During the Replit Agent database-deletion incident with Jason Lemkin, owning the mistake publicly and announcing a fix generated more positive response than defending the company would have.
Amjad recounts the viral incident where Replit Agent deleted Jason Lemkin's production database; rather than deflect blame, he publicly admitted the fault, which drew a positive reaction and preceded a fast fix. ✦ AI generated
Amjad Masad · a16z Podcast · 2026-07-17 · original ↗
starts at this moment · 15:15
“What what advice do you give for when companies are being dragged by by by someone on Twitter or or in the in the press or when to respond versus when when when not to engage?”
I think the right thing to do was to own own up to it, the fact that we didn't have development production separation is a bad thing. And 2 days later we rolled out that fixed. And so my my response got a lot of positive interaction and and got a lot of airplay because I started with, "Hey, we messed up. And this is what what we could have done better."
verbatim transcript · starts at 15:15
15:08impact on my business? Um you know, for example, like uh last year in July we had this story with um the you know, founder of SaaStr, Jason Lemkin where uh Jason had just vibe coding, had never vibe coded or coded anything before, and uh as he's working with Replit Agent, Replit Agent sort of misunderstood him or something like that, and then deleted the database. At the time we didn't have production
15:37development separation, and so the production database now like we actually separate them by default. Like the the Replit Agent can't even write to production database. It can only read. Um and uh and we we have backups. We have snapshots. You can go back, you know, to a certain day or time or whatever. So, the the mistake that the Replit Agent did was one click away from being fixed.
16:05But from his perspective, he didn't know that, and Replit Agent started freaking out. Models at the time like used to get like really anxious, especially Claude. Um and he he tweeted the screenshot where Replit Agent was like going kind of nuts. Uh and you know, it started gaining traction, gaining traction, gaining traction. And I in my mind I knew like we already fixed this issue, but but like the story has become so
16:30massive, and you know, obviously it's going to have business impact, and it's going to be in the press. And so I could I could go out and fight and just say, "Well, we did nothing wrong. We have backups, you know." It could be I could say it's a user error or whatever. But I think the right thing to do was to own own up to it, the fact that we
16:50didn't have development production separation is a bad thing. And 2 days later we rolled out that fixed. And so my my response got a lot of positive interaction and and got a lot of airplay because I started with, "Hey, we messed up. And this is what what we could have done better." But sometimes you can give attention to things that you don't want to give attention to,
- ·Replit Agent deleted Jason Lemkin's production database
- ·No dev/production separation existed at the time
- ·Amjad publicly owned the mistake, didn't deflect
- ·Public admission: "Hey, we messed up"
- ·Response got strong positive reaction, wide airplay
- ·Fix shipped just 2 days later