Google Labs Reshapes Product Playbook as AI Code Erases Software Moats

AI编码消除软件护城河，Google Labs重塑产品策略 由于前沿AI模型使代码几乎零成本，且容易被智能体在周末快速复刻，产品的持久竞争力如今取决于专有的用户情境与网络效应，而非孤立的软件界面。 Jaclyn Konzelmann详细介绍了像Gemini 4这样的前沿模型如何从根本上改变了Google Labs的产品开发模式，促使团队告别传统的40页PRD，转向快速原型制作、简明战略文档以及与最终用户的深度共创。她概述了Labs将叫停的实验视为学习资产而非失败的文化，其中成功孵化的能力会整合进Gemini应用等核心服务。讨论还探讨了在代码几乎完全免费的世界中产品的防御壁垒，认为真正的持久力在于上下文情境、网络效应，以及在消除用户繁琐操作的同时保留创作自主权带来的乐趣。最后，Jaclyn分享了她在家庭中如何利用AI将屏幕时间从被动消费转变为主动创造，与年幼的孩子共同制作原创播客、游戏和音乐。 Gemini 4已将Google Labs的核心产品考量，从某个想法在技术上是否可行，转变为它是否真正值得存在。 (0:00) Google Labs对AI生成的劣质文档执行严格规定：作者被禁止产出需要读者花费比作者撰写还要多时间去阅读的材料。 (7:48) 由于原始代码实际上已经免费且极易在周末被AI智能体搭建复刻，软件的长期持久力必须依托专有情境和网络效应，而非独立的UI。 (17:43) Google Labs的实验性项目在关停时不会被贴上失败的标签；代码、基础设施以及吸取的经验教训都会在Labs Live上向全团队共享，以此为未来的迭代奠定基础。 (12:53)

Jaclyn Konzelmann, Product Director at Google Labs, argues that advanced models like Gemini 4 have shifted product management from asking whether something is technically buildable to questioning whether it genuinely deserves to exist. Replacing 40-Page PRDs with Lightweight Artifacts (0:00) Because model capabilities now allow technical feasibility to be largely assumed, Google Labs has abandoned traditional 40-page Product Requirement Documents (PRDs). Instead, PMs anchor early initiatives around three concise artifacts: a strategy document averaging about one and a half pages that clarifies the core narrative, a presentation deck for formal reviews and casual hallway alignment, and an interactive prototype built to test real user excitement before committing production engineering resources. Filtering AI Slop and Cultivating Team Learning (7:48) To counter low-effort AI writing, Konzelmann enforces a strict standard: any document that takes longer to read than it took to produce is disallowed, and authors must be able to defend every line. Strategy documents originate via outdoor walking and voice memos rather than desktop typing, focusing on clarity over volume. Internal reviews demand iterative trimming, treating excessive length as a failure to refine the underlying idea. Experimentation, Feature Graduation, and Retiring Prototypes (10:45) Operating under a hierarchy of 'users first, Google second, Labs third,' Google Labs functions as an autonomous exploratory unit distinct from the core Gemini app team. Experiments regularly feed into core products—such as NotebookLM graduating into Gemini Notebooks, and the CC experiment evolving into the Daily Briefing. Projects are discontinued without stigma under the operational mantra 'keep the lessons, not the prototype,' sharing findings across regular forums like Labs Live and internal 'Methods and Metrics' (M&Ms) sessions. Defining Product Durability in an Agent-First World (17:43) With code becoming virtually free and easily replicable over a weekend, Konzelmann contends that software durability can no longer stem from the codebase itself. Instead, long-term defensibility relies on structural assets like data context and network effects. Furthermore, while automated agents can absorb administrative toil—such as repairing corrupted GPX files or handling routine forms—human users will still demand direct interfaces for creative exploration, learning, and entertainment. Eliminating the Blank Prompt Box for End Users (32:23) As user expectations shift from prompting individual tasks to managing end-to-end autonomous agents, the primary friction point remains the empty prompt cursor. For non-technical audiences, including small business owners, Labs aims to co-create specialized workflows rather than expecting users to master complex prompt engineering. The challenge for builders is designing scaffolded, intuitive interfaces that reveal agent capabilities without forcing users to navigate an unguided blank text field. While advanced models lower the cost of software creation to near zero, product durability hinges on identifying genuine human problems rather than shipping easily duplicated agent wrappers. Builders must navigate evolving social norms around synthetic delegation while replacing blank prompt boxes with intuitive, scaffolded user experiences.

Google Labs 产品总监 Jaclyn Konzelmann 指出，随着 Gemini 4 等前沿模型大幅降低编码和技术实现门槛，产品管理的核心挑战已从“能否做出来”转向“该产品是否值得存在”，团队工作流正全面转向更轻量、重体验的原型迭代与真实共创。 淘汰长篇PRD与抵制空洞AI文本 (0:00) 在 Gemini 4 时代，Jaclyn Konzelmann 放弃了早年长达40页的产品需求文档（PRD），改为由高阶战略文档、演示文稿和交互原型组成的三轨工作流。高阶战略文档聚焦于产品是什么、为何要建以及为何当前重要；演示文稿用于正式评审及茶水间或走廊中的即兴探讨；可交互原型则在正式工程投入前快速验证用户兴奋度与假设。由于生成式 AI 能在瞬间产出长达十页的内容，团队的难点反而转变为精简思考、提炼核心叙事。 为防止战略文档充斥空洞的 AI 拼凑内容，Konzelmann 践行一项明确规则：如果读者阅读产出物所花费的时间超过作者制作它所花费的时间，该文档便不被允许存在。作者必须能为文档中的每一个词进行辩护并展开实质探讨。目前她的战略文档通常控制在一页半左右，超出这一篇幅往往意味着核心思考尚未打磨透彻，文档会被退回重写。此外，语音转录已取代打字成为她理清思路的主要手段，她常在园区散步时通过语音记录想法，再整理成战略初稿。 用户共创与破解空白提示词陷阱 (4:02) Google Labs 秉持“用户第一、Google 第二、Labs 第三”的三元理念，从新模型能力与人类新兴行为模式两个维度评估产品构想。Konzelmann 强调产品经理必须走出“象牙塔”，避免预设中小企业（SMB）或大众用户懂得复杂的提示词工程或终端交互。以面向 SMB 的 Pomelli 项目为例，许多小商户没有时间从零调试复杂 Prompt，团队直接与营销人员、商户及受信任测试群体合作，甚至在 Mountain View 与纽约团队的线下活动中邀请本地商户进行实时测试，以发掘底层痛点。 针对当前大模型产品的落地瓶颈，Konzelmann 认为单纯告知用户模型功能强大远远不够，核心痛点在于“空白提示词输入框”（blank prompt box）。面对空无一物的光标，普通用户往往不知从何问起。产品设计的真正职责不是单纯给用户他们口头索要的表层功能，而是通过直观且富有巧思的交互界面，主动显性化模型价值，实现意料之外的惊喜。 代码贬值下的持久护城河与界面重构 (17:43) 随着 AI 生成代码令软件编写成本趋近于零，周末套用几个 Agent 就能复现的功能性产品正变得极易被淘汰。Konzelmann 提出，真正的产品持久性不再依赖代码本身，而取决于私有数据上下文与网络效应等结构性优势。在 Agent 优先的时代，甚至需要重新审视产品是否还需要图形前端（UI）。 对于传统用户界面的存续，Konzelmann 给出了清晰的判断标准：人类只有在享受创造、学习和娱乐过程时，才会持续需要丰富的交互界面；而在处理报税、PDF 整理或修复 GPS 轨迹等日常琐事（toil）时，用户会极乐意将其全权交给后台 Agent。产品设计应当在后台吸收这些摩擦力，同时保留带给用户乐趣和社交连接的触点。 保留经验而非原型与组织反哺 (10:45) Google Labs 与负责主界面的 Gemini App 团队保持分工：独立团队专注核心应用，Labs 则承担面向未来的探索与发明。Labs 的实验成果通过多种路径反哺核心产品，例如 NotebookLM 衍生为 Gemini App 内的 Gemini Notebooks，先前的 CC 实验演化为每日简报功能（Daily Briefing）并进一步拓展为 Family CC。由深耕该领域的项目团队自主决定实验是毕业晋升、转向还是关停。 为了克服产品经理对项目的心理依恋，Labs 建立了“保留经验，而非原型”（keep the lessons, not the prototype）的文化认同。关闭实验不被视作失败，团队会在名为“Labs Live”的全员会上复盘分享发现，或在定期的“M&Ms”（Methods and Metrics）研讨会上交流方法，其沉淀的底层架构、代码和洞察会被后续项目重新利用。产品副总裁 Josh Woodward 倡导正向与自主的探索氛围，促使成员将精力聚焦于用户价值与产品学习，而非内部考核政治。 AI家庭教育实践与制造必要摩擦 (24:00) 在家庭教育中，面对一个由 AI 抹平所有难度的世界，Konzelmann 对年幼的孩子采取了双轨教育策略。一方面，她刻意让孩子体验户外无聊并制造摩擦（manufactured friction），以锻炼他们应对现实挫折的情绪韧性；另一方面，她将 AI 作为实现无限想象力的引擎，让孩子从被动的屏幕内容消费者转变为主动的创作者。 具体实践中，她让五岁的女儿担任原创故事播客的“创意总监”，由孩子构思角色与剧情，利用 ElevenLabs 及 Gemini 语音模型挑选声音，并通过图像生成模型设计视觉，最后交由 Agent 混音并发布至 Spotify。此外，她还利用 Gemini Argon 通过自然语言提示词单次生成弹珠轨迹视频及互动式 3D 动物园小游戏。这些实践表明，当数字内容能够被真正公开发布并与同龄人分享时，AI 工具便能有效建立起下一代的主体创造意识与现实行动力。 当前用户对 AI 的预期正从单点任务自动化转向端到端雇佣完整 Agent，然而在大模型能力急剧膨胀的背景下，如何帮助主流用户跨越空白光标的门槛、在消除琐碎事务的同时保留人性化的体验，仍是产品团队在下一代前沿模型演进中需要持续探索的课题。

[0:00] Jaclyn Konzelmann: as the models get more powerful, it's really no longer a question of can the models do it or can we build it. It's a question of is it worth existing? Should we build it? [24:00] Jaclyn Konzelmann: In a world where things are easier and easier to do, I want to make sure that my kids still have that muscle to endure friction because there will still be frictional points that occur in your life.

