<aside> 👋
Thanks for joining Notion & Hustle Fund webinar!
We hope the conversation with Shir and Elizabeth sparked new ideas for building faster, smarter teams. Below you’ll find:
As a thank-you, we’ve got a special offer for you: up to 6 months free of Notion’s Business plan — including Notion AI (over $12k in savings!)
One workspace. Every startup tool. The AI workspace from idea to exit.
</aside>
Every meeting, event, or webinar should move your startup forward. We put Notion AI to work so you can truly stay present and focus on the things that matter.
Startups move fast. AI Meeting Notes keeps up - turning this webinar into organized takeaways, decisions, and action items your team can act on instantly. All in Notion.
Summary
This session featured a discussion with Shir, an engineering leader at Notion who previously worked at Google and Waymo, sharing insights on building and leading engineering teams in an AI-driven environment.
Notion's engineering hiring focuses on finding people with growth mindsets rather than just specific skills. Shir emphasized the importance of properly defining roles—neither too specific nor too vague—and assessing candidates based on their ability to learn and iterate.
For AI-related roles, Notion has introduced a "prompting interview" where candidates work on real problems that require effective AI interaction. This interview evolves as AI models improve, ensuring candidates demonstrate both technical skills and adaptability.
When competing with larger companies for talent, Shir recommends personalizing the pitch based on each candidate's specific interests, providing a more tailored experience than larger organizations can offer.
Notion's engineering team organization prioritizes flexibility over rigid processes:
Shir shared that they use a lightweight synced block in Notion to track weekly goals, allowing for easy adaptation as project priorities shift. For cross-geography collaboration, they maintain in-person teams in each location with complementary skill sets, connecting via video for critical meetings.
Projects are evaluated based on three key criteria:
The ultimate test for success is whether the team itself uses the feature. When projects don't succeed, Shir emphasized the importance of:
For major launches, the team conducts retrospectives to identify lessons for future projects.
AI has transformed how Notion's team works:
Shir noted that AI is shifting team members from being "writers" to being "editors" of content and code. The level of human oversight varies based on the complexity and stakes of each task.
Shir shared several leadership principles:
For new hires who underperform, Shir recommends first adjusting the problem to better match their strengths before having frank conversations about fit.
Notes
Transcript
I started my career, I joined this kind of small incubating team at Google, working on trying to integrate speech recognition into search. Which eventually became the Google assistant and I just thought it was like a really cool and exciting space. I wanted to build a computer you could talk to, and so that's kind of where I cut my teeth as an engineer and then eventually.
I thought self-driving cars were really, really cool. It would be awesome to get a robot that drives. So I joined Waymo, spent another few years there. That's where I became a manager for the first time. And then joined notion. I wanted to learn how to build a really great product and I thought that notion was just a really nice product while crafted.
And then after being an ocean for about a year, GPC 4 came out and everything changed in the most exciting ways. And so we created a small. That's what I've been doing for the last two years. Awesome. Thank you, Shir. Hi, everyone. I'm one of the co-founders and general partner at Hustle Fund. We invest in very early-stage startups, pre-seed. We're basically like institutionalized friends and family.
The rich uncle or aunt you may or may not have had. Um, we're going to talk a bit about a bunch of topics today in how she or these her engineering teams etc. Let's start actually with hiring because that's where it all begins, right? Can you talk a little bit about how do you even hire somebody? I think many people on this call may never have even hired an engineer. They're thinking about doing so for their startups. Yeah.
So I think there's kind of a couple different things to think about for hiring. The first is, what's the actual role that you need filled? And I've seen two kinds of mistakes, and I've made two kinds of mistakes for the role definition. One is being overly specific in a way that means that you'll never find someone, and the other mistake I've made is being overly vague, where you are just like, oh, I just want someone smart.
And you hire them and then you don't actually have something for them to do with the right amount of specificity. So like making sure that you're thinking at the right altitude there. Once you know the kind of role that you need or what skills gaps that you need to fill, well then that's where the hard work comes in of trying to find the person. Now, I've been really, really lucky that I've mostly been at larger companies that have recruiting structures and I think it's a lot harder if you're trying to do it all on your own.
