# Exa uses work trials to test first-principles thinking

Markdown copy of https://realalpha.app/s/exa-uses-work-trials-to-test-first-principles-thinking/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

- Speaker: [Will Bryk](https://realalpha.app/builders/will-bryk/) · Will Bryk is cofounder and CEO of Exa. · Co-founded Exa, a search engine built for AI agents with its own index and ranking algorithms.
- Host: [Nakul Mandan](https://realalpha.app/people/nakul-mandan/) · Host, Knuckle Up with Nakul · General partner at Audacious Ventures; his show goes deep with founder CEOs on how they build.
- Show: Knuckle Up with Nakul
- Episode: [How a first-time CEO is taking on Google: No process, no NPC hires, and extreme shipping velocity](https://realalpha.app/v/9S9UOjpGFUs/), Aug 4, 2026
- Moment: 16:26 to 19:50 of the episode (3 min), playing on https://realalpha.app/s/exa-uses-work-trials-to-test-first-principles-thinking/; the episode from there: https://realalpha.app/v/9S9UOjpGFUs/?t=986

## Quote

> “You have to give them **a real creative problem** and see how they think through all the trade-offs and stuff like that.”
> — [Will Bryk](https://realalpha.app/builders/will-bryk/) at [17:02](https://realalpha.app/v/9S9UOjpGFUs/?t=1022) · Knuckle Up with Nakul

## Takeaways

1. Work trials last one to two days.
2. Use real creative problems to see how candidates reason.
3. Observe how candidates interact with the team, not just their résumé.

## About this segment

Exa brings candidates in for one- to two-day work trials and watches them solve open-ended problems. This reveals reasoning and interpersonal fit that resumes alone miss.

From “How a first-time CEO is taking on Google: No process, no NPC hires, and extreme shipping velocity”, Aug 4, 2026. [Watch the full episode](https://realalpha.app/v/9S9UOjpGFUs/)

## Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[16:29](https://realalpha.app/v/9S9UOjpGFUs/?t=989) but we try as best as we can. How do you test for that first principle thinking? Because a lot of people look smart on paper. What is the recruiting process to assess this? Yeah. I mean, so for example, on the engineering side, we have a couple of technical interviews. Those are reasonably standard. You know, we've chosen for like interesting, not just like leak code type stuff. But then I think the big thing that helps us test for this is we do a work trial, meaning people like come to the office and work with us. For how long? For one to two days. And that's critical because then you really get to see what someone's like.

[16:59](https://realalpha.app/v/9S9UOjpGFUs/?t=1019) So you get to see first principle of thinking because you give them a real problem. So you can't really see someone's first principle of thinking if you give them some lead code style problem. You have to give them a real creative problem and see how they think through all the trade-offs and stuff like that. You could tell just from just like seeing the way they think through the problem. Awesome thing about a work trial is you also get to see all these intangibles.

[17:16](https://realalpha.app/v/9S9UOjpGFUs/?t=1036) Like, you know, how do they interact with people at lunch? Are they like optimistic? Are they, you know, can you imagine working with them every day? Are they locked in during the day? Like all these kinds of things you could test during a work trial. I read somewhere, correct me if I'm wrong, that you try to, for people who are coming for the work trials, you try to have a meal with them. Is that kind of the intangibles?

[17:33](https://realalpha.app/v/9S9UOjpGFUs/?t=1053) It's starting to not scale because we have multiple at the same time. I definitely walk with everyone, or I try to, or me and my co-founder. That's very critical. We should continue doing that for literally as physically long as we can because we've gotten good enough for a 15-minute walk. We could see a lot in someone. We have all sorts of methods that we do and places we take them and just see how they react.

[17:55](https://realalpha.app/v/9S9UOjpGFUs/?t=1075) double click on that like how have you gotten better what are traits that you are now double like looking more for and what are traits that you've stopped caring about that you may have over cared two years ago we really try to avoid non-playable characters is like the way of saying it like people who just go through life like as if they're on a stream and they have no like agency over themselves someone should be like they think about their life and think about the world and like adjust accordingly and don't just follow like non-followers and how do you test for that well you just literally just ask them about their life and like their decisions and stuff like that And you can kind of see.

[18:25](https://realalpha.app/v/9S9UOjpGFUs/?t=1105) Also, I just tell people about where EPS is going and just see how they react to it. Like, do they get excited? Do I see a sparkle in their eye? Reasonably easy to see a sparkle in their eye. Sometimes people are jet lagged because we have to fly them. So you have to adjust for that. And so all these subtleties that you learn over many trials. Are there traits you used to care a lot more about two, three, four years ago when you started the company that has just like you've gotten better at looking for something that now you think I was over-indexing it doesn't matter?

[18:51](https://realalpha.app/v/9S9UOjpGFUs/?t=1131) I'm sure there are a lot. One thing that just comes to mind, right, when you're asking that is like uh speed of talking like i happen to really like uh fast talkers just because like i talk really yeah and that is a surprise and i don't have time like you know my time i just want to move fast and make decisions fast and i get really bored but i think one thing i've changed there is like there are some people who speak slower but every token that they say is really high value so it's not actually the speed of words it's like valuable tokens per second it's kind of like the right metric so i've adjusted there is there a type of person who's clearly talented but you have realized it's not a good cultural fit for EXA? Yeah I mean there are a lot of really experienced it depends what you mean by talented but like definitely there's a lot

[19:31](https://realalpha.app/v/9S9UOjpGFUs/?t=1171) of experienced people and that could often be a negative thing. There are a lot of people who are like you know incredible pedigree even have done amazing work in the past but can they like adjust to this new agentic world whether it's like the tools they use every day to build or just like the way they think about retrieval. Sometimes people are you know locked in the past. I mean