[0:00] Jacine, thanks so much for joining me [0:01] today. Really excited to talk to you. [0:02] And for those of you who don't know, [0:04] Jacqueline is a product director at [0:06] Google Labs, arguably one of the fastest [0:08] moving teams inside Google. Is that is [0:09] that fair to say or? [0:11] >> We do value our ability to move fast and [0:13] get things into users hands. [0:15] >> Awesome. Awesome. So, let's talk about [0:16] Gemini 4 is is a big launch that [0:18] happened, you know, a week or so ago. [0:20] And how has it kind of changed your [0:22] workflow as a PM and as a team? [0:25] >> Yeah, [0:25] >> you know, I would I would say it's kind [0:27] of changed it in two big fundamental [0:28] ways. The first is just expanding the [0:31] surface area of what's possible to build [0:33] and what's possible to exist and what's [0:35] worth existing these days. And I think [0:37] that's taken a moment for us to really [0:40] recalibrate as a product team and just [0:42] think about what are the things that are [0:43] worth building. So there's a lot of time [0:45] we've been spending around just the [0:46] ideation phase and trying to plan out [0:48] what's next and figure out what can we [0:50] now achieve in this space. And then the [0:52] other is the actual day-to-day tactical [0:54] parts of the how do we do our actual [0:56] job. And on that front, a couple things [0:58] that we've kind of shifted towards I've [1:00] noticed. Number one, with a model like [1:02] this and as the models get more [1:04] powerful, it's really no longer a [1:06] question of can the models do it or can [1:08] we build it. It's a question of is it [1:10] worth existing? Should we build it? [1:12] >> And so instead of starting with what I [1:14] often refer to as the 40page PRD, of [1:17] which I have written many of those early [1:18] in my days as a PM, what I'm often times [1:21] finding is better to start with is more [1:23] of a highle strategy doc. something that [1:25] really asks the question on what are we [1:27] building, why are we building it, why is [1:29] it important, ask the bigger questions [1:30] and gets everybody aligned on the [1:32] opportunity space. In addition to that, [1:34] I find it helpful to create a [1:36] presentation cuz sometimes you want a [1:37] document for offline reading. Other [1:39] times you're actually presenting in [1:40] front of a small group and that could be [1:42] in an actual product review or it could [1:44] just be like bumping into somebody in [1:46] the MK or in the hallways. I know we had [1:48] lunch earlier and like you just see the [1:50] people that are walking around the [1:51] office and sometimes I like pulling open [1:53] a presentation and like bouncing ideas [1:54] off of folks that way. And then the last [1:57] one is a prototype before you actually [1:59] want to go seriously in depth and build [2:01] a production grade anything. Building a [2:03] prototype is just the fastest way to [2:05] gauge excitement. So those are kind of [2:07] like the three things I'm finding are [2:09] where the product journey starts these [2:10] days. [2:11] >> Awesome. And labs is like kind of the [2:13] experimental group in Google, right? [2:14] trying to prove our ideas and you have [2:16] products for SMBs, you have products for [2:18] consumers and all kinds of crazy [2:19] products. Like how do you just how Yeah. [2:21] Like you're totally right. It's not [2:22] about building, it's about deciding what [2:24] to build. So do you guys have some sort [2:25] of methodology to decide what to build [2:27] or is just kind of like, [laughter] you [2:29] know, what's the most inspiring idea? [2:31] Yeah, [2:31] >> I think it's kind of, you know, both an [2:33] art and a science. I would say we have a [2:36] couple great internal documents that [2:38] Josh has talked about many times as [2:40] well. We have some labs guiding [2:42] principles. We also just have this [2:43] general philosophy that we want to build [2:45] users first, [2:46] >> Google second, labs third. And so we [2:49] look out in the world and we say what is [2:50] worth existing at this point? What do [2:52] users want or what do we think that [2:54] they're going to want? [2:55] >> And then we figure out is this something [2:56] that makes sense for Google to be [2:58] building and then is it something that [2:59] labs should be actively contributing to? [3:02] But I think when it comes to new ideas, [3:04] there's usually like two lenses I look [3:06] at them at. One is what is this emerging [3:09] model capability that we're starting to [3:11] see and does that open up a new type of [3:13] product that we should be building as a [3:15] result of that or does it cause us to [3:17] revisit old ideas that didn't make sense [3:19] before that are now viable and then the [3:21] other is what are the new user trends [3:23] we're sensing in the world. I think as [3:25] we're all adapting to how AI is changing [3:27] our day-to-day life, it actually causes [3:29] new habits for people to create and that [3:33] can open up new surface area or new [3:35] white space for products or solutions to [3:37] exist. And so like good oldfashioned [3:39] userdriven product development, but with [3:41] the lens of not trying to see what [3:44] people are necessarily all doing today [3:46] and more about predicting where you [3:48] think things are going and taking a few [3:49] bets on that line. Maybe we could talk [3:52] about like a specific project maybe like [3:54] Pomet or whatever project you want to [3:55] talk about and talk about you know how [3:57] did you decide to build for this group [3:59] of people and what kind of problems you [4:00] were solving for them. [4:02] >> Yeah, I think that one was definitely [4:04] more it was a combination of both of [4:06] things coming together. We saw that [4:07] these image generation models were just [4:09] getting a lot better at being able to [4:12] create onbrand content. I remember [4:14] seeing even just like on my ex feed all [4:16] of these crazy elaborate prompts to get [4:18] image models to do things, but it [4:19] resulted in something super cool. But [4:21] you also know that like your average SMB [4:24] owner, of which my parents both ran [4:25] their own small businesses while I was [4:27] growing up, they don't have the time to [4:29] be constantly tracking everything that's [4:30] happening and figuring out how to use AI [4:33] and prompt the models in these specific [4:35] ways and, you know, every time starting [4:36] from a blank slate. And so in that case, [4:38] it was definitely about, hey, this [4:40] model's capable of doing these things. [4:42] you know, [4:42] >> we know there's a group of users that [4:44] could really benefit from that and we [4:46] think that there's a different product [4:47] shape that we can form around it. And [4:49] then you start to go a little bit deeper [4:50] and you realize, hey, to generalize this [4:53] capability and apply it to any user is [4:55] very different than just like a one-off [4:56] prompt that somebody spent a lot of time [4:58] curating. And so then that gets you into [5:00] the whole idea of, well, what if we [5:01] really understood your business so that [5:03] when we did want to generate