Model ML Targets Wall Street With Vision-Model Office Suite

Model ML携视觉模型办公套件瞄准华尔街 创始人认为，打造一家十人规模的十亿美元公司需要高强度的每周七天工作制，并让开发人员直接深入客户现场以绕过企业官僚体系。 在本次专访中，多次入选YC的联合创始人讨论了其代理式AI平台Model ML的飞速发展与企业采纳情况，该平台实现了复杂金融分析师工作流的自动化。他们分享了关于保持竭尽全力的创业工作伦理、从B2C转型至B2B，以及平衡兄弟合伙人职责等方面的核心见解。此外，他们回顾了以往创业项目Fancy和Fat Llama中关于坚持不懈的深刻教训，并极力主张早期开发者应当深度融入旧金山独特的创新生态之中。 由两位曾成功创立并出售两家 YC 公司的校友创立的 Model ML 正在经历垂直式爆发增长；该平台作为专为金融服务打造的 AI 工作区，能自动化完成分析师 90% 到 95% 的手动报告编制工作。 (0:00) 随着视觉模型的深度整合，系统能够从复杂的财务表格和图表中精准提取结构化数据，其准确度已超越传统人工处理水平。 (6:44) 金融机构的采购行为发生根本转变，AI 跃居高管层的头号优先事项，促使原本试水性质的采购快速转变为多年期、全公司层面的核心平台部署。 (10:49) 创始团队坚持高强度的每周六至七天工作制，并调整招聘重心，优先看重文化契合度、对独立项目的热爱与同频的奋斗精神，而非光鲜的过往履历。 (14:24) 回顾此前打造 Fat Llama 和 Fancy 的历程，两位创始人指出寻找 PMF 的路径各不相同，并凭借极强的韧性与面对面定制演示，成功在金融 B2B 市场建立起深厚的客户信任。 (16:56)

Founders Arie and Chaz Englander argue that rapid advancements in agentic AI and vision models have turned financial software from a historically difficult enterprise sell into a top C-suite priority. Drawing on their previous consumer startup exits, they combine an intense work ethic and lean team structure with consumer-grade user experience to automate complex Wall Street workflows. AI Workspace Adoption on Wall Street (0:00) Model ML operates as an AI-driven workspace featuring proprietary versions of Word, PowerPoint, and Excel built over an agentic architecture that mirrors an analyst's digital access across emails, CRMs, vendor feeds, and regulatory filings. The founders claim the software automates 90% to 95% of standard reporting tasks, generating formatted financial briefs in firms' shared drives immediately after public earnings releases. While enterprise financial institutions historically resisted early-stage software, the founders state that AI has become a primary executive mandate, shifting sales discussions from isolated pilots to multi-year enterprise contracts. According to the founders, Model ML signed as many contracts in a single week as it did throughout the entirety of Q4, expanding into approximately 10% of the world's largest private equity firms and investment banks. They attribute this shift largely to advancements in underlying vision models, which they assert now extract and process data from complex charts and financial tables more accurately than human contractors at traditional data vendors. Applying Consumer Product Principles to Enterprise Software (25:09) Having previously built the peer-to-peer rental platform Fat Llama and the delivery service Fancy, the Englander brothers apply consumer metrics—specifically direct user engagement and retention—to business-to-business software. To ensure product usability, the founders avoid detached software demos, preferring to sit side-by-side with finance analysts during trials to observe their real-time workflows and tool interactions. The founders maintain a deliberate goal to build a 10-person billion-dollar company, arguing that expanding headcount often introduces layers of management that disconnect engineers from end users. Looking ahead, they anticipate that enterprise workflows will shift from manual trigger actions to background autonomy, where software completes multi-step analysis tasks overnight before employees log in. Operational Intensity, Complementary Roles, and Cultural Hiring (14:24) The co-founders divide operational responsibilities along clear disciplinary lines: Arie