Granola Co-Founder Sam Stephenson Unveils AI-Native Meeting Workflow

借助编程智能体构建界面，Granola 将 Figma 使用率降至 20% 随着人工智能让软件代码生成变得毫不费力，竞争优势已从交付速度转向精雕细琢那些创始人往往过早放弃的产品细节。 Sam Stephenson 深入解析了 Granola 的产品哲学，详述该应用如何被设计为在高压、连续不断的开会工作日中营造平静氛围，并充当不引人注目的辅助工具。他解释了核心工作流，例如会前简报、被广泛使用的“我错过了什么？”功能模板，以及团队为何顶住频繁的用户反对意见、刻意隐藏实时转录文本。Stephenson 还分享了 Granola 的内部开发机制，包括强大的设计系统如何使智能体直接参与编写代码，进而将他使用 Figma 的时间从 80% 减少至 20%，以及他们的众包内部机器人“Nacho”如何自主解决工程故障。他摒弃科技圈严酷的盲目奋斗文化，认为在 AI 时代，竞争优势在于花时间深入思考应该构建什么，并对细节精益求精，而不是仅仅止步于首个可用版本。 Granola 围绕连续开会之间的混乱间隙设计其 AI 会议软件，刻意优先考虑平静的界面而非堆砌功能，以此实现了 15 亿美元的估值。 (0:00) 联合创始人 Sam Stephenson 通过为编程智能体编写统一的设计系统和 markdown token 规则，将其在 Figma 中的工作时间从 80% 缩减至 20%，转而通过 Cursor 直接在代码中进行 UI 原型设计。 (19:30) Granola 的内部 Slack 机器人“Nacho”与代码库、系统日志、Amplitude 和 SQL 数据库相集成，能够诊断生产环境故障并触发 Cursor 智能体自动提交 PR。 (26:14) 团队拒绝初创公司的“996”拼命文化，认为既然 AI 使得开发工作变得轻而易举，那么差异化竞争优势便来自于拥有足够的心力来决定该构建什么，并在最初的 V1 版本之后进行持之以恒的精细打磨。 (30:18)

Granola co-founder Sam Stephenson argues that AI meeting assistants should act as an ambient, calming 'handrail' that protects human presence rather than demanding attention. As software generation becomes trivial through coding agents, he contends that product defensibility hinges on knowing what to build and refining features until they deliver reliable daily value. Designing for Calm and Human Presence (3:45) Granola's design philosophy centers on reducing anxiety during chaotic transitions between back-to-back meetings. Stephenson describes the application as a supportive handrail rather than the hero of an interaction, intentionally keeping the software in the background. To keep users focused on making eye contact and listening to counterparts, the team made the opinionated decision to hide the live transcription feed into a tiny control, accepting frequent user complaints from those who assume transcription is missing. The interface instead emphasizes lightweight interventions. Stephenson relies on automated pre-meeting dossiers to quickly review candidate backgrounds before hiring calls without cognitive overload. During calls, the platform's most popular utility is the 'What did I miss?' prompt template, which brings users up to speed immediately if they become distracted or need to multitask. From Personal Scratchpad to Organizational Context Engine (5:36) While meeting tools often begin as personal note-taking utilities, spoken workplace conversations contain an organization's most timely operational knowledge. Stephenson explains that Granola leverages transcript details to track project trajectory, flag neglected initiatives, and draft deliverables such as slide deck outlines, job descriptions, and launch announcements based on historical context. To balance shared intelligence with individual confidentiality, Granola enforces boundaries that protect private one-on-one syncs while enabling shared team folders for standups, design critiques, and high-volume sales calls. In enterprise settings, workflows increasingly bypass the visual user interface altogether; external language models query Granola via APIs and the Model Context Protocol (MCP) to route meeting insights directly into internal stacks, bespoke agents, and scheduled weekly syntheses. Shifting from Figma to Agent-Driven Prototyping (19:30) The rise of coding agents has reshaped Granola's internal product design workflow. Stephenson estimates his daily time spent in Figma dropped from 80% to roughly 20%, with the team increasingly moving directly from physical whiteboard sketches into code prototyping using Cursor. To maintain design craft without bottlenecking engineering, the company introduced an agent-ready design system populated with markdown instruction files that guide coding agents on proper token and component usage. Stephenson highlights that static mockups are ineffective for evaluating AI features because UI text must be assessed under real-world emotional conditions. During the development of pre-meeting briefs, the team deployed a crude, functional version into the app within roughly a week. By forcing the internal team to experience the friction of an imperfect prototype on live personal data, Granola accelerated iterative improvements before public release. Internal Tooling and Rejecting Grind Culture (17:09) Stephenson rejects the intense '996' work culture prevalent among many AI companies, maintaining that high-quality product judgment requires mental space away from continuous coding. Instead, Granola relies on internal autonomy and custom tooling, most notably an internal agent named Nacho. Hooked into Slack, Granola, Amplitude, system logs, and the codebase, Nacho acts as the primary triage layer by analyzing system errors, writing SQL queries, and dispatching Cursor agents to author pull requests. Reflecting on product execution, Stephenson notes that his co-founder Chris emphasizes not stopping at an initial version. Because spinning up baseline features has become easy, Stephenson argues that builders must avoid treating an early release as the finish line, instead obsessing over post-launch details until users experience genuine daily utility. A persistent design hurdle for Granola remains managing the aftermath of meetings, where five daily conversations can generate an overwhelming stack of 15 to 20 action items. At the same time, the transition of meeting intelligence into headless API and MCP workflows leaves product designers with the unresolved challenge of creating value when end users rarely interact with the primary graphical interface.

[0:00] Sam Stephenson: kind of naked without them [3:45] Sam Stephenson: the number one feeling we want people to feel is calm when they use Granola. [3:45] Sam Stephenson: lower your blood pressure. [5:36] Sam Stephenson: keeps you humble [13:54] Sam Stephenson: definitely a pang of sadness in there in the UI designer part of me. [13:54] Sam Stephenson: It's been really interesting working with bigger and bigger enterprise customers, because I think just as soon as you become a big enough company... it's inevitable you have all of this internal tooling. And if Granola doesn't play nice with that internal tooling, like, you're a non-starter. [17:09] Sam Stephenson: There's just so much more to life than writing code or writing software... the work you do is better when you have some balance. [17:09] Sam Stephenson: fully like addicted to the slot machine of the coding agents [17:09] Sam Stephenson: We want to be a calming presence in your life, not the opposite. [23:54] Sam Stephenson: feeling the pain of this thing not being good like day in day out [26:14] Sam Stephenson: if you make Nacho better, you'll be everybody's best friend [30:18] Sam Stephenson: It's a cliche now, but like it's so easy to make stuff. The hard thing is knowing what to make. [30:18] Peter Yang: the AI is still too agreeable, man [31:58] Sam Stephenson: most people stop way too early [31:58] Sam Stephenson: really sweat the details and go the extra mile to make something that's actually going to be really useful to somebody.