things for [5:04] you, it was a lot easier for the user [5:06] and they could kind of be involved as [5:08] much or as little as they wanted into [5:09] those workflows. Yeah, that was kind of [5:11] the genesis of a lot of thinking on that [5:13] one. [5:13] >> And yeah, like I feel like the number [5:15] one mistake that I think Bay Area PMs [5:17] make is just like wearing our ivory [5:18] towers here. It's very nice here. And [5:20] then you know, we just think everyone's [5:22] going to use these terminal interfaces [5:23] and like these like complex agent [5:25] interfaces to prompt like a huge prompt. [5:27] And I think what I really love about [5:28] Pomei is like how it makes it really [5:30] accessible to people who are less [5:31] techsavvy and people who actually have [5:33] other things going on. Right. [5:35] >> Yes. I'm I'm curious. How much do you go [5:36] off and talk to a bunch of SSMBs and [5:38] local restaurants or how did you how did [5:40] you decide like you know how did you [5:41] kind of put yourself into their shoes? [5:42] >> Yeah. No, we were very big on [5:44] co-creation. [5:45] >> Okay. [5:45] >> In labs and in this case we talked to [5:48] both people that worked in the marketing [5:50] sector as well as businesses. [5:52] Absolutely. We work with a small group [5:54] of them. If you look back on all the [5:55] launches we've had we almost always [5:57] feature a brand and that's somebody that [5:59] we've worked closely with. But we also [6:00] have groups of trusted testers and [6:02] before we launch anything we do a [6:04] certain amount of user research and [6:06] bring in actually one of my favorite [6:07] stories is we'll occasionally have [6:10] on-sites with the team and this team [6:12] happens to be split between Mountain [6:13] View and New York and the team's [6:15] actually going to New York next week to [6:17] do a like group onsite there and part of [6:20] that workshop is going to be bringing in [6:22] a few small businesses and actually [6:24] sitting with them and using the tools [6:26] and showing them the stuff that we're [6:27] building and getting feedback directly [6:29] from them because we think it's super [6:30] important that it's not just our opinion [6:32] that's impacting the product. Like we [6:34] want the users who are using it. [6:35] >> Yeah. [6:36] >> Helping us co-create and figure out what [6:37] to build and what's working and doing [6:39] that iterative loop to make sure that [6:40] what we're throwing out there isn't just [6:42] a reflection of what's technically [6:43] possible, but is presented in a way [6:45] that's actually accessible to the users [6:47] that we're going after. [6:48] >> Yeah, I really believe that's probably [6:49] my number one calling as a PM just to [6:51] the co-creation thing. [laughter] [6:53] and like you know getting them into the [6:54] community and like just talking to them [6:55] every day showing all the designs and [6:57] stuff and just like you know just like [6:59] it's kind of unfair you just get the [7:00] best ideas from the actual customer like [7:02] you know why why not [7:04] >> I think it's interesting because that is [7:06] so important and then the hard part now [7:08] is also you hear everything they want [7:10] >> but don't just give them exactly what [7:12] they want like think about what the root [7:13] problem is they're trying to solve and [7:15] then solve it in a way that they [7:16] probably wouldn't have thought to solve [7:18] it and that's when you get that like [7:19] delight factor the second somebody uses [7:21] it you kind of walk away being like, [7:22] "Oh, I didn't think that this was [7:24] possible, [7:25] >> but this is actually exactly what I [7:26] wanted." Like, I think that's what we're [7:27] always trying to strive for. [7:28] >> That's right. That's why you need both. [7:29] You need both your input and also like [7:31] kind of your understanding of what the [7:32] model's going and all that kind of [7:33] stuff, right? [7:34] >> Exactly. [7:35] >> I'm curious with Gemini 4 and everything [7:36] else. You talk about not writing 40page [7:39] PRDs and thank God, but like is there [7:42] still like a lot of like, you know, [7:44] strategy documents or what whatever. Is [7:45] there still like a lot of reviews and [7:47] approvals and like you know let's check [7:48] the boxes and everything or is it like [7:50] you know a little bit more nimble now? I [7:52] know Josh is very into like moving fast. [7:54] >> Yeah, [7:55] >> definitely into moving fast. I would say [7:58] it's less about having to check the [7:59] boxes, but I do want to be careful here [8:02] because writing those strategy docs I [8:03] think still does take a lot of time. [8:05] >> In fact, I think the person writing it [8:07] should sit with it for a while and make [8:08] sure that what they're actually coming [8:10] up with is worth other people's time to [8:12] read. [8:12] >> That's right. That's right. And so I [8:13] think a lot of the work that happens is [8:15] now just on that initial phase of like [8:18] honing your thoughts and making sure [8:19] that your strategy doc is not, you know, [8:21] 10 pages cuz it's so easy to write 10 [8:24] page documents these days. Like trim it [8:26] down. Figure out what the narrative is. [8:27] Figure out how to position it in a way [8:29] that resonates with folks. And then once [8:31] you get that going, like once you've [8:32] really unlocked a good idea, it's easy [8:34] to get people on board and build. [8:37] >> This episode is brought to you by [8:38] Whisper Flow. Whisper Flow saves me at [8:40] least three hours a week and is one of [8:42] my favorite AI apps by far. It's just [8:44] [music] so much faster to dictate to AI [8:46] using your voice than to type. You just [8:48] talk naturally and it outputs clean, [8:51] ready to send text. [music] [8:52] Whisper Flow even removes filler words [8:54] and formats your sentences for you. I [8:56] use Whisper Flow for everything [8:57] including drafting newsletter posts, [8:59] writing product specs, replying on [9:01] Slack, and more. It works on Mac, [9:03] Windows, iPhone, and Android across all [9:05] of your favorite apps. Try a free at [9:07] whisperflow.com and use my code [9:10] peterwisperflow to get six months free. [9:13] That's peter whisperflow. Now back to [9:15] our episode. Have you like come across [9:18] we don't have to name any names but have [9:19] you come across like these dogs where [9:20] like you read halfway through and it's [9:22] like oh this is written by AI like this [9:23] like [laughter] do you have like a you [9:26] know principle on that? Like you have to [9:27] be able to maybe it's written by AI but [9:28] you got to be able to defend the idea. [9:30] >> Yeah I would say in the early days this [9:32] was definitely more common. I think [9:33] people are starting to adjust adjust and [9:35] adapt, but I have a few rules at this [9:38] point. And one of these actually I got [9:40] from talking to somebody during an SF [9:41] tech event last week, but