← sources

We are being overtaken by a staggering mental regression

source: reddit (reddit:AI_Agents)published 2026-11-07 14:48ingested 2026-11-08 02:35

retained text

Read the news lately? Imagine buying a fridge and finding out you can’t turn the Wi-Fi off. Or finding out your car's hood won’t open because the battery you’re trying to get to is dead. Shall I mention trains running at 300KMH not having autonomous radar systems? And what about automatic doors with no manual override?

You would ask what idiot designed it.

Now imagine a customer service agent decides you are committing refund fraud and freezes your account. You click APPEAL, and the bot looks at your account, sees the fraud flag it put there five minutes ago, and tells you the decision is final.

At what point in that story was someone supposed to say, 'Wait a min, why are we letting the bot do that?'

I want AI to save me three hours of paperwork. I don’t want to spend those three hours trying to find the one person who can tell it to knock it off.

Engineering an agent means thinking through what can go wrong and deciding up front which calls it gets to make. If you leave every foreseeable failure for the model to sort out in production, you have delegated the engineering too. And I'm sorry, but that's really dumb.

Worst thing? I see this daily with big corps product's workflows, which brings me back to the initial real life examples. Why dumb people get hired to engineer things that could carry dangerous outcomes?

document fetched 2026-11-08 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xq838h/we_are_being_overtaken_by_a_staggering_mental