What separates the teams winning in the AI era?

Wei-Wei Wu: In the age of AI, everyone wants to tighten the feedback loop. Being able to adapt, be flexible, and be willing to change my opinion. Instead of waiting weeks to get user feedback, we can just ship it and see what happens. The price you pay for not experimenting is you don't actually know what's possible.

Emma Webb: You're listening to We Built What? The podcast for engineering leaders building in the agentic era. I'm Emma Webb from Augment Code, and I'm so pleased to be here today with Wei-Wei Wu, CEO of Momentum. Thank you so much for joining us.

Wei-Wei Wu: Thanks for having me.

Emma Webb: Wei-Wei, you just recently closed your Series A — congratulations. Can you tell me a little bit about the thesis behind Momentum? What are you building?

Wei-Wei Wu: Momentum is the AI verification layer for software. We are the guardrails to make sure your software actually works the way you intended to for your end user. We're powering customers like Notion, Webflow, Cora, Zero, and more.

Emma Webb: This is such an important problem right now. Why do we need a standalone company to do it?

Wei-Wei Wu: As teams adopt more and more AI tools, they're shifting toward plain English specs for how they talk to their AI coding agent. We want to make that spec also the way you verify it — along with all the success criteria and edge cases. The way you describe how you would build it is exactly how you would verify the AI agent's work as well.

Emma Webb: So the spec is part of the verification step, not just the design stage.

Wei-Wei Wu: Exactly. The spec is the source of truth for how your product or feature is supposed to behave.

Emma Webb: One of the things people talk about in the agentic era is that you really don't want the same agent that wrote the code to also be verifying it. Is that part of why Momentum exists?

Wei-Wei Wu: Yes. I'm tasking all of my AI agents to write code for me — we have them all hooked up in Slack internally at Momentum. But I also want proof that they did their work correctly. That goes beyond static code review and unit tests. I actually want to see it work from the user's perspective. It's important that it's a separate tool doing the verification. But it also works well with the AI that's building the feature in the first place. It's an ecosystem that grows together.

Emma Webb: There's a big conversation right now about what it means to be AI native. What does that look like for you at Momentum?

Wei-Wei Wu: When I first started coding in high school we were using Notepad++. In college it was Sublime Text — before Cursor or any AI coding tool existed. I would not consider myself AI native. AI native is someone who grew up with AI. But for me, being flexible and willing to change my opinion about what's possible is the thing that matters most, especially as a founder of an AI native company.

Emma Webb: The way you describe it, there's a before and after. And if AI native means someone who always built that way, that's a pretty small cohort of people.

Wei-Wei Wu: I agree. Having experience in a non-AI-native era is actually important for context. Instead of prompting in my editor, I used to have to type everything out by hand. That seems insane now.

Emma Webb: In my day, we coded by hand.

Wei-Wei Wu: In the future I imagine I'll have an electrode attached to my temple, just thinking software into existence.

Emma Webb: It's become a bit of a problem because I build so many apps now that nobody wants them. I'm like, look at this cool thing I built. And they're like, you get one per week, that's your maximum app distribution. What is our appetite for this level of creation? Where's the limit?

Wei-Wei Wu: We should be testing the latest and greatest and finding what's actually bringing 10x, 100x productivity gains. That means experimenting a lot, building a lot of different apps and integrations, and seeing what works. The price you pay for not experimenting is you don't actually know what's possible. Even if 99% fail, that 1% is going to be a huge lever for your team.

Emma Webb: And part of that discipline is getting comfortable killing those apps, killing that code, saying that was a good idea that just didn't work out. We've never had this proliferation of code that's so disposable. We haven't fully figured out the cleanup stage yet.

Wei-Wei Wu: Instead of waiting weeks for user feedback, we ship Momentum multiple times a day. We can just ship it and see what happens. There's a lot of value in just getting it out there and seeing what works, because it's very easy to fix or update.

Emma Webb: Your customers have run 2 billion test steps. What are you seeing in the data about how teams are actually behaving?

Wei-Wei Wu: The rate of code output has obviously increased a lot. But more output is not necessarily better. It's important that more code output directly translates to new features and better user experience. Everyone wants to tighten the feedback loop. The teams shipping the most value are the ones optimizing for that feedback loop, through tests on critical flows like authentication or billing.

Emma Webb: How do you think about the quality of the signal? In a pre-AI world, part of the discipline of an A/B test was knowing how long to run it to get statistically significant data. How do you think about that now?