I just love [9:43] the way they summarized it, so I'm going [9:45] to recote them, which was basically if [9:47] it takes somebody longer to read the [9:50] thing you produced than it took you to [9:51] produce it, [9:52] >> not allowed. [9:53] >> Like a I think that just shows you [9:55] didn't even take the time to read it [9:56] yourself. And so that's like a big red [9:57] flag in my books. But then the other [9:59] thing I would say is you should be able [10:00] to actually defend every single word [10:02] that you've written out. And you should [10:04] be able to have an active conversation [10:06] about the idea. And if somebody reads a [10:08] doc and they get frustrated because it [10:10] feels like it's too lengthy or not like [10:12] compact enough, I would push it back on [10:14] them and say, "Hey, can you take another [10:15] try at writing this?" And I think it [10:18] just it takes a bit of a muscle to learn [10:20] how to do that and to not rely on AI, [10:22] but like to use it as a thought partner, [10:24] but you really want your thoughts coming [10:26] out in what you produce at the end of [10:28] the day. [10:28] >> Yeah. Like you should be able to you [10:29] should get the message across in like, [10:31] you know, like one page basically, [10:33] right? You don't have to write like six [10:34] pages or anything like you know. [10:35] >> Yeah. I keep calling it a one pager. I [10:37] feel like mine average one and a half [10:38] these days, but like if it's getting [10:39] beyond that, then you probably haven't [10:41] honed the idea enough to effectively [10:43] communicate it. [10:44] >> Got it. Okay. All right. All right. [10:45] Well, let me ask you maybe some harder [10:47] questions. Let me ask you for labs. [10:49] There's a lot of experimentation going [10:50] on. One argument you can make is like, [10:53] you know, it's great there's also [10:55] experimentation, but why does the team [10:57] just focus on making the Gemini app [10:59] better or something like that? Or maybe [11:00] let the experimentation actually lead [11:01] the way to make the Gemini app better. [11:03] >> You know, I think one of the great parts [11:05] of Google is that we have a lot of [11:07] amazing teams that are working on [11:08] different products. And so we have an [11:10] awesome Gemini app team that is working [11:12] on that product. And then we have a [11:14] bunch of people in labs whose goal is to [11:16] try to [11:17] >> invent and build for the future. And [11:19] sometimes it is actually a great story [11:21] that's something that we've been working [11:23] on helps make the Gemini app better. And [11:25] that could be the actual product or [11:27] experiment that makes its way into it. [11:29] It could be just the learnings that [11:31] we've had that help influence how the [11:32] rest of us are thinking. So you see a [11:34] lot of stories like that where something [11:36] from labs might help influence the [11:37] Gemini app, but also just the way we're [11:40] trying to, you know, carve out the space [11:42] in the future, I think generally helps [11:44] overall with just the mindset and the [11:45] culture and making sure we're all [11:47] thinking with that that point of view. [11:48] >> Yeah, I think notebook LM is now part of [11:50] the Gemini app, right? [11:51] >> Yeah, you have Gemini notebooks in the [11:53] Gemini app. You have CC was an [11:55] experiment we launched last year [11:57] initially and that kind of helped turn [11:59] into the daily briefing that's in the [12:01] Gemini app. And then that's actually a [12:03] great story because I recently launched [12:04] a family CC as well. So that kind of [12:06] shows [12:07] >> there's many paths to success for what a [12:09] lab's experiment is and I could probably [12:11] list out like five or seven other ones [12:12] at this point too. [12:14] >> How do you guys decide whether a lab [12:15] experiment, you know, should be wound [12:17] down or like mix it into Gemin [12:20] case by case? [12:21] >> I think it always ends up being case by [12:23] case what makes the most sense. But more [12:25] often than not, it's the team itself [12:27] that's building it that happens to know [12:29] what the right path is for that. And [12:31] that comes down to we want the people [12:33] working on these so deeply embedded in [12:35] the problem space that they're trying to [12:36] solve. And it's it's them that kind of [12:39] help us figure out, all right, like [12:41] where is this going? Does it grow? Does [12:43] it graduate? Does it, you know, reh [12:45] harvest into the next great thing? But a [12:47] lot of just autonomy and relying on the [12:49] fact that these teams know what's best [12:51] for the products and experiments they're [12:52] working on. [12:53] >> I see. Interesting. But like I feel like [12:54] as a PM it's kind of hard to kill your [12:55] own product, right? Or like you know [12:56] like I don't want to give up on it. Like [12:59] do you guys have some lab principles or [13:00] like to kind of encourage you know [13:02] objective thinking around this stuff? [13:04] >> Yeah, I mean I think it's definitely a [13:05] muscle that grows with you [13:07] >> the more you're in labs and that is part [13:10] of the culture that we really try to [13:12] curate here. We have a few fun sayings [13:14] that are helpful to point back to. I [13:16] think I've said one already like build [13:17] to learn. We just very much believe in [13:19] building to learn. We also another [13:22] favorite saying of mine is keep the [13:25] lessons not the prototype. [13:26] >> Like you learn so much as you're [13:28] building things. And I think a key thing [13:30] is never look at an experiment ending as [13:34] a failure. [13:35] >> Like you always take something away from [13:37] it and it's helpful for the teams to do [13:39] a reflective piece on like what are all [13:40] the lessons learned? How can we share [13:42] that out with the rest of labs? So we've [13:44] also done things like we have our labs [13:47] live which is our whole team meeting [13:49] that happens periodically and during [13:51] some of these we'll invite the [13:53] experiments up that have since moved on [13:55] to come and talk about all of the things [13:57] that they learned and also invite other [13:59] people in labs who might now be [14:01] exploring a similar area but with a [14:03] different lens or a different like point [14:04] of view on how it might be worth solving [14:06] it to talk to the team like how do we [14:08] continue to pass on the learnings as [14:10] well as sometimes it's just the [14:11] infrastructure and like the actual code [14:13] that was created. is still able to be [14:14] reused in other areas. But I think just [14:17] celebrating learning is a really key [14:19] part of it. [14:20] >> Yeah, I think that's like really [14:21] important like the culture and the [14:22] incentives like because people want to [14:24] build their careers and stuff. You got [14:25] to make sure that [14:27] >> like I always believe