[0:00] I think the number one feeling we want [0:01] people to feel is calm when they use [0:03] Granola. The main time we hold our heads [0:05] as we're thinking about the product is [0:06] like the messy [music] in between [0:07] moments between back-to-back meetings. [0:09] You're probably running late, probably [0:10] forgot who you're going to be speaking [0:11] to. We want to be a calming presence in [0:13] your life. You know what would be the [0:14] ultimate feature request from me? I'll [0:16] have a Peter agent and you have a Sammy [0:18] agent and then we go send our agents [0:19] have a meeting first to align. And if [0:21] they can't align, then we come in. You [0:22] get a lot of like CEOs asking for that [0:23] kind of thing. Can you make Granola be [0:25] my first line of defense from all the [0:27] questions I get all the time? [0:31] >> All right, everyone. Today I'm really [0:32] excited to welcome Sam, the co-founder [0:34] of Granola. Granola started as an AI [0:36] meeting notes tool and now they're worth [0:38] $1.5 billion. So, I actually really want [0:40] to get Sam to show us how he personally [0:42] uses Granola beyond meeting notes and [0:44] also how Granola runs internally using [0:46] agents. So, welcome Sam. [0:48] >> Uh thank you, Peter. Nice to be here. [0:49] I've I've been looking forward to this. [0:51] >> Why don't we get right into it? I think [0:52] most people use Granola just to take [0:53] meetings with AI, but I'm curious how [0:56] you personally use Granola. Can you show [0:57] us your personal Granola setup? [0:59] >> I've become very dependent on these [1:01] pre-meeting briefs. I find these most [1:03] useful. I've kind of under-prepared for [1:05] for a meeting with somebody new, like [1:06] often like a new candidate. Um if I'm [1:09] I do a lot of like cultural interviews [1:11] as part of the um hiring process and if [1:14] if I'm not up to speed on who the [1:15] candidate is I'm talking to, this saves [1:16] my ass and embarrasses me a lot of [1:17] times. [1:18] >> Nice. [1:18] >> And uh yeah, I've I've like it it was a [1:20] bunch of work to make this like just the [1:21] right information that you need right as [1:23] you're going into the meeting, but I [1:24] feel kind of naked without them. That [1:26] you know now if I'm meeting somebody [1:27] new. [1:27] >> I I think it'd be next level if if like [1:29] Granola can actually help suss out like [1:31] disagreements that people are not [1:33] expressing and then be like, "I actually [1:35] Sam and Peter, like do you guys actually [1:37] agree on this thing?" Like during [1:38] [laughter] the meeting. [1:39] >> So, there is one of our uh uh investors [1:42] um has a really good recipe, which [1:44] designed to run like after the meeting, [1:46] um [1:47] not really during the meeting, but like [1:49] uh he uses this a lot, I think, after [1:51] meetings to kind of like kind of go back [1:53] and do like a post-match analysis sort [1:55] of and figure out like, "Oh, what what [1:57] did I miss? What should I have asked? [1:58] How could I have gone deeper during that [1:59] conversation? Yeah, I I I I think that [2:02] that kind of thing is really powerful. [2:04] Like our sales team will do this as well [2:05] on a on a sales call. You know, the [2:08] after the meeting they'll be like, "I [2:09] what what should I have done better that [2:10] time Granola?" And you know, it has the [2:12] history of previous uh [2:14] previous sales calls you've done and so [2:15] it can sometimes give you a pretty smart [2:17] answer there. [2:18] >> Got it. Do you just like uh you know, [2:20] take take some notes here and then [2:20] Granola fills in the rest of the notes [2:22] and then you apply a recipe to it [2:23] afterwards or like do you apply some [2:24] sort of template to it afterwards? [2:26] >> For me like the automatic notes sorts me [2:28] out like 95% of the time. Um [2:31] the things I will use like recipes and [2:32] stuff for is I I'm I use this what did I [2:34] miss thing um a lot. Like if I'm yeah, I [2:38] shouldn't be but if I'm end up [2:39] multitasking during the meeting or [2:42] start you know, zone out thinking about [2:43] something else like I go [2:45] >> [laughter] [2:45] >> I'll be reaching for this to to get me [2:47] back on track. We uh yeah, we're talking [2:49] about adding like making this more of a [2:51] more of a productized feature like um [2:53] you have this [2:55] I'm saying over here which um has a live [2:58] transcript and I kind of I kind of want [2:59] to just make it so that you hover over [3:01] this, we we tell you what you missed [3:03] like straight away cuz it's um it's it's [3:05] by far the most used recipe like people [3:07] click on this thing all the time. [3:08] >> Yeah, I don't know if I should admit [3:09] this but like sometimes I'm interviewing [3:11] a candidate for like product too and um [3:13] they'll give me an answer and and then I [3:14] I probably should not admit this but [3:15] like I like to double check with Granola [3:17] and be like, "Hey, you know, is is this [3:18] a good answer? [3:19] Like what what what what do you think?" [3:21] Yeah. It's especially if you like give [3:23] it like a pretty good template [3:24] beforehand. Of the rubric that you're [3:26] assess. [3:27] Like it's it's actually pretty pretty [3:28] good. [3:30] >> Uh I I think for folks who have more [3:32] kind of structured types of meetings um [3:35] they lean on a template a lot for this. [3:37] Like it's a you know, when you when you [3:38] know you have the certain questions you [3:39] want to ask, they'll be like making sure [3:41] the notes come out in this structured [3:42] way and [3:43] >> Yeah. [3:43] >> Yeah, being able to do repeatable things [3:45] is important. [3:45] >> Um can you talk about some of the cu- [3:47] cu- cuz you're a designer by trade, [3:49] right? Or [3:49] am I wrong or Yeah. So, do you guys have [3:52] some like product principle, design [3:54] principles when building this app cuz it [3:56] was like very simple and like there's [3:58] I'm not sure there's a lot of trade-offs [3:59] that went into this, right? [4:00] >> We like I don't know. We We wrote these [4:02] down like a month ago, so that they're [4:04] pretty fresh. [4:04] >> Okay. [clears throat and snorts] [4:05] >> All right. Codify like [4:07] you know, when we have new people join, [4:08] we're trying to codify like what [4:10] For every new engineer that joins, like [4:11] what what few things should they have in [4:13] their head, you know, as we like think [4:15] about Granola. And then uh these were [4:16] the three we got to. I think the number [4:18] one emotion or feeling we want people to [4:20] feel is calm when when they use Granola. [4:23] We design Granola's designed for like [4:24] incredibly busy like moment in your [4:26] life. I think the main time we hold in [4:28] our heads is when thinking about the [4:30] product is like the messy in-between [4:32] moments between back-to-back meetings, [4:33] you know, where you're