AI, a year in

I stopped trusting AI summaries after this one mistake. The summary looked clean. It was confidently wrong about the one number that mattered, and I almost sent it. Now I read the source every time, and the time tax is worth it.

I'm not claiming this is true for everyone. It's true for me, and the version I'd tell a friend is shorter than the marketing version.

The pushback I get is always the same, and it's usually right at the edges, not the center. The center holds: the boring practice beats the clever one most days.

If you take one thing from this, let it be the unglamorous part. The flashy bit is what gets demos; the plain bit is what gets used.