But even then, a lot of the hires that I've made are just like people in my network, people I've met at events like these. You can tell a lot by a person just by how they talk. And then the third part is figuring out if they have the skills that you need, and that's where I think one of the most important things that I overlooked earlier in my hiring.
I only focused on do they have the skills that they need? And I didn't focus on can they learn the right ways? And I think again and again, I'd so, so much rather have somebody that can learn. but doesn't have the skills because then they'll be curious and be a lot more effective in the long run than somebody who has maybe more skills than another person but just has no agency or motivation to learn. So I always pick the growth.
And how do you test for that? Is that like interview questions? Do you do tests or projects? Yeah, like peeking behind the covers now if I ever interview any of you, you'll be prepared. I just ask, almost like overly, what did you learn? What would you do differently? Did you, if you went back and did it again, what would you do the same?
What were the mistakes you made? And I think if somebody is able to think about a project that they worked on, even one that was really, really successful, and still be able to extract a learning from it. Even like the most basic of small tasks, if you're able to reflect on something that you learned from doing that, that shows that you have a
Kind of a learning mindset and you know in this day and age with AI and vibe coding. Are you like have you changed anything about your hiring process is is that something someone needs to. Demonstrate that they can use or can't use or something. Um, so yeah, we do we've changed which skills we look for. So, um, we have a prompting interview for anyone interested in doing prompting. Um, and it's really funny. We actually we've got to continue.
iterating on that interview because the new model comes out and then the prompting is not as hard because it's just better and so it's like this kind of fun constant game. And then in the prompting interview, again, it's like you're looking for the curiosity. If it if the model doesn't get it right on the first try, do you get like paralyzed? Or are you able to iterate on it? So that's a new interview that we've kind of replaced.
But the rest of the principles are quite similar. You're looking looking again for that growth mindset on the technical side for engineers looking again for can you think rigorously and robustly about a problem. And I think you can measure that regardless of if it's like an AI-assisted code or just sort of a thought exercise.
Do you guys do like project based sort of interviews or testing so we don't do like a come to a project with us, although I think that that's probably the absolute best way you can tell. If someone's a good fit, we'll do take-home assignments, especially for our intern class, all went through take-home assignments. Okay, okay. And how do you test for
I don't know how easy someone is to work with or hard or challenging. Yeah, again, it's to me it also really comes down to that learnability and that growth. Mindset is this somebody who can, when being given a problem, have humility for the problem and not necessarily just think that they know all the answers.
And again, I just test that by asking, what did you learn, what did you learn, what did you learn? And I find that if somebody is curious and has respect for the difficulty of the problem, They usually can get along with everyone, but if someone comes in with an overly rigid way of thinking, that's really, that's often the source of problems. But I think conflict is good, so I don't necessarily need to be hiring somebody that agrees with me 100% of the time, but they just have to be able to listen and really mull over a new way of thinking.
Now, a lot of these folks in this webinar are running startups. So, you know, budgets are probably a little bit tighter than certainly Google when I was there as well. Or, you know, cachet, right? Like Google and now Notion both have great brands that people know and love.
How, but you know, how do you win somebody over like after you've decided you want to work with them, especially if you're competing with like a Google, a big behavior? Yeah, every single person is different. And so I think the way and this is, I think, especially useful in a startup environment where you're competing with.
Are they looking for autonomy? Are they looking to try something new? Are they looking to be working really closely with a group of people? Are they looking to be in person? Like, really understanding their... needs and then really just talking to that specifically and like porting that side of them and you can give them a really personal experience, which a big company can never do. And so we've tried.
As much as we can at Notion to really, really hold on to that, I'll talk to every single candidate we give an offer to, I usually talk to multiple times, just to try and figure out what it is, what is it that you're actually interested in? Also, no hard feelings if notion's not the place for that. That's okay, but making it personal. Yeah, I mean, I know that sales is often a dirty word, especially engineering, but it sounds like that is what you were doing, you're catering your pitch.
Yeah, absolutely. Yeah. Okay. Interesting. And like, do you guys measure your sales wins like on the candidates you do offer? Yeah, we definitely measure our overall accept rate. I can still probably list out every person by name that I extended an offer to that said no. There's like a...
