We Are the QA for AI

We Are the QA for AI In my last post I wrote about Anthropic’s postmortem and how AI bugs do not look like the bugs we are used to. They hide in the answers. They are not obvious crashes. That leads to this. In the past if you wanted to complain or suggest a feature you went to a support page. Maybe you submitted a ticket. Maybe you took a screenshot of the bug and attached it. Then you waited. Maybe someone read it. Maybe they did not. AI is different. The thumbs up and thumbs down buttons are right there in front of you. No forms. No support tickets. Just a quick signal. And that signal matters. It is the easiest way to shape how AI learns and how reliable it becomes. This is still new. Most people do not think of themselves as part of QA. But with AI we all are. Every tap is a way to make the system better. I am an advocate for this. If you want better AI at work and in life, smash those buttons. **Labels:** AI Feedback, AI Reliability, Human Feedback, AI QA, Responsible AI

Bugs in AI Do Not Look Like the Bugs We Are Used To

Bugs in AI Do Not Look Like the Bugs We Are Used To When software would break in the past it was often obvious. The screen froze. The app crashed. You knew you had a bug. AI is different. When it breaks the system keeps going. It still gives an answer. But something is off. A word out of place. A weaker response. A sentence that does not sound right. Anthropic’s postmortem showed this. Claude slipped Thai characters into English text. That is not a crash. That is a bug hiding inside an answer. This matters. The risk is not that AI stops working. The risk is that it keeps working while giving you something that is just a little wrong. And at work a little wrong can be a big problem. This is why the like and dislike buttons matter. They look small. But they are signals. They are how the system learns what can be trusted and what cannot. Smashing those buttons is part of keeping the AI useful. The old bugs broke software. The new bugs break trust. **Labels:** Anthropic, Claude, AI Reliability, AI Bugs, AI Trust