like ready or [4:34] managing your windows or might even be [4:36] rushing through rooms. [4:38] You're probably running late. Probably [4:39] forgotten who you're you're speaking to. [4:41] Uh you might also have a couple of like [4:43] quick things you need to get on top of [4:45] before you jump into the next meeting. [4:47] And Granola trying to like insert itself [4:49] into that moment like really chaotic [4:50] moment. And so, yeah, we really try to [4:52] hold ourselves to this idea that like [4:54] Granola should Looking at Granola should [4:55] lower your blood pressure. It does not [4:56] add to this chaos. It should feel like [4:58] it's a it's helping you there. Yeah, I [4:59] mean, I think such a blank canvas is [5:01] like [5:02] a lot a lot of that. And like as a I [5:04] think we've we've made quite a few [5:06] aggressive design decisions here. [5:08] >> Like the transcript is so hidden. You [5:10] got a lot of people find kind of [5:11] ridiculous. And honestly, we get a lot [5:13] of complaints that people don't even [5:14] realize the transcript is there. Like we [5:16] We pay a price for having it [5:18] be that small. Um but I think uh we [5:21] really really want to like just strip [5:22] away everything that's not what you need [5:24] to have in front of you during a [5:25] meeting. And [5:26] yeah, I think that's super important. [5:28] >> Because when you're having a meeting, [5:29] you want to focus on the other person, [5:30] right? You don't want to be looking at [5:31] like a live transcript, you know? It's [5:33] like distracting. [5:36] >> Yeah, we had this [5:37] metaphor of like [5:39] Granola should feel like a hand rail. [5:40] Like that's kind of like one way of [5:42] talking about Granola's role in your [5:43] life. Like it's it's Um, it's not the [5:46] main character, you know? Like you [5:47] you're the main character. You're trying [5:48] to get your work done and [5:50] being present in this meeting, you know, [5:52] the deciding the things you need to [5:53] decide or talking about the things you [5:55] need to talk about. And Grainer's really [5:56] just going to be there on the side and [5:58] support you and [6:00] help you be better, you know? Um [6:03] And uh [6:04] yeah, I really I really like this [6:05] metaphor. It It's [6:07] >> [snorts] [6:07] >> keeps you humble. [6:08] >> [laughter] [6:09] >> Yeah. Yeah, no, it's it's it's [6:10] fantastic. And um I I totally agree. [6:12] Like back-to-back meetings you're trying [6:13] to like jot down notes in the last 5 [6:15] minutes, trying to send some takeaways [6:16] on Slack. And it it it's just like a mad [6:18] rush, right? Then you have then you have [6:19] to contact switch to the other meeting. [6:21] So, like yeah, so the some reason the [6:23] pre-meeting briefings are super helpful [6:25] for for that. [6:27] >> Yeah. Yeah. Yeah. Yeah. [6:28] >> This episode is brought to you by Google [6:30] AI Studio, the easiest and fastest way [6:33] to build and share your idea with real [6:35] users. [6:37] For example, I use AI Studio to create [6:40] three different landing pages for this [6:42] bakery using images generated by [6:44] Google's built-in Nano Banana model. One [6:47] highlights fresh bread, another [6:49] highlights awesome cakes, and the last [6:51] highlights the bakery story. I can [6:53] publish the website easily and get a [6:55] custom URL, in this case [6:57] corgiconfections.ai.studio, [7:00] to share all three variations with [7:01] potential customers for feedback in one [7:04] tap. When I was a product manager, [7:06] Google AI Studio was my go-to tool for [7:08] building prototypes like this. And it [7:11] really is an easy and free way to share [7:13] ideas with team, customers, and everyone [7:16] else. Check it out at [7:18] aistudio.google.com. [7:20] Now, back to our episode. [7:22] >> Another thing I think is very [7:23] interesting is when you guys started, it [7:25] was just like kind of taking meeting [7:26] notes, right? But now, and I think Chris [7:28] said this before, meetings are the most [7:30] useful and up-to-date context that a [7:31] company has in in some shapes, right? [7:33] So, now I feel like there's a big [7:34] opportunity here to just evolve into a [7:36] whole, I don't know, like like a wiki or [7:39] like a contextual platform for the whole [7:41] com- com- company. [7:42] >> Yeah. [snorts] Yeah. [7:43] >> Yeah. [7:43] >> Yeah. And that that like the direction [7:44] that you guys are going down, right? Or [7:45] or like can you talk more about that? [7:47] >> Yeah. Yeah. Sure. Yeah. I think [7:49] this is a bit of a I think we kind of we [7:51] feel this could be [7:53] an opportunity for us, but I don't think [7:54] we realized how big of a thing it would [7:56] be, but um [7:57] uh I think so much of [8:00] when you use Grain of like day in day [8:02] out [8:03] for all your meetings like just [8:04] accumulates so much context about who [8:07] you are, who you who are the people in [8:08] your life, what projects you're working [8:09] on, what's happening in your company. Um [8:12] yeah, you know, like like and and not [8:14] just at like high-level bullet points, [8:16] but like real nitty-gritty details and [8:18] you know, Grain of can tell me like what [8:21] projects are likely to be going off the [8:22] rails soon or like what things am I not [8:24] giving enough attention to, you know, [8:25] and and and it [8:26] >> Yeah. [8:27] >> has enough nuance detail in the [8:28] transcripts to be able to give smart [8:29] answers to that. [8:30] >> I also love like just like the [8:31] pre-meeting meeting and post-meeting. Do [8:33] you run some recipes like on a regular [8:34] basis or like what what what are some [8:36] stuff that you kind of like, you know, [8:37] check with all this context? [8:39] >> Uh I I do like I will use things recipes [8:42] for quite specific things. So, like [8:45] um I don't know if I'm going to be able [8:46] to get get a good example now, but like [8:48] uh [8:49] let's say if I just enhanced the notes. [8:52] I'm to make a presentation [8:55] internally about our [8:58] um design principles. Can you make a [9:02] slide deck outline for me? [9:04] Um [9:06] I'll often like use the conversation [9:08] buddy as a way to Yeah, as draw things. [9:11] Um I've been doing this with a [9:13] like job description just earlier today [9:15] for a for a new role we're looking for. [9:17] I've used this [9:20] one of the ways. A blog post like blog [9:22] posts and new feature announcements have [9:24] been written by us all having a meeting [9:26] about what we think needs to go in the [9:27] blog post and then [9:28] >> Okay. Grain of has that, but it also has [9:30] like the whole history of project, you [9:31] know, so it's like the real [laughter] [9:32] problem is that zero. [9:34] >> Um yeah, and so it's I I find it very [9:36] useful that kind of [9:37] >> I see. And and and like uh if if I want [9:39] to write a blog post in my format, like [9:42] I I can make my own recipe or like save [9:44] some