# Agent-led product work depends on automated testing Markdown copy of https://realalpha.app/s/agent-led-product-work-depends-on-automated-testing/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy. - Speaker: [Vlad Tenev](https://realalpha.app/people/vlad-tenev/) · Vlad Tenev is co-founder and CEO of Robinhood. · Co-founded Robinhood in 2013, then steered it from a $500M loss in 2023 to a $1.1B profit. - Host: [Alex Heath](https://realalpha.app/people/alex-heath/) · Founder & host, Sources · Left The Verge to start Sources in 2025; it has made over $1M since, mostly from sponsors. - Show: Sources - Episode: [Robinhood’s CEO on letting AI trade your money](https://realalpha.app/v/rkTMQ3oHzFE/), Oct 13, 2026 - Moment: 23:08 to 27:14 of the episode (4 min), playing on https://realalpha.app/s/agent-led-product-work-depends-on-automated-testing/; the episode from there: https://realalpha.app/v/rkTMQ3oHzFE/?t=1388 ## Quote > “everyone should have a team of agents that's **productive while they sleep**” > — [Vlad Tenev](https://realalpha.app/people/vlad-tenev/) at [23:38](https://realalpha.app/v/rkTMQ3oHzFE/?t=1418) · Sources ## Takeaways 1. Agents need context and a way to test their own work. 2. Tenev expects product development to become increasingly automated. 3. Live-customer adoption is harder to test than model performance. ## About this segment Vlad Tenev wants employees to manage agents that work around the clock, with enough context and tests to finish tasks independently. He sees testing changes with live customers as a key product-development bottleneck. From “Robinhood’s CEO on letting AI trade your money”, Oct 13, 2026. [Watch the full episode](https://realalpha.app/v/rkTMQ3oHzFE/) ## Transcript Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there. [23:12](https://realalpha.app/v/rkTMQ3oHzFE/?t=1392) just make that easy for you? I'm curious how you're using agents in your personal life for running the company, also managing your money. And I heard from your team, you really love loops. Yeah. You're doing a lot of loops. We could talk about that. But yeah, curious how you're using it and how it's evolved for you these last few months because things are moving so fast right now the first thing i started out with was just writing code i became an engineer again and you know through that i guess my my north star is that everyone should have a team of agents that's productive while they sleep so we should be like a 24 7 company and not just in aggregate but kind years 24 7 push you're doing on like trading windows and all that yeah each person should be a 24-7 agent manager. Your job becomes managing your agents, make sure they're well-fed and [24:07](https://realalpha.app/v/rkTMQ3oHzFE/?t=1447) they're not hungry and they have the context. And you ask, all right, well, what does it take for your agent to be running fully autonomously? They have to have context. They have to be able to test their work because if they don't have the ability to test their work, eventually they stop and they're like all right i'm done you know you you you do the testing for me human so it really just requires rethinking our entire infrastructure and making sure everything from uh the continuous integration continuous deployment the testing um like that that's all really really good and it can scale and that you know we can actually automate pretty much everything and and i think that the future is all product development will be looped in this way, right? And your bottleneck will be how quickly can your changes be tested against live customers? And how quickly are you [25:10](https://realalpha.app/v/rkTMQ3oHzFE/?t=1510) going to be able to tell, is this a good change or a bad change that we should throw away? Industry-wide, software engineering is completely transforming. You're going to see an end-to-end automation of product development and infrastructure development. One of the first places you're seeing it is actually AI research. People thought that would be the hardest and the last to go because it's just complex work. But really, if you think about it, it's pretty neatly sandboxed. I mean, AI models, there's already a way to test them without having them live. We have evals. So you have automated evals. The experiments are pretty sort of like narrow in some ways. Like you can run experiments, change the AI model, deploy a new feature, retrain it in some cases. So it turns out some of the first work that's being automated is the AI [26:06](https://realalpha.app/v/rkTMQ3oHzFE/?t=1566) research work. And I think eventually it'll happen with products too. But product work is a little bit harder because you have to go and get adoption from real users. They have to test it in the wild and human adoption and behavior change changes a little bit more slowly than, you know, things in the pure agent world. And there's an intuition to products and an art really to it that I think also is going to be hard to replicate by AI. Maybe so. I don't know. I mean, they said AI would be the last to uh art would be the last thing to be uh to be replicated but you know some of the first things you saw right when chad gpt launched was stable diffusion yeah you could say make this but in the style of vincent van gogh right and yeah it was it's pretty good um now again that doesn't mean that it has the the taste and people kind of get used to things and realize that this [27:04](https://realalpha.app/v/rkTMQ3oHzFE/?t=1624) is an AI generated image. But yeah, I think a lot of the design work benefits from AI as well.