court's AI review system. The text told the AI to respond superficially and not challenge the documents. In a separate post on the AI cost of JavaScript frameworks, Astro came out with the best time, the second-lowest spend and zero client-side JS, while Angular did badly for agents. HN is also debating a flag for AI-generated articles, with the guidelines now saying not to post generated or AI-edited text, and Cliff Stoll's "Rumors of my death" post was about AI slop writing his obituary.

The pattern: AI reads, judges and writes in places that used to be human-only, and every one of those places gets gamed. Courts use it to review filings, so people inject. Agents write code, so framework choice becomes a cost line. Social feeds fill with generated text, so communities ask for flags.

The debate is about enforcement. Commenters note AI detectors are unreliable, voting can be gamed, and heavy stigma drives away people who use LLMs lightly.