that if you want [14:28] to build a career idea you you don't [14:30] have to think about all the career [14:30] performance stuff. You just focus on the [14:32] product and make it good [14:33] >> or or like you know learn something and [14:35] then you build a career naturally, [14:37] right? [14:37] >> I very much agree with that. [14:39] >> Yeah. But I think as big companies it's [14:40] kind of hard to like just refocus on [14:42] that. you kind of have to encourage [14:43] that, you know, behavior. [14:44] >> Yeah, you definitely have to encourage [14:45] it. I think, you know, a lot of credit [14:47] to our VP Josh who has just done such a [14:49] great job helping to create this culture [14:51] and this, you know, natural curiosity. [14:54] And then what I found is like you tend [14:55] to attract the types of people that do [14:58] actually thrive [14:59] >> in these areas. And so like the more [15:01] culture and the more you prioritize [15:03] building for that team like the more [15:06] you'll be able to continue to like [15:07] recruit and the right people tend to [15:09] find you which then helps like in a [15:10] positive way to continue on those [15:12] traditions. [15:13] >> Yeah. He kind kind of sets the example [15:14] to a certain extent, right? I think he [15:16] started here as an intern or something, [15:17] right? [15:17] >> I think so. Yeah, he's been here a [15:19] while. [laughter] [15:20] >> Yeah. Yeah. Let me ask you a question [15:22] about Josh. Like he seems like such a [15:24] nice guy. What happens when he actually [15:25] gets a little bit disappointed in you? [15:27] like does he like get mad or what? Give [15:29] them some feedback or [15:30] >> he's just such a positive person. I [15:32] think [15:32] >> what he does is he tries to empower the [15:34] team. [15:35] >> And I mean like I think if you love the [15:37] products that you're building like [15:38] you're going to be the ones that hold [15:40] yourself to the bar that you want to [15:42] ship at and that's kind of the culture [15:44] we've tried to create. So I've [15:46] >> Yeah. like it's I don't know how else to [15:49] describe it except that I sometimes find [15:51] like I'm my own harshest critic [15:54] >> and that as a result means that like I'm [15:56] always going to try to do things and if [15:58] things don't go or if things take a [16:01] different turn you learn something and [16:02] you're like oh this thing I thought was [16:03] going to work out actually there's a [16:04] whole bunch of challenges it's a safe [16:06] place to talk about it and then like [16:08] we're all we're all in this together so [16:10] like how do we course correct like how [16:12] do we move on to something different I [16:14] think there's always a positive way to [16:16] look at situations because at the end of [16:18] the day we're all here to try to build [16:19] great things for the users. [16:21] >> Yeah. And then it's not like you know [16:22] it's not like life or death or anything. [16:23] It's like we're trying to you know [16:25] >> we're building Yeah. We're building [16:26] great products, great experiences and [16:28] everybody is motivated to make that [16:31] happen. So [16:32] >> if anybody ever like similarly is [16:35] >> has a question or like feels stuck on a [16:37] team, there's always people in labs that [16:40] are there to help and you know share [16:43] what they've learned about a space or [16:44] like provide input on like how they [16:46] might think about it. And actually it [16:48] was great earlier today. We do a lot of [16:50] these also internal uh meetings within [16:54] the team to try and just knowledge share [16:56] cuz there's so many of us building so [16:57] many interesting things. So we call one [16:59] of them M&M's which are I think methods [17:01] and metrics and you'll have a team that [17:03] will come and present something that [17:04] they've done and it's a great way to [17:05] just knowledge share out and a great [17:07] chance for them to also let other people [17:09] see what it is they're building. So [17:11] that's one example of you should never [17:13] feel like you're stuck or you don't know [17:15] how to solve a problem. Like I always [17:17] say go to AI to help brainstorm in that [17:18] but also go to the people that you're [17:20] surrounded with and that's something [17:21] that we really tried to cultivate in [17:22] labs. [17:23] >> Yeah, that's what I actually really miss [17:24] after doing my job. Like it's it's fine [17:26] brainstorming with different models, but [17:28] it's not the same as brainstorming with [17:29] actual smart humans. It's just not the [17:30] same, you know? [17:31] >> I know. Well, that's why that's why I [17:32] wanted to hang out with you today also, [17:34] and that's why I try to make it a point [17:36] to not just talk with people at Google, [17:38] but also just like continuously go to [17:39] different industry events and just like [17:41] bounce ideas. I think there's a lot to [17:42] be said about that. [17:43] >> Have you with all this like I'm I'm [17:45] assuming everyone's like, you know, [17:46] using anti-gravity or using AI like all [17:48] the time. I know in the engineering side [17:50] like the problem becomes the bottl net [17:52] becomes like people are almost pushing [17:54] too much code and then like you know you [17:56] get AI to do all the code reviews and [17:58] then you don't even know what the hell's [17:59] going on anymore if you're not careful [18:00] like I'm curious on the prod side is [18:02] there something similar like that is [18:03] there like a lot of has a bottleneck [18:05] shifted [18:06] or [18:07] >> I think [18:09] >> what it's really shown is we need to [18:11] think about what types of products to [18:13] build [18:13] >> okay [18:14] >> differently [18:14] >> you can just build 100 different things [18:16] >> yeah like it's so like you can just [18:17] build it doesn't I mean you should be [18:18] building but it's also what is the value [18:20] of a product I think is shifting [18:22] especially as we're moving towards this [18:24] agent first world where now you have [18:26] users that aren't just people but [18:27] they're agents and is it the agent that [18:29] belongs in the product is it a consumer [18:31] agent that's using a product on behalf [18:32] of a user all of that is in flux at the [18:35] moment and it really just makes you want [18:38] to realize that you should be building [18:40] as we say like you want to build [18:42] products not features and I think a lot [18:44] of things that used to be products [18:45] before could be easily reproducible [18:47] throw a couple agents at it over a [18:49] weekend. So, that's not a durable [18:50] product. [18:52] >> What is that's the question that we're, [18:54] you know, bringing up more and more [18:55] these days. [18:56] >> Do you have any do you have any initial [18:58] thoughts around that or [18:59] >> I mean, I think there's some good [19:00] frameworks to look at. You know, network [19:03] effects