Post by @danielpunkass@mastodon.social

2026-04-15 03:34 UTC

AI skeptics often express a fear about how it might do something bad without you knowing. You can&#39;t &quot;trust&quot; it. It&#39;s absolutely, 100% true. But the question of whether you trust it or not is entirely in your hands. You CAN review every word, code snippet, etc., before it does it. If you choose to.

Replies (2)

@jon_alper@mastodon.online 2026-04-15 03:57

@danielpunkass@mastodon.social The problem isn’t when someone capable of verifying its output does. The problem is that more and more, it’s used by people incapable of and or unwilling to verify the output fail to in ways that cause harm.

Open ##3812189
@arroz@mastodon.social 2026-04-15 08:00

@danielpunkass@mastodon.social That’s not as clear as it seems. Humans are notoriously worse at verifying and monitoring things than doing them. That’s one reason code reviews aren’t worth much especially in large PRs. If you didn’t think through every little detail that resulted in a code line, you’re always implicitly trusting a lot of assumptions implied in the output. And you’re also missing those “wait, this could be better” moments when you have all the context in your head at the code abstraction level.

Open ##3812195