Wei-Wei Wu: When Jeff and I first started Momentum in YC, we had maybe two paying customers. What does statistically significant mean with two? How we approach it today is through shared Slack channels with customers, regular meetings, session recordings, qualitative and quantitative feedback. Product shaping at Momentum is vibe-based. It's based on conversations with prospects and customers. Do we have to survey 300 users to get 95% confidence with plus or minus 5% error bars? Probably not. Based on all the conversations we've had, I'm pretty confident the change will make their life better.

Emma Webb: If we're experimenting and shipping at this rate, the data in many cases will never catch up, or it'll be too noisy to distinguish signal. Taste and judgment become even more important. How do you develop that?

Wei-Wei Wu: We're building the tools we wish we had when we were engineers. There's incredible value in that because you may not be able to quantify it, but you have an intuition for what the workflow should feel like as an engineer using a developer tool.

Emma Webb: How is the role of the QA engineer changing?

Wei-Wei Wu: The lines between roles are blurring more and more, and the barrier to entry into what we used to consider software engineering is blurring too. Titles will need to change in the very near future. Technically, anyone could be an engineer today. But the QA skill set is incredibly valuable right now. If you used to be a QA engineer and now you're building full-fledged features, you have a much higher bar than someone with no experience testing or verifying product. It's thinking from first principles: we built this cool software, but how are people going to misuse it? What are they going to type into this text box? That type of thinking is incredibly valuable.

Emma Webb: In bigger organizations, that person often saw the problem, had the skills to fix it, but for whatever reason wasn't allowed to. Wei-Wei Wu: Exactly. We want to make the feedback loop as tight as possible with as little middleman as possible. That's how you iterate quickly and move fast. How is your engineering team structured? If you can output at a higher rate, if your feedback loop is tighter, if you're closer to both customers and code — what does the team need to look like?

Wei-Wei Wu: Momentum is still pretty small — 13 people, about to be 14, with seven engineers. The amount of parallelization we can unlock with AI is massive. Instead of executing on one or two projects, we can execute on four or five times more. We just launched Android support, and iOS is coming in a week. Two or three years ago that would have taken six months to a year. It took a few weeks, maybe a month and a half. It's incredible.

Emma Webb: What is the most overhyped thing in AI testing right now?

Wei-Wei Wu: Crawling your whole app to figure out everything your users are doing and generating tests from that. It sounds great in marketing — the AI is just going to figure out everything that's broken. But using production as your source of truth is fundamentally flawed because production will have bugs. What's live in production might not be what you actually want. That's why we go back to the spec.

Emma Webb: What exists may not be your ideal. The opportunity space is the delta between your ideal and what's out there today.

Wei-Wei Wu: Being specific about what you want and what you don't want is incredibly important. Whatever you tell AI, anything that's ambiguous, it's definitely going to go there. Specs are your north star. What you want your software to look like, how you want your product to behave. When a test fails, how do you know if it's the test's fault, or if the feature change was intentional, or if it was a breakage? Support docs are out of date, the Jira ticket was implemented two months ago, there have been a few thousand commits since then. Figuring out what the software is actually supposed to do is going to become more and more important as we delegate more to AI.

Emma Webb: How should engineering leaders think about where the quality bar is?

Wei-Wei Wu: I don't want to dictate where the bar should be set, but I think the bar is higher than before. Just because you can ship a lot more code doesn't mean you're shipping higher quality code. There's real value in stepping back and thinking from first principles: what are the problems we're actually trying to solve for our customers, how do we want to solve them, and how do we present that to the end user. To have high quality, you need to be specific. When I create a Linear ticket, I think: if I gave this to an AI agent to build, would it actually know what I want? I have a random idea in my head, but my engineers and our AI can't read my mind. I need to be specific about what I want and what it needs to look like.

Emma Webb: Ambiguity is not a good thing for your agents.

Wei-Wei Wu: It can spend money now. It can send emails. I'm like, I don't know if I wanted it to do that. There's all this hype about AI agents being able to do everything for us, and not as much thought about whether you actually want it to do that. I don't want it to pay my rent or sign a lease for me. I want to tour the apartment. I don't really want to be in the Wall-E movie.

Emma Webb: I think about that movie so much. Wei-Wei, this has been really fun. Thank you. You've been listening to We Built What? If you like what you heard, please like, subscribe, and share with your friends — humans or agents. And if there's someone you think we should interview, let us know.