and, you know, context is [19:05] interesting to think about. So, I think [19:06] there's like a few common ways to slice [19:10] the problem. But I also think just a lot [19:13] of that is changing in real time. And [19:14] that's why I mentioned around like what [19:16] are new user patterns that are emerging [19:18] and like if you play that out and assume [19:21] a future that you know executes on that [19:24] like what new problem spaces open up and [19:26] then you can figure out products that [19:27] might fit in there but also figure out [19:30] what is the durability of them and the [19:32] code itself is not the durability of [19:34] them. Yeah, code is going to be like [19:35] kind of just abstraction, right? Like I [19:37] think I don't know. I feel like as a [19:39] I've I've also been a PM for a long time [19:41] and I've always prided myself in like [19:43] you know caring about the UI and like [19:45] making the user experience great and [19:47] kind of stuff and now we're moving to a [19:49] world where I think San Francisco and [19:51] the Bay Are these people are just using [19:52] their agents like they're even looking [19:54] at UI anymore. They're not they're not [19:56] going to your website or download your [19:57] app. They're just like using some API or [19:59] something. I think it really makes you [20:00] question why do you want a person [20:02] >> in your product? Like is your product [20:05] something that needs a UI? And I do [20:07] think in some cases it absolutely is. I [20:09] think there's a lot of things that [20:10] people will want to still be involved [20:11] in. Anything that involves me creating [20:14] something, I have a lot of fun doing [20:15] creativity projects with my kids. [20:17] Anything where I want to learn [20:18] something, there might be a reason for [20:20] me to interact. Anything I want to or [20:22] anytime I want to be entertained or [20:23] consume something, I'm still the one [20:25] using the product. But you're right, [20:26] there are definitely products that I use [20:28] today that I would be more than happy to [20:30] not use or at least not be the main [20:32] person interacting with the interface, [20:34] but perhaps the underlying capability or [20:36] what that product or service offers is [20:39] something I would still want. [20:40] >> But that's what I mean where I'm like [20:41] there's a lot to be thinking about [20:43] differently when it comes to building [20:45] products. [20:45] >> There's definitely like a ton of like [20:47] just like manual like just filling out [20:48] an online form or like do some PDF or [20:51] tax document or something that's that's [20:53] like just stuff that we don't want to [20:54] do, right? So I think agents can help a [20:56] lot of that and especially like SMBs [20:58] that you were building for Pomei, they [20:59] have a lot of paperwork. [21:00] >> Yeah. [21:01] >> All the toils and tasks as we like to [21:03] say. [21:03] >> Yeah. So I definitely use AI to handle [21:05] all that stuff. But then at the same [21:06] time sometimes when I like give feedback [21:08] to someone like I won't mention what the [21:10] lab is but like the lab and then their [21:12] agent respond to me instead of the human [21:14] and then I get a little bit offended. [21:16] I'm like hey why did you respond to me [21:17] yourself? Why do you get your agent to [21:19] respond to [laughter] me? So it's kind [21:21] of like a double-edged sword cuz I'm [21:22] using the stuff myself too you know. I [21:24] mean, I think that's a great example of [21:25] user behavior is shifting [21:27] >> right now. And I think we're right in [21:28] the thick of it at the moment. And it's [21:30] interesting to think about how this all [21:32] plays out and what are the products and [21:34] experiences that should exist once we've [21:36] all kind of gotten used to this whole [21:37] new set of capabilities and how we want [21:39] to integrate them into our lives. [21:41] >> That's right. Yeah. [21:42] >> Yeah. I'll give you another like quick [21:43] silly example. I I were chatting [21:46] earlier. So, I have three kids and we [21:48] love hiking in the mountains and I took [21:50] my daughters up hiking near Yusede and I [21:53] recorded it on Straa and my three and a [21:56] half-year-olds, this is the first time [21:57] they did this Yusede hike all on their [21:59] own and it was like a pretty lengthy [22:00] hike and at the end of it I was really [22:02] frustrated because although it was all [22:04] stored nicely in Strava, there was this [22:05] like weird jut out on the GPS that was [22:07] just like a glitch. Okay. [22:08] >> And in order to fix it, I did look into [22:11] it like you have to download the actual [22:13] GPS like upload. You have to like splice [22:15] the section that doesn't work. You have [22:16] to like stitch it back together. You [22:18] have to reupload. It's like this huge [22:19] manual process. And like that's the [22:21] perfect thing where it's toilome. I want [22:23] to just throw my agent at it. And full [22:25] disclosure, I absolutely did throw my [22:26] agent at it. It fixed everything. [22:28] >> But I do still like seeing a feed of [22:30] what my friends have accomplished. [22:31] >> So it's not that I don't want to use [22:32] Strava. It's just I didn't want to use [22:34] it in that particular case. And as I [22:36] look at products and what you're [22:37] building, there's questions around like [22:38] what are the toilsome things that users [22:40] might need to do? How do you make them [22:42] not have to do that? And then how do you [22:43] play into more of what they might want [22:45] to be spending their time doing in your [22:46] product? [22:47] >> Yeah, we were just talking about during [22:48] lunch about how I think both our [22:50] daughters like to paint and draw. [22:52] >> Yeah. [22:52] >> And like you can easily imagine a future [22:54] like even like a year from now where you [22:57] mostly focus on her painting and you [22:58] just fire some commands to the agent to [23:00] do your work. [laughter] You go back to [23:01] painting or like you know DD these days [23:03] when I work a lot of times I'm like [23:04] actually outside just talking to the [23:06] agent and I I think it's like a big [23:08] quality of life improvement to just like [23:09] stare on my screen the the whole time. I [23:12] am such a big fan of voice. Like that [23:14] has become my primary way of getting [23:16] thoughts out of my head and into [23:18] whatever I happen to be using. Like you [23:20] should see my voice memos at this point. [23:22] It is just and I love going for walks [23:24] too cuz you're right, it is a quality of [23:26] life improvement where I can literally [23:27] like do laps around the building and [23:29] just like talk to offload my thoughts [23:31] and then I get back and I'm like, "Okay, [23:32] now I can massage this into exactly what [23:34] I want to turn it into." Most my [23:36] strategy docs start off this way [23:37] actually. [23:38] >> Yeah. Cuz your