oversees engineering and product based on his computer science background, while Chaz manages finance, commercial operations, and product from an accounting background. Both emphasize that radical transparency and complementary interests prevent co-founder conflict, which they identify as the leading failure mode for early-stage companies. Their hiring approach prioritizes cultural alignment, shared work ethic, and intrinsic curiosity over traditional CV credentials, screening engineering candidates by evaluating personal side projects. The founders report working seven days a week for 18 months, with their broader team working six days. They view relentless execution as essential during rapid inflection phases, drawing on early crises at their previous companies—such as manually tracking down a stolen £1,800 drone for Fat Llama and handling telephone transactions during a multi-day Stripe suspension at Fancy—to illustrate the necessity of logical perseverance. Navigating Ecosystem Trade-offs Between London and San Francisco (41:58) Although Model ML maintains its engineering base in London, approximately 80% of its customer base is located in the United States, prompting the founders to divide operations between New York, London, Hong Kong, and San Francisco. They observed that a significant share of global decision-makers for financial technology implementation reside in the Bay Area rather than traditional financial centers like New York or London. The founders contrast the high peer intensity of San Francisco against the hiring landscape in Europe. While they regard San Francisco as uniquely supportive for ambitious founders, they note that local engineering talent is prohibitively expensive and highly contested. In contrast, the UK provides high-caliber engineering talent with significantly lower hiring competition, though the founders find it more challenging in Europe to recruit workers willing to sustain intense startup schedules. The founders maintain that early-career builders should embrace high-risk entrepreneurship to accelerate learning and product impact. However, their operating approach relies on continuous seven-day workweeks and direct founder oversight, leaving open whether an ultra-lean team can indefinitely sustain multi-year enterprise deployments across global financial institutions without expanding its operational footprint.

Model ML 创始人 Arie 与 Chaz Englander 兄弟指出，随着视觉与推理模型的成熟，华尔街顶级金融机构对 AI 的采购态度已从试点转向全员铺开；初创企业应将 2C 产品的体验追求引入企业级软件，并依靠小规模团队与高度同频的极客文化快速抢占市场。 重构办公套件与垂直增长轨迹 (0:00) Model ML 作为入选 Y Combinator（YC）2024 年冬季批次的企业，将自身定位为服务于金融行业的“AI 工作区”，自主开发了对标 Word、PowerPoint 和 Excel 的专用套件，并结合模仿人类数字访问权限的智能体认知架构，直接集成企业内部文件、邮件、CRM、公开披露及 FactSet 等付费数据源。此前分析师需要耗费数日手动抓取数据并制作演示文稿，而该系统可在财报发布后自动生成格式规范的多页报告并同步至 SharePoint 或 Google Drive，直接完成 90% 至 95% 的工作量。 两位创始人此前曾创办并出售 Fat Llama 与 Fancy 两家 YC 孵化公司。凭借消费级产品对留存率和真实使用情况的敏锐度，他们观察到企业客户出现从“技术概念验证”向“实际采购使用”的关键转折。Model ML 在短短一周内签署的合同数量相当于此前整个第四季度的总和，客户群体已涵盖全球约 10% 的顶级私募股权机构和投资银行，以及主权财富基金、资产管理公司和风险投资机构，且企业高管往往绕过小团队测试直接签订多年期全员合同。 视觉模型落地与企业采购心智重塑 (6:44) 底层模型的持续进步极大降低了产品迭代成本，尤其是视觉模型的成熟打破了传统 OCR 的局限，使系统能够直接解析复杂的财务图表与多维表格。创始人强调，在数据收集和结构化呈现方面，模型展现出的准确率已经超越了数据供应商雇佣的人工团队。更深远的变化在于销售环境：以往 YC 经常告诫创业者投行、私募及律所极难采购外部软件，但如今 AI 已经跃升为这些机构首席执行官层面的头号战略议题。 为了赢得高风险偏好行业中高管的信任，两位创始人采取高频现场拜访的策略以降低决策者的职业风险。公司目前在伦敦、纽约设立办事处，并在印度、香港和新加坡部署团队，两位创始人分别驻扎纽约和伦敦，分头覆盖北美占 80% 的客户群体。