Tech Titans: Rethinking Career Five-Year Plans with Francisco Trindade, VP of Software Engineering

**Francisco Trindade** (0:00)
The big question I think that's facing, the companies are facing is like, can we get to a point where humans are out of the loop in the way they are right now, which is like, you need to have a code review to get products out.

**Joel** (0:11)
Now you have like one person doing the work of 100, and then 1,000, and then 10,000.
Yeah, you had a pretty popular article on Medium about will humans still review code? What did you learn from writing that?

**Francisco Trindade** (0:30)
Yeah, that was an interesting experience as well, and I want to continue that article. But I think it was also the result of, I mean, we all have been experiencing this shift in AI in this year, all models became much better, and people, I think, we came back from the end of the year of thinking about, realizing, okay, this is actually moving faster than a lot of people expected. And I think I started experience at work, so talking to people and seeing this discussion happening in the industry, that's really interesting. We're going through a big shift of what's happening with software engineering.
And I do think that as AI became the tool of choice to writing code, and developing a lot of the code that's happening nowadays, the bottleneck has quickly shifted, I think, to code reviews. So I think a lot of companies experience the idea of, okay, now reviews are a problem, and reviews are the thing that we need to optimize. And that article was really a summarization of these conversations, and getting this understanding of, okay, there seems to be, I talk about three versions of AI so far. I think V0 being last year, when we all gave tools to engineers, and no engineers had tools, and there was a lot of code being produced, but I don't think we've seen a lot of companies haven't seen a lot of the productivity gains, to some extent they have, but not meaningful. I think V1 now is what we're going through now, which is companies are trying to optimize processes around AI. There's like, okay, code review is a bottleneck, so how do we deal with that? And if product or the planning is becoming a bottleneck, how do we deal with that? And also tooling to optimize specific things. So a lot of companies are deploying agent systems so that engineers can use that directly and faster, or bug automation, or issue automation, so that small things can be solved automatically. I think the big question for me is just, can we get to a point? The big question I think the companies are facing is, can we get to a point where humans are out of the loop in the way they are right now, which is being kind of this, you need to have a code review to get products out. And I mean, in some companies that is already the situation, I think if you're starting a company now, if you have a small code base, I think everyone knows at this point that AI can effectively do most of it and deliver pretty fast. I think some large companies are also experimenting with that, and I think everyone has heard the Plot CTO Podcast saying about how code you solved and how that's not a problem anymore. But you still have this, in a lot of situations, I think that hasn't been in a lot of companies like mine or different companies that I work with, that's still not a reality. I think still code reviews are a meaningful part of guaranteeing quality and guaranteeing stability in the technology environment.
So I think the question is like if we get, and I think the prize at the end of this is meaningful. I think if we get to a point where we can take humans out of the loop actually and humans still participate in coding but in a way that's more direct in agents. We're talking about 5, 10 X gains into how fast we can deliver. But I think that is still, can we get there in a large complex revenue drive in codebase? I think that's the question that I think is interesting. If we can, I think that will be a meaningful productivity shift. But I think that's the world that we end, but also requires meaningful investment from companies to get there, because I think there's a lot of how do we actually structure that, to structure code bases and structure process to achieve that. But if you don't do that and if you don't believe in it, but you don't invest in it and someone else does it, you're going to have competitors running much faster than you. I think that's kind of like the, it's an interesting kind of bifurcation, but I think it's an interesting choice that I think companies are going to start to have is like how much they invest in this direction and what is the outcome if they do or they don't.