mind's clear when you're [23:39] outside and just ram blank, right? just [23:41] yapping. This is [23:42] >> scary. [23:42] >> 100% 100%. [23:44] >> But then maybe our neighbors are like, [23:45] "Who heard these weird people just like [23:46] yapping to their like talking to [23:48] themselves?" [23:48] >> You know, I've had that thought more [23:49] than once. But it's okay cuz I feel like [23:51] people have always talked into their, [23:53] you know, headphones when they're on [23:55] phone calls. [23:55] >> So now I just assume they think I'm [23:57] talking to somebody [laughter] on the [23:58] other side right now. [23:59] >> That's so funny. That's funny. Yeah. [24:00] Well, I guess now let's move to like the [24:02] family side of the interview. So, you [24:04] know, that's pretty like techsavvy, I [24:06] guess. Parents, like how are you [24:07] thinking about raising your girls, you [24:09] know? Yeah, that's a great question. I [24:11] know it's it's on a lot of parents' [24:12] minds these days. [24:14] >> There's a couple things I'm trying to be [24:15] really deliberate with. [24:16] >> Okay, [24:17] >> first, as much as I love AI and lean [24:19] into it, and we'll talk a bunch about [24:21] that, I do think it's still incredibly [24:23] important that my kids get outside and [24:25] that they learn how to be bored. I talk [24:26] a lot about trying to also manufacture [24:28] friction for them because I think it [24:30] just teaches you skills that are [24:32] necessary in life and that will be [24:34] helpful in life. And in a world where [24:36] things are easier and easier to do, I [24:38] want to make sure that my kids still [24:39] have that muscle to endure friction [24:42] because there will still be frictional [24:43] points that occur in your life. [24:45] >> But then when it comes to how I [24:47] incorporate AI into my day-to-day with [24:49] them, for context, my eldest just turned [24:51] five and my twins are three and a half. [24:53] I want them to grow up in a world where [24:55] they think anything is possible. [24:57] >> Like I want to use AI to make their [24:59] wildest imaginations come to life. Mh. [25:01] >> So, a couple side projects I started [25:03] with my eldest. We started a scary [25:06] stories podcast because she loves scary [25:08] stories and we started running out of [25:10] interesting podcasts to listen to on [25:11] this. So, I was like, well, why not try [25:12] to make one with her? So, she helped me [25:14] literally choose out all of the voices. [25:16] We had some fun doing like voice design. [25:18] I was using 11 Labs for this one. [25:20] Although the new Gemini TTS model, I've [25:22] been using that also. And like voice [25:23] design is fun to play around with. But [25:25] she chooses all the voices. She names [25:27] the characters. I then use an image [25:29] genen model to generate different images [25:31] of the characters and like she gets to [25:32] help me refine them. She gets to help me [25:34] come up with the story lines for each [25:36] episode. And then once we get the first [25:38] version of the podcast done, she listens [25:40] to it and she tells me what she likes [25:41] and if there's any changes she wants to [25:43] have made and so then I'll send it off [25:45] to my agent. It'll remix everything and [25:48] once her and I are happy with it, it's, [25:50] you know, now posted and live on Spotify [25:52] for others to listen to. [25:53] >> Spotify. [25:54] >> Oh, you want to know why I did that? [25:55] because initially I was starting to make [25:57] music with her using flow music and we [26:00] made this really funny song about banana [26:02] slugs after going camping in Redwood [26:03] National Park [26:04] >> and one of the first things she asked me [26:06] was it real? [26:08] >> I was like what do you mean it is it [26:09] real? Like we're sitting here in the car [26:10] listening to a song we just made like [26:11] yes it's real. She's like no but like [26:13] can my friends listen to it? [26:15] >> And I was like to her real is a shared [26:17] experience with her friends. [26:19] Spotify [26:24] and she'll grow up in a world where [26:25] stuff's [26:27] also sharable and shared with people. [26:29] >> That's super cool. Yeah. [26:30] >> Yeah. So, that's been one fun one. And [26:32] then I've been having a bunch of fun [26:33] with Argon over the past few days. Also, [26:36] we built my kids love marble runs, the [26:39] videos of them. And my husband likes it [26:41] cuz intuitively he's like they're [26:42] learning physics and he's really big [26:44] into physics. And so I had Argon [26:47] literally oneshot a marble run video for [26:50] them that they got to see. And next I'm [26:52] going to like customize it with them so [26:53] they can choose the colors they want and [26:55] feel like they're shaping it. And then [26:56] this morning I because I'm a mother of [26:59] three young kids, I wake up randomly [27:01] early. And so as I was getting ready to [27:03] come to work today, I was like, "Oh, [27:04] what would be really cool is if I could [27:06] build an interactive like petting zoo [27:09] game for my kids." And so in a couple [27:11] prompts, I ended up building this like [27:14] whole interactive kind of game that you [27:16] can walk around. It has six different [27:18] animals and they're very like esoteric [27:20] animals that the model decided to put in [27:22] this petting zoo, but you can walk [27:24] around and as you approach them, it [27:26] tells you a fun interesting fact about [27:27] them. [27:27] >> Oh, that's really cool. Yeah. [27:28] >> And so then that gets you into, well, [27:29] how are my kids going to learn as they [27:31] get older? And like, can we create these [27:33] bespoke experiences for them? Yeah. [27:35] >> And it's just really fun to see them [27:36] grow up like with this being a normal [27:39] thing for them. [27:40] >> Yeah. Kids are kind of spoiled these [27:41] days, right? Cuz like when we have to [27:42] learn, we have to read like dry texts. [27:44] But now you can just make like, you [27:45] know, clips and videos and games of [27:47] everything. You can learn everything [27:48] through that. [27:49] >> I mean, any way that gets my kids [27:51] excited about learning, I am more than [27:54] more than happy to indulge that side of [27:56] them. [27:57] >> Yeah. I think all your stories share a [27:58] commonality and like I think it's [28:00] important to try to get the kids to go [28:02] to creation as opposed to consumption. [28:04] >> Yes. You know, like like when my kids [28:06] watch like these like shorts, I'm like [28:08] it's like very brainwashing. But like [28:10] when they actually make their own games [28:11] and actually you don't actually need to [28:13] type to make games, right? You just like [28:15] speak to a phone or speak to the [28:16] computer and they'll make a game for [28:17] you. [28:18] >> And when you do that kind of stuff, I'm [28:19] like if you make your game and you make [28:20] it good, you can have like unlimited [28:22] screen time in like making it good. You [28:23] know, you can just do whatever you want. [28:25] >> Honestly, I think it just connects [28:26] different parts of your brain when [28:27] you're actively involved in the creation [28:29] point. And like 100% agree with like if [28:32] you're being productive and you're [28:34] learning and you're challenging [28:35] yourself, like that's what I want to see [28:37] them using AI for, not just like as a [28:39] way to make their life easier right now. [28:41] There will be plenty of time to smooth [28:43] those things out later. For now, I'm [28:44] like I want you actively learning and [28:47] being curious about it. And then also [28:48] expanding your mind cuz if you're [28:50] growing up thinking that all of this is [28:52] just normal and natural, like you said, [28:53] we grew up in a very different era. like [28:55] I can't wait to see what the next [28:57] generation comes up with and like what [28:59] they bring into the world. [29:00] >> Mhm. I also think like it's very [29:02] important to yeah to pursue your [29:03] curiosity and also show that you can [29:05] turn your curiosity into like a real [29:06] career you know like when I was younger [29:09] like I I make games on like Apple 2 or [29:11] whatever like the old Apple games [29:12] [snorts] but I always thought like it [29:14] was like not a real career. I was like [29:15] okay you know I have to go get a real [29:17] job. I have to go become like a [29:18] investment banker or something. But I [29:20] actually want my kids to like not only [29:21] explore their creativity but also like [29:23] you know we're talking about like [29:24] actually make some money online and see [29:26] how it goes so that they can actually be [29:28] ambitious and pursue whatever they want [29:29] to do you know. [29:30] >> Yeah. I think it's important for them to [29:32] connect the dots on right now your kids [29:34] are a little older than mine. So I'm [29:35] like curiosity is the main thing that [29:36] I'm pushing them to think into. But [29:39] absolutely how does that tie into the [29:40] real world and like where do they want [29:42] to spend their time? What do they want? [29:44] And how do they help create the life [29:46] that they want to live at the end of the [29:48] day? um is it's important lessons. [29:50] >> Let me wrap up with this. Now that we [29:52] have, you know, Gemini Argon and I'm [29:54] sure have a bunch of like Gemini 5, [29:55] Gemini 6 internal models going on, what [29:57] are you most excited about, you know, [30:00] building labs or with your kids or what [30:03] are you really excited about? Yeah. [30:05] >> Okay, I'll start with the kids side [30:07] because that's easy and then I'll answer [30:08] the lab side. With my kids, I'm just [30:10] like I am a neverending faucet of ideas [30:12] of things I want to be building with [30:14] them. And I'm really excited as they get [30:16] a little bit older for them to be even [30:18] more active participants in the ideas [30:20] that we're making. And I just like [30:22] there's so many things I want to run [30:23] home and just start building with them [30:25] right now. Time is like my limiting [30:26] factor these days. [30:28] >> And then in labs, I just this next class [30:31] of models that feels different. The [30:34] possibilities are so much bigger. And [30:36] I'm really excited that we get to be in [30:38] a place where we can think about what [30:41] that future is and start working towards [30:44] it. [30:45] >> Yeah. [30:46] >> And I'm just so grateful that like this [30:47] is my job and I have an awesome team [30:50] that's also thinking about these really [30:52] interesting hard problems and coming up [30:55] with like some really creative [30:56] solutions. like the best solutions and [30:58] the best ideas I find somehow connect [31:01] the technical capability with that user [31:03] need but come at it from an angle that [31:04] you wouldn't expect and kind of make you [31:07] go huh that's interesting [31:08] >> interesting [31:09] >> I kind of like it and I'm seeing more of [31:12] that coming out these days. So yeah, [31:14] just generally I think it feels [31:16] different right now and I'm excited for [31:18] all the new stuff that we have cooking. [31:19] >> Okay, we're going to keep it competition [31:21] until it's actually [31:22] >> Yes. [laughter] [31:23] >> Okay, got it. Yeah, I think I think [31:25] almost like any kind of persona you have [31:28] like any kind of customer you're [31:29] building for I almost feel like these [31:30] models at a point where they can almost [31:33] kind of you hire a person, you just like [31:34] solve your entire problem for for you, [31:36] right? It's not like it's not like doing [31:37] a task anymore. It's just like you just [31:38] hire a person to just take care of this, [31:40] you know? I just that's the way I'm [31:42] treating a these days. I'm like instead [31:44] of like doing a customer support task, [31:46] I'm hiring a customer support agent just [31:49] take care of everything. Yeah, I think [31:50] it's it's an interesting behavior shift [31:53] even in the way you're talking about it [31:55] right now versus what likely were [31:58] talking about AI a year ago, even [31:59] myself. And you're right, I think like [32:02] us [32:03] >> as users are we're changing our [32:05] expectations. They're rapidly shifting [32:06] and the way we're thinking about how we [32:08] want to apply it into our lives is [32:10] changing. I will say I think there's [32:12] different users and they all have like a [32:13] diversity of what it is they want but [32:16] user behaviors are definitely shifting [32:18] and that's been something to calibrate [32:20] against constantly as we're thinking [32:22] about those new ideas. [32:23] >> Awesome. Yeah. And I think just to wrap [32:25] up I feel like part of our job or like [32:27] part of part of like maybe labs in my [32:29] opinion is like just teaching users who [32:31] are less techsavvy that they can [32:32] actually ask a lot from these models. [32:34] Like you can actually ask a lot and they [32:36] will just do it for you. [32:37] >> Yeah. And I think it's more than just [32:38] asking a lot. It's like how do we show [32:40] them? How do we give them the [32:41] experiences that they want? Cuz when you [32:43] ask, that's exactly where people get [32:44] stuck is they don't know. Yes, the blank [32:47] prompt box problem basically. How do you [32:49] solve that in a delightful way so that [32:50] people are getting the most out of them [32:52] cuz there's so much value that can be [32:54] gotten out of them. [32:54] >> Awesome. Awesome. Well, where can people [32:56] find you online or you know? [32:58] >> Yeah, absolutely. You can find me on X [33:00] at Jackalu. Don't ask. There's a [33:02] backstory behind that [laughter] one or [33:04] on LinkedIn as well and Substack for [33:06] that matter. [33:06] >> Awesome. Thanks so much for your time. [33:08] Yeah, [33:08] >> thank you. Blessed and pray.