prompt somewhere. [9:46] >> Um [9:48] I [9:48] >> a recipe on here, yeah. [9:50] >> You [laughter] do? [9:51] >> I don't see it. [9:53] >> Uh let's see. Let's see what Let's see [9:55] what other folks on the team have done. [9:56] Um [9:57] >> Oh, this is this is the from from the [9:58] team. Nice. Just that was advocated. [10:00] >> Yeah, these are work with so these are [10:02] from folks on the team. Yeah. Um [10:04] >> Is there like a Sam review recipe or [10:06] like what would Sam think or or Chris [10:08] think? [10:10] Oh, it's more like brainstorming product [10:11] ideas. Got it. [10:13] >> Yeah. [10:13] >> Very interesting. [10:14] >> is funny. This is This is pretty [10:15] accurate. [10:16] >> [laughter] [10:18] >> It's a simple [snorts] suggestion we can [10:19] share. How do you measure it's working? [10:21] What's the two way to do it? [10:23] >> Yeah. Nice. [10:24] >> Nice. Yeah, yeah. You got it. [10:25] >> You got to open source this. [10:27] >> Yeah, you got to open source this. [10:28] >> Yeah, yeah. [10:30] >> Got it. [10:31] Um [10:32] But what do you think about like okay, [10:33] so Granola is for like it's really good [10:35] for me to be a little bit more [10:37] self-aware about like how I show up in [10:38] meetings and stuff, right? [10:40] But like how do you think about like for [10:41] the team? Like like you know, like like [10:43] I guess it's just kind of like privacy [10:44] issue, but like [10:46] like how how do I know like how like if [10:48] you and I were on a team and we have a [10:49] bunch of meetings together or or like [10:51] you have meetings with other people, [10:52] like how do I know how Sam op- operates [10:54] or or maybe like I'm your report and you [10:56] want to know how I show up in meetings. [10:59] Is there like an opportunity there? [11:01] Or something? [11:01] >> Yeah. [11:02] >> Yeah, there's um we obviously like I [11:04] think one of the one of the like [11:06] surprisingly hard things about Granola [11:07] is like preserving your privacy at every [11:10] at every moment. So like um yeah, in the [11:12] appropriate way. So like Granola [11:13] definitely can't like go and look into [11:15] the meetings of my reports and tell me [11:18] what's going on in them, you know, like [11:19] how they should be. [11:20] >> Yeah, that would not be good. Yeah. [11:22] >> But but I mean I mean not all the [11:24] private ones at least, but like um [11:26] you definitely can ask stuff like that [11:28] of like um the more you know, anything [11:29] that's been shared. Like we have this um [11:32] Yeah, we have like we have like our team [11:33] space and there are [11:35] I won't open any, but there's like tons [11:36] and tons of folders in here um of people [11:39] pouring team notes into here. Um [11:42] you can see stand-ups, you can [11:44] uh this is like sales calls and [11:45] >> so so there's like a team so there's [11:47] like a de- design notes channel and then [11:49] people proactively share stuff in here [11:51] and then and you can look stuff up. Is [11:53] that kind of [11:53] >> Yeah, yeah. So, I this one isn't used [11:55] very much. I guess it's like design crit [11:57] podcast thing. [11:58] Uh design team weekly. [12:00] Um [12:01] Like, the sales folder has many, many [12:04] calls a day. [12:05] Uh yeah. [12:05] >> Okay, yeah. I can believe it's that. [12:06] Yeah. [12:07] >> Yeah. [12:08] >> Mhm. [12:08] >> And uh yeah, this I mean this this I use [12:10] a lot. Like, [12:11] usually you don't you actually don't [12:13] really need to go to this folder. [12:14] Whoever is good at the chat can just [12:15] access all of it, but um [12:16] >> Yeah. [12:17] >> I go to like [12:20] uh [12:21] >> Be- Be- Be- because the sales is [12:22] actually the customer conversations, [12:23] right? So, you can look for like [12:24] insights and trends and stuff. [12:26] >> Yeah, exactly. Yeah, you know like what [12:28] that's things uh customers [12:31] say. [12:34] Um [12:35] >> Yeah. [12:36] >> This is going to look across like Yeah, [12:37] not just my meetings, but across [12:39] everything that the team you know the [12:40] team's been having recently. [12:42] >> Okay, got it. [12:43] And and and is there a way to or maybe [12:45] this is coming like at some point like [12:46] is there a way to set up au- automations [12:48] so that like you know it does the search [12:49] every Monday [12:51] or or something? [12:51] >> Uh [12:52] we Yeah, we um [12:55] I hope so soon. Like, we've prototyped a [12:56] lot of these things. Um yeah. [12:58] Uh I think [12:59] it would be easy for us to do average-ly [13:02] and like I think quite hard to do to [13:03] develop. Like, our summary is so far [13:05] like um just in [13:06] >> Yeah. [13:07] >> being able to get out of the box, making [13:09] the quality of the output very good. Um [13:11] So, yeah. I would I definitely think [13:13] we'll we'll get there. [13:14] We'll get there at some point someday. [13:15] Yeah. We have um [13:17] >> Okay. [13:17] >> Uh because of the API and the MCP I like [13:20] we have a bunch of users right now using [13:21] those to create [13:23] you know like a weekly report of uh [13:26] uh things like this. [13:28] Um, [13:28] >> Yeah. Yeah, I do it for myself actually. [13:30] I have a I I just go to Granola MCP and [13:32] then it it's it gets the information [13:34] from Granola but also from my other [13:36] sources and it sends me like a email [13:37] summary of what's going on. Yeah. [13:39] >> Nice. [13:39] >> Yeah, super helpful. [13:40] >> Right. [13:40] >> Yeah. [13:41] >> Yeah. Yeah. Uh, like I lots of other [13:44] people use the MCP and uh, [13:46] I think like a a bunch of our other [13:48] members probably just be like figure [13:50] out, you know, what are people using MCP [13:52] for the most and productize it back into [13:54] the app. [13:54] >> I have a [13:55] I I actually have a question for you as [13:56] a designer. Like um, [13:58] cuz like people are just like using [13:59] these like people are just getting their [14:01] agents to use these products, right? [14:02] Like just use the MCP or use the API. [14:04] And then and then and then how do you [14:06] think of that? So then maybe your [14:08] beautiful UI won't even be shown to [14:09] people. Like [14:10] >> [laughter] [14:11] >> How do you think of that? You know, like [14:12] as a designer, like do you design APIs [14:15] instead? Or [14:16] >> There's definitely a There's a pang of [14:18] sadness in there. [14:19] In the UI designer uh, part of me when [14:22] when I think about that for sure. Yeah. [14:23] Yeah. I I like [14:25] We honestly I think because of that [14:26] reason we fought it for a while. Like we [14:28] um, we were kind of like no, we we [14:29] really need to keep everybody in the [14:31] Granola UI for everything. [14:33] I think um, it's been really uh, it's [14:35] been really interesting working with [14:37] bigger and bigger enterprise customers [14:39] um, because I think just as soon as you [14:41] become a big enough company you