It's like a special place in my heart for them and maybe some learnings from there, too. Yeah, and I think a lot of, I mean, two adjustments from the rejections. One of them is just, was I looking for the right person? So maybe that role setup or the people I was talking to was actually the wrong fit. And then also changing the pitch. So maybe somebody was like, I don't believe in notion.
Now I know that in my first conversation with you, I have to show my genuine excitement for Notion's business. And making sure to incorporate those learnings, especially for things that people get wrong. Um, earlier in the process, and I think that is absolutely key because when I often hear founders say, oh, I can't win this candidate or that candidate, there's always something an angle to sell on. If it's comp, then you're probably selling on the wrong angle.
But OK, so actually this is a good reminder to everyone because I'm starting to see questions in the chat. Definitely this is meant to be interactive, so throw your questions in the chat and I will pull from them. So actually I'm going to take the first question here.
Um, Noah's asking how do you do the prompting interview questions? Yeah, and Noah says that actually you're the first person who's mentioned this. Oh really? Pretty cool. So, without giving away what we have, basically we took a problem that we genuinely had to solve.
with prompting, and we just kind of broke it down into something that fits into an hour. We, just like tactically, we have our own kind of separate API key that we'll give the candidates to use. So there's a little coding involved as well. But yeah so it's a real problem because as you probably all know there's a ton of prompting that you have to do in your day-to-day and so there's many to choose from that we just bring to the interview process.
Cool, alright, we're going to switch gears a little bit so now you have a team of people whom you've hired. How do you organize your team and run sprint? I guess first question at Notion, are you guys? You know all in person you guys have wonderful headquarters or do you have some distributed folks or how does that work? Yeah, so we're all in person and I actually.
Um, uh, prepared something little small to answer this question just to show rather than tell how we how we run the team. So give me a second. Let me share. My screen. So the way that I organize the team, process-wise, we have a daily stand-up. That's it. We don't do...
Anything fancy in terms of big process. And then we organize our standups to be around a project. And so I try and keep it really, really loose because with, as you all know, with AI, it's kind of impossible to predict what new capabilities are gonna get unlocked.
And so we sort of swarm into these sort of Tiger teams all around a specific project or feature we're trying to build. And our stand-up process is quite simple. People list out their updates ahead of time. We'll read through them asynchronously. Maybe there's something to talk about, maybe there's not.
If there is something to talk about, then we'll add a discussion topic, and oftentimes there's no discussion topics, and that's fine, and we'll end early. And then at the beginning of the week, we'll come up with goals for that week, stick it in a synced block, and then usually it's around whatever prototype we're trying to build or feature we're trying to launch.
And then we'll take the goals for the week because it's a sync block. I can then have this sync block live in many different places. So if like my manager asks me what's And then you heard it here first, we have a new feature launching this week, hopefully. But we just re-skinned it to give you the ability to ask questions that pulls from information, not just from your current page, but from other places. And so, for example, if you have a particular Slack channel, Where all of your updates are, you can select that channel specifically.
So we've been using that in our stand-ups every day to basically pull from Slack and Notion what each person works on. And so like this whole section gets automated, which is pretty fun. Oh, that's so cool. Yeah, we are big users of Notion ourselves at Hustle Fun.
I think along similar lines that people are looking for ideas. With our investment team meetings, we don't quite have a stand-up in the way that maybe an engineering team might, but we are notoriously awful at trying to all You know, update stuff in a centralized location, but people use slack and so our workflow is set up such that we can just actually type in pound topic and then add the whatever and it gets pushed a notion so that it is in that centralized table.
For our meetings, so that's another way if your team is not so organized like perhaps the notion team is that there are ways you can get things organized in notion without everybody needing to be sort of a. And that's what's like so cool about AI is that you can be. It's like very amenable to my kind of chaotic disorganized mind because as long as the information exists, so can you.
Somewhere I can find it. And so it's really enabled me and maybe not so good of a way to be more disorganized than I used to be. But it's quite convenient. Yeah, that's super cool and I'm excited to see that launch later this week. So so when you have this team stand up, it's it's in person. It sounds like how much time you spend on stuff. Is it like rapid fire? Like what does this look like?
