Moritz Sudhof
speaker
399 appearances
1 recordings
1 series
first heard Jul 2026
last heard 8 Jul
Moritz Sudhof’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.
Appearances
And I think the danger there is exactly what you said for most of the things that aren't straightforward code execution or f straightforward tasks, for most of the things that are actually require judgment, creativity, expertise, the things that we really want AI to help with.
For most of those situations, we just don't know.
At the start, exactly everything that what good looks like.
We don't know all of our requirements.
It's only once we start engaging with the work product that we realize, oh no, but I actually I want this the tone of this email to be a little more warmer and more casual because that's my voice.
Or I forgot to mention, yeah, the other PM said that we needed this, or you first time you see something in action, you realize, like, oh, that's a really bad user experience.
We don't want to do it that way.
And so it's how do
do we design products and design our own usage of AI so that we iterate more and
iterate with lower friction, iterate more easily so that we can hill climb towards the thing that is actually our deliverable.
And I think not not get in a world where we just try to have AI working for us in the background for twelve hours and then we hope to review the results.
Because a lot of the times there's going to be a lot of things about that output that aren't actually perfect.
They don't actually satisfy your requirements and you're just going to go back to the drawing board anyway.
Yeah, I think it's a real danger to fly blind.
The research shows and look at actual product metrics, around three percent of users, if you're lucky, give you feedback.
Um, that means they're pushing thumbs up or thumbs down or they're escalating.
Um, most users do not give you feedback.
And most of the failures in AI and ninety nine percent of the failures that we've discussed in this podcast will never show up in a dashboard.
There will never be a dashboard that has a CSAT number or has a um a failures number or a regressions number that catches these.
Because these aren't technical failures.
Showing 321–340 of 399 · page 17 of 20
← Previous
Next →