have [14:43] it's inevitable you have all of this [14:44] internal tooling and you if Granola [14:47] doesn't play nice with that internal [14:49] tooling, like you kind of you're a [14:50] non-starter. Like companies like [14:52] and uh, yeah, I think you know, I think [14:54] we basically just like I think at this [14:56] point we're at peace with the fact that [14:57] like [14:58] for a bunch of work for best way to use [15:00] Granola is it to [15:02] use it to ingest the context, you know, [15:04] and use it during the meetings but then [15:06] it's like powering workflows off in all [15:08] kinds of other apps because it's [15:09] connected to your internal agent or Ford [15:11] or ChatGPT or whatever. [15:12] >> Yeah, exactly. Yeah. I mean there's [15:14] always going to be like more normal [15:16] users who just use the UI, right? But [15:17] like for for even as a solopreneur for [15:19] me like Granola has super useful context [15:22] but I have a bunch of contacts in Google [15:23] Docs and other sources." So, like, the [15:25] power is combining all that stuff [15:26] together. Like, helping the AI [15:27] understand everything. [15:29] >> Yeah. Yeah. Yeah, it is. It is. Like, [15:31] as it's um it's as [15:34] as the context you give it, right? And [15:35] the amount of connections and stuff you [15:36] have. Um yeah. Yeah, I'm excited to do a [15:38] little more of that. Like, with the [15:40] pre-meeting briefs was great for us [15:41] because it um I guess it it forced us to [15:43] build a bunch of infrastructure under [15:45] the hood to make Gronela good at sifting [15:47] through emails to find the right thing. [15:49] Um and uh yeah, I don't I don't know if [15:51] that's been built in a similar way so [15:52] that we can make that work for other [15:54] other tools, too. You know, Slack and [15:55] and CRMs and everywhere else a company [15:59] has like a a world of knowledge that we [16:00] could be looking at. [16:02] >> Yeah, dude, there's so much opportunity, [16:03] man. Like, you you can do like a weekly [16:05] brief, like weekly [snorts] meeting [16:06] brief, or like you can do like a [16:08] post-meeting inside brief. There's so [16:09] much. You can do so much briefs. [16:11] >> [laughter] [16:12] >> It'll be amazing. [16:15] >> The briefs is kind of like the um it's [16:16] like the obvious first use case or [16:17] whatever, you know? So, I feel like [16:18] there's there's a there's a risk of us [16:20] If you're the kind of person that's [16:21] going to try out the AI tools that then [16:22] you can write just draft briefs [16:24] from everybody left, right, and center. [16:26] Yeah, yeah. [16:27] That's not a a regular person's problem. [16:29] Like, yeah, I think [16:31] >> But, before we move on the product, [16:32] like, you you know what would be the [16:33] ultimate feature? The ultimate feature [16:35] request from me. You know [16:36] >> [laughter] [16:37] >> The ultimate feature request from me is [16:38] like uh this is probably bad, but like, [16:40] you know, you and I just make our own [16:42] like I have a Peter agent and you have a [16:44] Sammy agent, and then we we go send our [16:46] agents we go send our agents to have a [16:47] meeting first to align, and if they [16:49] can't align, then we come in. [16:51] >> [laughter] [16:52] >> That'd be the ultimate feature to do. [16:53] >> I feel like people are having a major [16:54] thing a lot. Like, um CEOs we get a lot [16:56] of like CEOs asking for that kind of [16:58] thing. Like, you know, can you can you [16:59] make Gronela be my first line of defense [17:01] from all the questions I [laughter] get? [17:04] Yeah, yeah. [17:05] >> [clears throat] [17:05] >> I think it's a it's a good idea. Yeah. [17:07] Yeah. [17:07] >> Yeah, there's there's something there. [17:08] Yeah. [17:09] >> [snorts] [17:09] >> All right. Well, I mean, now let's let's [17:11] let's kind of switch switch gears and [17:13] talk about kind of like how you guys [17:14] work internally. [17:15] >> Yeah. I'm kind of like a little bit [17:16] pessimistic about like this whole AI [17:18] cycle of thing. Like it feels like a lot [17:19] of the AI native companies are just [17:21] like, you know, working 996 and just [17:22] like [17:23] >> [laughter] [17:24] >> working non-stop, dude. And I feel like [17:25] as a parent, like do I really want that? [17:28] Uh but like I I I think Granola, maybe [17:30] you guys have a different philosophy [17:32] on how this stuff works. [17:33] >> Yeah. [17:33] >> Yeah. [17:34] >> I think on that front I really enjoy [17:35] being a Londoner, honestly. Like I think [17:37] um the team works a lot and and very [17:39] hard. I mean, they pour a lot of [17:40] themselves into it. But I think I don't [17:41] know. I [17:42] I think you should all you should have a [17:43] life outside of work and you should like [17:45] you should this just this just so much [17:46] more to life than than than writing code [17:49] or writing software. And then I don't [17:50] know. I think it's so important. Like I [17:52] think you the work you do is better when [17:53] you have some balance. [17:55] >> Mhm. [17:55] >> Working hard though. Like I feel like [17:56] that I'm I'm fully like addicted to the [17:58] slot machine of the coding agents. You [18:01] know, the [18:02] I'm like every every night I'm on my [18:04] laptop like, "Oh, one more problem. One [18:06] more problem. One more problem. One more [18:07] problem." So [18:08] Yeah. [18:08] >> Yeah. [18:09] >> Yeah. [18:09] >> Yeah, it's it's actually it's actually [18:10] cuz we talk about like Granola helping [18:12] with context switching with me meetings, [18:13] but like actually context switching [18:14] between the different agent threads is [18:16] even harder. Like you literally context [18:18] switch like every 5 minutes. [18:19] >> [laughter] [18:19] >> Yeah. [18:20] >> Yeah. [18:20] >> Yeah. [18:21] >> So [18:22] >> Yeah. [18:22] >> Yeah. [18:22] >> Yeah. [18:23] >> It gets pretty exhausting. [18:24] >> It's like it's the thing we're grappling [18:26] with at the moment, I think, is like [18:27] with designing Granola cuz we we want to [18:29] help Granola be better for you like [18:31] outside of meetings for [18:33] and um [18:34] the stuff you can use chat for is like a [18:36] little hint of that, but I think there's [18:37] a lot more we could do with especially [18:38] like um when you need to follow work [18:41] around meetings. [18:42] Um [18:43] once Granola is connected to a couple of [18:44] tools, [18:45] but really I think there is like if you [18:47] have like three things that come out of [18:49] an average meeting that you might need [18:50] to go away and do or think about and [18:51] then you have five meetings during the [18:53] day, it very very quickly becomes a big [18:55] big stack of things to do. [18:56] >> Yeah. [18:56] >> And uh how you like present that to the [18:59] user or like how Granola can help you [19:00] get through that stack of things in a [19:02] way that feels like uh it makes your [19:03] life nicer and better and and more [19:06] calming rather than just like adding [19:08] stress to like uh pressure and more more [19:11] and more to do on top of you is like a [19:13] hard thing to figure out. [19:15] >> Uh yeah, yeah, yeah, that's that's [19:17] that's right. Cuz yeah, if if you guys [19:19] if Granola has like three takeaways from [19:20] each meeting, but at the end of day [19:22] there'll be like 20 takeaways. [19:24] It would be too much. [19:25] Yeah. It would be too interesting. [19:27] >> We want to be a calming presence in your [19:28] life, not not the opposite. [19:30] >> Um and and and like how how do you guys [19:32] cuz I I feel like nowadays like any [19:33] engineer can just start shipping like [19:35] the the shipping stuff is not hard [19:37] anymore. It's easier than before. [19:39] Uh so in some ways like almost everyone [19:40] in the company can ship basically. And [19:42] and like now like you know as a designer [19:43] how how do you make the craft and [19:45] quality still good? Like it doesn't like [19:47] >> [laughter] [19:47] >> how do you avoid just like a bunch of [19:48] slop going to prod all all the time? [19:50] Yeah. [19:51] >> Yeah, it's uh it's hard like cuz there's [19:54] a lot of benefit to letting [19:55] uh like everybody in Granola is like [19:57] writing code shipping code in one way or [19:59] another. Like a lot of it's for internal [20:01] tooling not not into production, but a [20:03] surprising amount is in production as [20:05] well. And um [20:06] there's like some power in having [20:08] anybody be able to go and do that. Uh [20:10] and I want to encourage that and not [20:11] make design like a bottleneck for that. [20:14] Yeah, we like I guess um some things I [20:16] think we've done that I think really [20:18] helped there that um with the rebrand [20:21] that we did in in February, we did a [20:23] bunch of work to [20:25] overhaul the look and feel of the app, [20:26] make it feel in line with the brand, but [20:28] in doing that we also [20:30] built that out on a new design system. [20:32] And so there like everything behind the [20:33] scenes is tokens and components and uh [20:36] yeah, it's all it's all reusable. Not [20:38] just the human, but for agents to do it [20:39] as in B files all over the place to [20:41] teach the agent how to use the design [20:43] system. That's I mean that's helped a [20:44] lot for like stuff that's not super like [20:46] design heavy or doesn't need thinking [20:48] about super hard. Like it usually comes [20:49] out looking decent. [20:50] >> Uh okay. [20:52] >> Okay, just from the agent writing the [20:53] code. A lot of the I don't know, for [20:55] building enterprise features in the [20:56] settings menu like you can just ask [20:59] Granola to you can just ask the coding [21:01] agent to um [21:02] you know, add a add a new page in [21:03] settings that does X Y and Z and when [21:05] this toggle's turned off this then and [21:08] uh [21:08] does all the right things like comes out [21:09] with the right [21:10] >> It was the components. [21:11] >> Yeah. [21:11] >> Got it. [21:12] >> Yeah. Yeah. Yeah, for the designers, I [21:14] don't know. I feel like they're It's an [21:16] interesting time. I feel like [21:18] in real Figma had like a complete hold [21:19] on the whole design industry, you know, [21:21] for good reasons like it's just a great [21:23] tool. Um [21:24] I I don't know. Even for us, 80% of my [21:27] day was spent in Figma both like [21:29] catching ideas, iterating on different [21:31] approaches, and also like refining and [21:33] handing off designs. And I feel like [21:35] that's just all been thrown apart like [21:36] in the last year or something. And it's [21:38] honestly now like a I don't know all the [21:40] designers use all kinds of different [21:42] tools all day long. Like Like um we're [21:44] an in-person company. We do a lot of [21:45] like initial thinking on whiteboards [21:47] together in rooms. I mean sometimes like [21:48] a whiteboard sketch is enough now to [21:50] give to an engineer and they're able to [21:52] like cuz like the coding engine can [21:54] with the design system go and do the [21:56] thing. You know, like do the right [21:57] thing. [21:57] >> Yeah. [21:57] >> Yeah, and then and then I don't know [21:59] like [21:59] I have I have an example actually. We [22:00] have like some some mocks and stuff [22:02] premium briefs I could talk you through. [22:05] >> Yeah. I mean I mean how how how much do [22:06] you use Figma now? Is like 10% or you [22:08] just don't use it anymore? [22:09] >> Uh [22:10] 20% maybe yeah, something like that. [22:13] It varies. Like I I some designers on [22:15] the team will be a lot more than that [22:16] and I think it depends on the project [22:18] too. Like when [22:19] when Tanya was working on the a paywall, [22:23] you know, like subscription flows and [22:25] upselling you to the the business tier [22:28] and stuff like that. That's all like all [22:29] of the hard design stuff there is like [22:31] does this flow [22:32] make sense and does the copywriting [22:35] ideas in your head at the right moments? [22:36] And it's far faster to iterate on that [22:38] by just looking at it all in Canvas and [22:41] it's not [22:42] it's not heavy on like interaction [22:44] detail or like you know, there's not too [22:46] many states to have to worry about. It's [22:47] just like a linear flow with copywriting [22:49] being the hard thing. Um whereas premium [22:51] and briefs example is much more [22:53] interactive and [22:55] integrated in the core flow. And so it [22:56] was a lot less Figma and a lot more [22:58] prototyping in [22:59] in code for for premium and briefs for [23:01] example. [23:02] >> Yeah, makes sense cuz it's just a couple [23:03] bullet points, right? You got to make [23:03] sure the bullet points are actually [23:04] good. Yeah, so. [23:05] >> Yeah, yeah. And it's like where do you [23:06] where do you show them in the app and um [23:08] Yeah, so like this was a this was one of [23:10] Tanya's early prototyping um [23:14] explorations. Like I think we basically [23:16] went from like chatting about where the [23:18] where the brief could live in a room on [23:20] a whiteboard or something to this to to [23:22] her kind of like mocking this up. So [23:24] this was like [23:26] what if it's in the notification and uh [23:29] this, you know, like [23:30] >> Okay, got it. [23:31] >> from here. You can ask Cursor to be [23:33] like, "Hey, I want to I have this brief [23:35] to test these texts. Um can you go like [23:38] put this in the notification that [23:40] appears in the top right of the screen [23:41] before a meeting?" And can you also [23:42] please go and put it in the bottom of [23:44] the screen? Can you also go put it in [23:45] the floating transcription indicator on [23:47] the side of the screen? Do all three of [23:48] those things now, you know? And you get [23:50] like um instantly [23:52] get so much more signal on like is this [23:53] going to