Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.
Soft Skills Engineering · Hands-On Engineering Podcasts · October 2026
Dance applies the Gell-Mann amnesia effect to reviewing AI-written code. You see how wrong the output is where you know the code well, then trust it where you don't. He says it's a reason developers still need to know the code well enough to have opinions about what the model produces.
…effect. I was furiously googling to remember this. This is the name for the situation where you read headlines about a subject you know very well and you realize just how utterly wrong it is in many areas and then read other headlines about something you don't know and think, huh, how strange.
That's probably right. That's
the way it is. Yeah. And I think there's something like that with the code as well. The less I know the code, the more I say, yeah, it's probably fine. And the more I know it, the more I have that interaction of, no, this will be bad for us. We do not want it that way. I don't know. There's probably some capability increases that maybe make that not matter. But it is, I'm convincing myself still that it's useful to know and be able to have opinions about the output because sometimes it seems bad.
You're convincing yourself that you have a moat around you.
Yes. I am. We're dog fooding. Our own product. Yeah, at all levels.
Yeah, I'm sure you have a moat. I'm sure you paid the fee.