他们指出，在金融软件采购决策中，面对面的现场演示不可或缺，即使是经验丰富的创始团队，也必须依靠逻辑推演而非盲目坚持来推动交易落地。 两段创业经验：从三年破局到闪电退出 (16:56) 兄弟二人此前的创业历程展示了截然相反的两种产品市场契合点（PMF）探索路径。2017 年创立的 P2P 租赁平台 Fat Llama 提供带有保障机制的个人器材互租服务，但在早期天使轮融资中遭到近千名投资人以“设备必被盗窃”为由的拒绝。在平台首笔 600 美元租赁中，价值 1,800 英镑的无人机险遭租客侵吞，创始人通过应用内置的经纬度搜索数据定位到伦敦北部一所敞开大门的房屋，亲自取回设备并伪装成官方取件服务交还失主，该业务耗时三年才最终验证了单位经济效益。 相比之下，2019 年底启动的垂直一体化即时杂货配送 Fancy 仅用 4 至 6 周便完成了 MVP 开发，通过自建前置仓向学生配送平价日用品，上线即迎来爆发，并在成立 18 个月后被美国行业巨头 Gopuff 收购。即便在早期遭遇 Stripe 突然断联支付服务三四天的危机，团队仍通过电话收款、货到付款和直信 Patrick Collison 渡过难关。早期亲自完成前 1,500 笔配送的经历，让他们坚定了“构建者必须紧密接触一线用户”的原则，并立志将 Model ML 打造为仅有 10 人的十亿美元级精干企业。 兄弟合伙契约与极端工作强度筛选 (14:24) 在创业合作机制上，兄弟二人形成了明确的权责分工：具有计算机科学背景的 Arie 负责工程与产品开发，拥有会计背景的 Chaz 统管财务、商业运营与产品，二者的重叠交集仅保留在客户互动与产品定义层面。他们认为，创业失败的主要诱因是合伙人沟通不良与信任缺失，直系亲属间“毫无滤镜”的坦诚交流有效规避了内耗。 在高强度运转下，创始团队已连续 18 个月保持每周 7 天的工作节奏，普通员工目前保持每周 6 天的工作制。这也促使他们彻底推翻早年依赖名校名企简历的招聘模式，转向以文化契合度和真实技术热情为核心的筛选标准。在面试工程候选人时，他们重点考察其工余时间自发编写代码（如 vibe coding）与个人项目的经历，借此筛选出真正享受创造过程、能够适应初创期极端节奏的核心人才。 伦敦与旧金山的生态反差与地域抉择 (41:58) 在探讨欧洲与硅谷的创业生态时，创始人指出两地在工作文化上存在巨大鸿沟。旧金山弥漫着一种在健身房跑步机上边跑步边开视频会议谈融资的极端拼搏氛围，而欧洲社会普遍缺乏这种无休止的进取心，导致在欧洲招募愿意承受超常工时的员工极为困难。尽管欧洲和英国拥有极具性价比且竞争相对平缓的优秀工程人才，旧金山的人才成本高企且争夺激烈，但硅谷的整体创新协同效应依然不可替代。 此外，创始人发现全球金融科技落地决策者中，有相当大一部分实际驻扎在旧金山湾区，而非传统认知中的纽约或伦敦。他们给年轻创业者的核心建议是尽全力搬到旧金山，或者至少进驻具备高密度雄心勃勃人群的一线城市；在 20 岁出头阶段全情投入创业的下行风险极低，即便失败也能收获密集的经验积累，而大公司的晋升通道未来始终敞开。 Model ML 的经验表明，当前金融 AI 的突围关键在于兼顾消费级交互质感与企业级底层安全，但创始人将全员超负荷工作视为竞争壁垒并追求极度扁平的 10 人规制，在跨国扩展与更复杂的全自动无感交付落地过程中，其可持续性与组织容错率仍待时间检验。

[6:44] even if we do nothing, theoretically our product will improve [11:30] If something logically makes sense, you should probably continue doing that thing, right? And not let anything stop you. [14:24] to be honest I didn't really have a clue what I was doing [14:24] it's suddenly... working but you haven't won the market yet [14:24] we just don't like losing [19:58] we're going to the moon [19:58] The lender: This is awesome, man [23:20] We got hit with... a lot of cricket bats to the face [25:09] one of the first... 10-person billion-dollar companies [25:09] making money... just gets boring so quickly [28:27] no money that can buy those moments [33:53] Paul Graham: If you're 26 and you're poor, that will be the worst outcome. And probably most 26-year-olds don't have any money anyway. So it doesn't make a big difference. [37:45] Everything that you have learned for 5 years try and unlearn immediately [37:45] Michael Seibel: What the hell are you doing? Like, let's talk weekly if not daily, if not hourly. [41:58] the best place in the world [262:56] frankly want to work this hard

[0:00] In the last seven days, we've signed the [0:02] same number of contracts as we signed in [0:03] the whole of Q4. There is clear tangible [0:06] value being driven by these products, [0:07] and it's only going to get better and [0:09] quickly. [0:10] >> Ultimately, you've got to be very [0:12] passionate about what you're building. [0:14] You've got to have that perseverance. [0:15] And if that sounds good to you, then [0:16] then build a startup. If something [0:18] logically makes sense, you should [0:20] probably continue doing that thing, [0:22] right? And not let anything stop you. [0:25] And I think like the consistency that [0:26] we've noticed of founders that we've [0:28] some of that we've invested in or work [0:30] with is like the ones that kind of do [0:32] that and really persevere tend to win. [0:35] [Music] [0:39] Today