feel good or not? [23:54] >> Yeah, got it. [23:55] >> And then, you know, like [clears throat] [23:56] uh I this this pro this whole project [23:58] like flip-flopped between prototypes [23:59] like this and then back into Figma and [24:02] then prototypes again, back into Figma. [24:03] Yeah, it was really uh not at all [24:06] linear. [24:07] And that's more and more okay now. [24:08] >> Yeah, cuz you want to I guess you want [24:10] to get as close to the actual end user [24:11] experience as possible, right? So like [24:13] this is more it feels more interactive [24:15] than like static mocks. [24:16] >> Yeah. [24:17] Yeah, exactly. And and I think like [24:19] especially with um uh [24:22] Well, let's see if I go back to mine to [24:24] us, for example. Um like with something [24:26] like this, you know, it's it's pretty [24:27] simple. It's just it's just a bunch of [24:28] text. Uh and [24:30] there's there's a bunch of nuance in [24:32] like [24:33] how loud you make it and where do you [24:34] put it on the screen and whatever, but [24:36] really the thing you're designing for is [24:37] like how does it feel when you're [24:40] stressed and you're late and you're [24:41] arriving to this meeting and you've [24:43] forgotten who the person is to to like [24:45] receive these words in this in this way. [24:47] And um [24:49] it's very hard to get a realistic [24:51] feeling of that from looking at a [24:52] mock-up of like somebody else's brief, [24:54] you know, or like a made-up brief. Like [24:56] it really it really matters that it's [24:57] your data and that it's your stuff [25:00] you're looking at to make a judgment [25:01] call on whether it works or not. [25:02] >> Yeah, that that that makes a lot of [25:04] sense. Yeah. Yeah. Yeah, and and also [25:06] like how how how good the brief is also [25:08] matters, right? It's not just about the [25:09] UI. It's like how how good is the brief. [25:10] >> Yeah. [25:11] >> Yeah. Yeah, exactly. Yeah. Yeah. [25:13] Yeah. [25:13] >> Do Do you guys like like another way I [25:15] found to just like still let people ship [25:17] but still have higher quality is like, [25:19] you know, just like a ton of dog [25:21] food internally and like, you know, [25:22] showing to beta users and getting their [25:24] feedback before you do GA the product, [25:26] right? Like I I I assume you guys do [25:27] that, too. Yeah. [25:28] >> Yeah. Yeah. Yeah. Yeah, we do. We we [25:30] really like try and push to get like [25:32] when a project starts to get like a [25:33] shitty end-to-end working version of it [25:35] in the app as quickly as possible. Um [25:37] yeah, like we had brief we had a version [25:40] of this brief in the app within like a [25:42] within a week a week and a half of of um [25:45] starting the project. [25:46] >> Nice. [25:47] >> and it was incredibly ugly. Like it was [25:49] just in a big box at the top of the [25:51] notes and um [25:53] >> Yeah. [25:53] >> you know, the beta wasn't good and the [25:54] the yeah, it wasn't well written. But [25:56] like I I think you you want people to be [25:58] feeling the pain of this thing not being [26:00] good like day in day out. It's a very [26:02] It's a very effective way of like of you [26:04] know, making you give a about about [26:06] being good. [26:06] >> Yeah, you probably have a bunch of like [26:07] internal Slack channels giving feedback [26:09] and stuff like that, right? It's like [26:10] yeah. [26:10] >> Yeah. Yeah. Yeah. Yeah, it's Yeah. [26:13] >> [laughter] [26:14] >> Um [26:16] cool. Well, I mean, uh I I I guess maybe [26:17] you can show us one more thing. Like I'm [26:19] um you know, like uh Cloud Launch Tags [26:22] recently and people are saying like, oh, [26:23] you know, moving from the IDE to like [26:25] the app and now we're moving to just [26:26] like people giving commands to agents in [26:28] Slack. [26:29] And um and like I I think you guys have [26:31] a bunch of internal agents, right? Like [26:32] there's like um [26:34] >> Yeah. [26:34] >> Na- Na- natural or something or like [26:36] Honeybot. [snorts] [26:37] Yeah. Can you show us some of this? [26:39] >> Yeah. Yeah. Um [26:41] So, [26:43] uh [26:44] yeah, natural So, we have this whole [26:46] like internal dashboard thing. Um so, [26:49] this is [26:50] like all kinds of um features people go. [26:54] That's for all internal tools [26:56] all kinds of things. This is I'm I'm I'm [26:59] like really I'm really proud of [27:01] this. So, like I think it's [27:04] everyone on the team is adding stuff in [27:06] here and building it. And you can see [27:07] like we have this little edit with [27:09] cursor data like you can [27:12] turn this on and I'm like [27:14] make this red. [27:16] It like you can basically like point and [27:19] click be like hey, I need to change the [27:20] internal tool to go do this thing for [27:22] me. [27:23] >> Oh nice. Okay. [27:24] >> You got to get the spinner up cursor [27:25] agent to go do it. So, we really tried [27:26] to like like a bunch of people have put [27:28] in work to lower the bar to making [27:30] internal tools. Yeah. Natural is like I [27:32] think by far and away the most used [27:33] internal tool thingy we have. Um It's a [27:37] it's a chatbot here [27:39] and it's highly up with like connected [27:41] to [27:42] as many things as we can give it like [27:45] Slack, granola, amplitude, all of our [27:47] logs. [27:48] >> Okay. [27:49] >> website [snorts] [27:50] analytics like as many things as we can [27:51] as many things as we can. And then [27:53] people use it. I guess you can use it [27:55] here. It's it's mostly used in Slack. [27:57] I'll show some examples that hopefully [27:58] safe to share like look at the code, [28:00] figure out how the code works, give a [28:02] smart answer. Yeah, you know, and then [28:04] like okay. This this takes so long to go [28:06] actually takes them to do it yourself. [28:07] It's like often So, this was a minor I [28:10] think summary generation had a little [28:12] dip for a second or something like a [28:13] minor incident and [28:16] yeah, so we have like video into the [28:17] natural channels starts out. [28:19] Tells me also what's going on. So, this [28:21] is like what's it doing? This is [28:24] looking at logs I think to figure out [28:26] you know, where are these errors popping [28:27] up? What's going on there? It's then [28:29] able to then go like sense check that [28:30] against the code and figure out like why [28:32] would that error be happening? [28:33] >> I see. [28:34] >> Yeah, like this this [28:37] natural basically gets brought into [28:38] every like every question that pops up [28:40] is how many states uh [28:41] as the first protocol. And [28:44] yeah, I don't know. We're really [28:45] extremely mad at this point. [28:47] >> But [snorts] basically you just linked [28:49] it up to multiple internal company [28:50] sources and has a you you can it sounds [28:52] like it can pick the model they want to [28:53] use, but then like it it kind of does [28:55] the first pass