Replit's CEO says his company "entirely changed how we operate". The question that asks whether you can say the same about your own work.
On October 16, 2026, TechCrunch reported on a talk Replit CEO and co-founder Amjad Masad gave at TechCrunch Disrupt in San Francisco, where he described what he called a "self-driving company". The phrase, TechCrunch noted, refers not to robotaxi makers but to a workplace where agents do much of the "grunt work" while people are valued for their "judgement and taste". Masad said his team had spent the past year reflecting on how AI had changed its own operations, and summed up the result bluntly: "Our company entirely changed how we operate". His example was support. By his account, a ticket at Replit is "first examined by an agent, which performs the initial triaging by looking at GitHub issues, the codebase, and Linear tickets", and only then reaches a person, who decides what to do and can deploy another agent to ship the fix. "You need human judgment; you need taste," he told the audience. All of this is Masad's own description of his company, delivered from a conference stage, and TechCrunch reported it as such.
Set aside the forecast and look at the behaviour underneath it. A company looked hard at how it had actually been working, decided the old arrangement was wrong for the tools it now had, and rebuilt the loop so that the expensive part, judgment, sat in the middle. Masad said the visible effect was that people now spend "a lot more time figuring out what to work on", with "a lot more discussions, a lot more whiteboarding". That is a retrospective that produced a different way of operating, not a regret. It is also, almost exactly, the behaviour one of the most common senior interview questions is designed to surface.
Why they ask it
The question reads "Describe a project that you wish you had done better and how you would do it differently today", and it is asked because the interviewer cannot see a candidate's judgment directly. They can only see what the candidate does with evidence about their own past work. A manager who can say precisely where a project went sideways, and can describe the change they would make with the same precision, is showing the thing Masad was describing: the capacity to examine a working system honestly and redesign it. The question also tests calibration. Senior interviewers have all run projects that disappointed, and they can tell the difference between someone who has processed that experience and someone who has filed it away.
The trap
The weak answer picks a project that went fine and manufactures a small regret. "We shipped on time, but in hindsight I would have communicated more." That is not a project the candidate wishes they had done better. It is a project they are proud of, wearing a humility costume, and interviewers discount it on sight. The second failure is the opposite: a candidate names a genuine disaster and then narrates it as a sequence of things other people did. The team was understaffed, the requirements moved, the vendor slipped. Each may be true, yet the answer leaves no evidence that the candidate's own decisions shaped the outcome, which means there is nothing to learn from. The third failure is the vague remedy. "Today I would plan better" tells the interviewer nothing about what the candidate now knows that they did not know then. A strong answer names one decision that was theirs to make, shows why it seemed right at the time, and replaces it with a specific alternative that carries a cost of its own.
Applying STAR-T
Situation. Keep it short and make the stakes real. "I led the migration of our billing service out of the monolith into a set of services. It was a nine-month project with four engineers, and the goal was to let the payments team ship independently of the core release train."
Task. State what you owned, not what the team owned. "I owned the sequencing. I decided which capabilities moved first and how we would cut over, and I set the expectation with leadership that the migration would be invisible to customers."
Action. This is where the honesty lives. Describe the decision as it looked then, and the moment the evidence turned. "I chose to migrate the ledger first because it was the best-understood component. That meant running dual writes for months while everything else caught up. Around month five, reconciliation discrepancies started appearing weekly, and each one pulled two engineers off the migration to investigate. I treated them as bugs to fix rather than as a signal that the sequencing was wrong, and we kept going." Then the pivot to today, stated as concretely as the original choice. "Today I would move the read paths first and leave the ledger as the last piece, because the read paths were low-risk and would have given the team a working pattern before we touched the component where mistakes cost money. I would also set a tripwire up front: if reconciliation issues exceeded a fixed rate for two consecutive weeks, we would pause and re-plan rather than absorb the cost."
Result. Give the honest outcome, then the honest consequence. "We finished three months late. No customer was mischarged, but the payments team got its independence two quarters after I had promised it, and I spent the last stretch rebuilding trust with their director instead of designing the next thing."
Trade-off. This question rewards naming the cost of the better choice, because that is what separates a lesson from a slogan. "Migrating the ledger last would have meant the most valuable capability was the last one unlocked, and I would have had to defend a plan whose payoff arrived late. I think that was the right trade, and I did not see it then because I was optimizing for the part of the system I understood best rather than the part that could hurt us."
The follow-up that breaks weak answers
The interviewer waits, then asks: "What did you actually change on the next project?" The manufactured regret has no answer, because nothing was wrong. The blame-shifted disaster has no answer, because the fix lived with other people. The vague remedy collapses into restating itself. The candidate who genuinely processed the experience can point to a specific later decision, made differently, and say what it cost. That is the loop Masad described on stage: the work was examined, the judgment changed, and the change showed up in what the company did next. The follow-up is not a trick. It is the interviewer asking whether the retrospective ever left the room.
Score your answer against the director’s bar
Q: Describe a project that you wish you had done better and how you would do it differently today
Write this story down while the details are still sharp, then rehearse it until the pivot to today is as specific as the original mistake. Try it free →