Yeah, so, um, we optimize for, um, kind of fast decision making for the most part. And so there's kind of two, two parts of that. One is trying to make the stand-ups the time to make decisions and to talk through issues. And that's part of why it's so important to have it every day so that you don't get, if you come across some problem in the afternoon, you don't have to wait until like next week's meeting to figure it out.
You can just solve it tomorrow. Or even later today, depending on what time of day it is. So I'd say we really, really optimize for decision making. And we'll also try, the reason we have that lightweight synced block. is that our goals change. You try something, it doesn't work. You have to abandon that direction and go in a different direction. And I found the more classical two-week sprint planning
To just be too rigid for the pace at which things move and the pace at which you discover, okay, this architecture does or doesn't work for the capabilities of this new model. And so I think like extreme fluidity and extreme flexibility is what I'm trying to organize around. I don't know if that answered your question, but that's the like philosophy. Yeah. I mean, I think that makes sense. While you have everyone there to make quick decisions.
What decisions do you think can be made async versus live? Like, how do you parse through that? Um, uh, I think, uh, maybe I shouldn't say this too publicly, but I usually say ask forgiveness, not permission. And anything async that doesn't have a lot of consequences, or that's, we'll talk a lot about one-way doors versus two-way doors.
So I'm trying to think of an example. In the AI blocks feature that I just showed you, we sort of went back and forth on what kinds of blocks should be supported in and out of that AI block. And we actually decided asynchronously. You know what we're just going to do the easy ones that don't require a lot of work and then the harder ones when users ask for them. That's when we're going to add them.
And that was an async decision. It's something that we could have agonized over a long bit, but it was like, okay, just take the easy or the path with the least friction. Because it's a one way, it's a two-way door. Once you decide you can add more things later, it's not a big deal.
But then synchronous decisions would be things like maybe changing the overall direction of the project. Like, when we were building our enterprise search, having it connect to many, many different third parties, there's some decisions that impact the end-to-end latency of the whole system.
And so there's that like quality speed trade-off and so that actually you want we wanted a live discussion because we're you sort of if you're taking debt in one place you want to make sure that it was worth it. Yeah, well, I guess like for these for these meetings and stuff like it sounds like it's just a very limited number of meetings. Would you say that for your individual contributors is basically one?
This is the one meeting, or do they have other meetings with other subgroups? That's the goal. The goal is that they only... So, I attend many different stand-ups because there's different projects. But globally, I want to give people as much maker time as they can. Yeah. And how do you think about one-on-one meetings? Like, how often do you meet with your reports? Yeah, so it depends on the report.
Try and just do whatever is most enabling for them. And so I have some who are like really asynchronous processors. And so I actually want to meet with them less often, but for longer. so that they can come up with a list of pre-thought-through things that they want to talk about. I have other folks that require a human to talk to in order to get their ideas out, so I'll meet with them more often. And so it really depends. What is sort of the range, like once a week, or?
The range is, I think for my leads, I'll meet once a week. For anyone that's working like deep, deep on a project or trying to bother them. And so I'll do more closer to bi-weekly. But it also it like changes. If there's anything you'll take away from this, it's flexibility. Some months, some quarters, some weeks. We do need to meet, we don't need to meet. I try not to be too rigid. Just because we did it like this, we have to keep going.
Yeah, I, you know, I think one of the things that I love that you said earlier about how you have everyone come in essentially with a prepared mind with what they want to talk about. Do you have your one on ones do that as well? Like, hey, we're going to have the meeting if you are bringing something or like, how do you think about that? So I start every single meeting with what's on your mind.
And sometimes there's like a long laundry list, and we'll read first, we'll write everything down, and then we'll walk through it one by one. But sometimes I'll come in and I'll be like. You know what, I don't have anything top of mind, and then my next words are, okay, let's cancel. Um, so, um, I think, like, again, we're often
to attach to, oh, there's a meeting on the calendar, I must have it. And I think that it's really maybe bad to get trapped in that. I think there's a lot of meanings that actually are wonderful that don't have agendas ahead of time because you kind of come in with this problem that's burning and then you just want to talk about it and get it off your chest.
<aside>
</aside>