response, right? And then [28:56] you can talk to it properly. [28:57] >> Yeah. Yeah. Yeah. And then you know, [28:59] it's it can it's it's really reliable at [29:01] SQL queries and things like that. So it [29:02] can pretty Yeah, you can pull data in [29:04] and it's going to be cheap like just [29:06] just way way faster than a human being [29:07] your dashboards would be able to. [29:09] >> Yeah. And and do you have like some guy [29:11] or some person work working on that? So [29:12] like a internal tools team or something? [29:15] >> Nobody owns it. Like I like um uh JK JK [29:18] one of our engineers I think started it [29:20] as a weekend or an evening project. Um [29:23] >> Yeah. [29:24] >> And no, people just add things here and [29:26] there to it like a um yeah, someone [29:29] like Nacho was incredibly popular and so [29:31] if you make Nacho better, you'll be [29:32] everybody's best friend. [29:34] So uh there's a strong incentive to [29:37] to [29:38] you know, if you if you've got 10 [29:39] minutes to try and figure out a way to [29:41] make Nacho more powerful for everybody. [29:42] >> Yeah, I I I feel like this is the right [29:43] way to do things, man, cuz I I feel like [29:45] um like I guess you can also just bring [29:47] Claude or you know, like Cursor in here [29:49] too, but like I I feel like companies [29:50] should just build their own internal [29:51] thing for that. It's not too hard to [29:52] build, I feel like and like you have [29:54] more flexibility. Like if if if [29:55] someone's model becomes better or like [29:57] the price gets too high or something, [29:58] you can just switch, right? [30:01] Yeah, it's like a snake. [30:01] >> And it's I mean we have um we have I [30:03] think Nacho can create Cursor [30:06] agents. So like it will often, you know, [30:08] like in an doing an investigation will [30:10] often lead into a PR being written to [30:13] fix the thing. And I think that's just a [30:14] great way to do lessons we added Cursor [30:16] and made that not two tool two different [30:17] tools. Yeah. [30:18] >> So I guess uh like I I guess just to [30:20] recap like, you know, how to be AI [30:21] native and and still give people a life [30:23] is is like [30:24] >> [laughter] [30:25] >> it's kind of like anyone can build, but [30:26] you have like design system and [30:28] components to make sure that at least [30:29] the the UI is like looks good like from [30:31] any engineer and then also having a [30:33] bunch of internal agents that's hooked [30:34] up to internal systems for people to ask [30:36] questions, right? That that that's kind [30:37] of [30:38] >> Yeah, I think so. Yeah. Yeah. And then [30:40] uh [30:40] yeah. Yeah, don't don't I don't know [30:43] and trust people to use their time as [30:45] they as you know like as they as they [30:47] wish. Like uh [30:48] >> Yeah. I mean yeah [30:50] like you know you're not trying to [30:51] you're not trying to make sure people [30:52] are working like you know 8 hours a week [30:53] or like a day or something. You're just [30:54] trying to see the output, right? [30:56] You look at the output. [30:57] >> Yeah. Yeah. [30:57] >> Yeah. [30:58] >> Yeah. Yeah, I really think like [30:59] especially maybe now like more than more [31:02] than in the old days like it's so easy [31:03] to make stuff. You know, it's a cliche [31:04] now but like it's so easy to make stuff. [31:07] The hot thing is knowing what to make [31:08] and like that kind of thinking comes [31:10] from like doesn't come from just like [31:12] grinding as hard as you can the whole [31:13] time. Like the the real interesting [31:15] stuff comes from I don't know going for [31:17] a walk or taking your time off or or [31:19] um [31:20] you know space to go and think about [31:21] something. So [31:23] Yeah. [31:24] >> That that part is still very human dude [31:25] because I I I [31:26] try to brainstorm like what to make with [31:28] like you know code codex or something [31:29] and then [31:31] like the AI is still too agreeable, man. [31:32] It is like you know, here's some ideas [31:34] and like this is a great this is a great [31:35] idea. Is it are you sure it's a great [31:36] idea? No, it's not a great idea. It is [31:38] like it's not the same as talking to a [31:39] human. [31:40] >> Yeah. [31:40] >> Yeah, not the same. [31:41] >> Yeah. Yeah. Yeah. [31:42] >> Well, well Sam this is this is a [31:43] pleasure man. Do do you have any like [31:45] closing words of advice for AI builders [31:47] or people trying to build just like AI [31:48] tools or companies? I guess it is about [31:50] like trying to figure out what you want [31:51] to build and maybe [31:53] >> I think I think the I mean yeah, the the [31:54] thing I the thing I really like [31:56] appreciate about what we have here which [31:58] I mostly credit Chris my co-founder with [31:59] instilling in me and in everyone else in [32:01] the company is like it's it's just you [32:04] know, it's so easy to build things [32:05] nowadays. Like one you got to build [32:06] something interesting. You know, a lot [32:07] of [32:08] I think figuring out what's going to be [32:09] the interesting thing. But two like most [32:11] people stop way too early I think when [32:13] in building stuff. Like they you know, [32:15] they get to the version one and they [32:16] have push it out the door and ship it [32:18] and I think the easier it gets to build [32:20] stuff the more it matters to like really [32:22] sweat the details and go the extra mile [32:23] to make something that's actually going [32:24] to be really useful to somebody. To hold [32:26] yourself accountable not just to [32:27] shipping something but to like does the [32:29] person actually get like daily value out [32:32] of it? Um [32:33] and to not stop until you achieve that [32:35] that goal rather than just the shipping [32:37] goal. [32:37] >> I I see. Yeah, yeah, yeah. Cuz you you [32:39] want to ship some stuff to get some [32:41] market feedback. You don't want to just [32:42] like plan it forever, right? You want to [32:44] ship. [32:45] >> Yeah, I know, yeah. [32:47] We encourage to like ship We ship stuff [32:48] internally like as quickly as we can, [32:50] but um we, you know, and and to like [32:52] small groups of friends and beta users [32:53] and things like that very very early, [32:54] but we do like I think we pull the like [32:57] a a very high bar for like is this thing [32:59] actually working for people, you know, [33:00] before we put stuff out often. [33:01] >> Got it. Well, well, [33:03] add me to the janky and the mobile [33:05] builds, man. I I I I think the problem [33:06] is it's too polished right now. Add me [33:07] to the janky and the mobile builds. [33:08] Yeah. We're not in the builds. Yeah. [33:09] >> [laughter] [33:10] >> Yeah, I know. [33:10] >> Yeah. [33:11] >> It's really fun. Yeah. Yeah. [33:12] >> Cool. Thanks so much, man. I I feel like [33:14] I learned a lot from this chat. Thanks. [33:15] >> This is really fun. Thanks. Thanks a lot [33:17] for doing it.