we're here with Arie and Chaz [0:41] Englander. Uh they are the founders of [0:43] Model ML uh from Winter 24. Uh prior to [0:46] Model ML they started two other YC [0:48] companies that both were successful and [0:50] sold, Fancy and Fat Llama. And this is [0:53] probably the first time I've worked with [0:54] a company where both of the founders had [0:56] had a previous successful uh YC company [0:58] before. So I'm super excited to welcome [1:00] Chassen Arie here to YC. Uh welcome [1:03] back. [1:03] >> Thanks for having us. [1:04] >> Thanks very much. [1:04] >> Thanks for having us. [1:05] >> Tell us what you guys are building. [1:07] >> So Monal is an AI workspace for [1:10] financial services. So that's our [1:11] oneliner. What that actually means in [1:13] practice is we've so we built a [1:15] workspace that's akin to kind of the [1:17] office suite. our own version of Word, [1:20] PowerPoint, and Excel with the major [1:22] difference that it's built on top of an [1:24] agentic system that kind of mirrors what [1:26] a human has access to at the firms we [1:29] work with. So quite specifically, if [1:31] you're a human at firm X, right, you [1:34] will have access to your files and [1:36] folder systems, your emails, your CRM, [1:39] any data vendors that you might used and [1:41] pay for, uh, real-time publicly [1:43] available information, public filings, [1:45] your internal custom data sets, etc. [1:47] Right? So then we kind of build this, we [1:49] call it a cognitive architecture. It's a [1:50] fancy word of saying, you know, kind of [1:52] like a brain that mimics what you have [1:54] access to digitally digitally. um uh and [1:58] we overlay that with our user interface. [2:00] Um the general idea being well, if you [2:02] had an Excel spreadsheet that was [2:04] already connected into those data [2:05] sources, you'd probably spend less time [2:07] going and gathering information and [2:09] analyzing it. [2:10] >> I can tell that you guys are excited [2:11] about how things are going right now. [2:13] Would you put some words on how how [2:14] things are going? [2:15] >> Vertical. [2:17] >> Uh look, I mean, in the last seven days, [2:20] we've signed the same number of [2:22] contracts as we signed in the whole of [2:23] Q4. [2:24] >> Wow. And [2:25] >> congratulations. Thanks very much. And I [2:26] think it's it's really just the turning [2:29] point I think in the sector whereas as [2:31] we keep saying it's like there is clear [2:33] um tangible value being driven by these [2:36] products and it's only going to get [2:37] better and quickly. [2:39] >> What were people using MML using before? [2:41] What were their tools that they were [2:43] using in the daily daily work? [2:45] >> So they would have their um data sets uh [2:48] and then they would they would spend a [2:50] lot of time in the office suite or in [2:51] Outlook, right? Um, and you know what [2:54] that meant was, you know, a lot of that [2:56] process is super manual and super [2:58] repetitive. You know, I think, you know, [3:00] the key here is we're definitely not [3:02] saying that humans should like never do [3:03] these tasks, but you know, if you're [3:05] organizing logos in a PowerPoint [3:07] presentation and you've done that [3:09] hundreds of times before, [3:11] >> uh, and you're a very well educated, you [3:13] know, analyst or associate, it's [3:15] probably not a great use of your time, [3:17] >> right? And I remember from I think maybe [3:18] from our interview or sometime on the [3:20] first office hour you had a unique story [3:21] on why you wanted to solve this problem. [3:23] Remember what that was? [3:24] >> Yeah. So we um so we sold our first two [3:28] companies and um after we sold our [3:31] second company we did a bunch of kind of [3:33] investing ourselves which I think on the [3:35] whole we were pretty bad at. But we [3:37] became very interested in automating as [3:40] much of those processes as you possibly [3:42] could, right? And so it it started [3:45] whereby we we literally were um when we [3:48] would receive an opportunity via email [3:50] and having sold a company you kind of [3:51] tend to receive a bunch of opportunities [3:53] every day. It would then enter into this [3:55] like a genensic system that to be honest [3:57] we were just building for fun um and [4:00] kind of produce like a one pager for us. [4:02] But the interesting thing about the one [4:03] pager is whilst you know we received [4:06] some information via email what was in [4:07] the one page probably 90% of that [4:09] information was unrelated to what we [4:11] would have received via email. In other [4:12] words, you know, let's say we got an [4:14] opportunity to invest in a startup, [4:15] right? This thing would go off and look [4:17] at their LinkedIn and their background [4:19] and then look at like comparable [4:20] companies on Crunchb or S&P and so on [4:23] and kind of produce this one pager that [4:25] you would probably go and do as a human [4:27] as your first port of call, you know, [4:28] little things like if it was a consumer [4:30] company, it would go and look at review [4:32] websites just as an example. Um and [4:35] yeah, we thought it was pretty cool and [4:36] then people became interested in it and [4:38] then we kind of agreed that we were [4:40] quite bad investors and uh okay at [4:43] building stuff. Um and that's how it [4:45] started [4:46] >> and that then turned into you selling [4:48] this product or new version of that [4:50] product to the top financial firms in [4:52] the world basically. [4:53] >> Yeah. So um you know we can say now that [4:56] we're about 10% of the largest private [4:59] equity firms and investment banks in the [5:01] world use our product. [5:02] um you know we have other customers [5:04] asset managers sovereign wealth funds um [5:07] you know a bunch of venture firms but [5:09] yeah it's pretty it's been pretty [5:10] exciting [5:11] >> and how would they do how would they [5:13] doing their work before before you start [5:14] using model ML [5:15] >> so as crazy as it sounds now right so [5:18] let me use an example whereby [5:20] um let's say you're you're tracking a [5:23] public company right and every quarter [5:26] or every time there's a release um maybe [5:29] as part of your job as an analyst or [5:30] associate you'll go to uh the release [5:33] the filings and you'll put together [5:35] what's called an earning summary. So [5:36] think about it like a beautifully [5:38] present like a slide effectively a [5:39] single slide with a bunch of information [5:41] but all that information is coming from [5:43] the same piece of information. So for [5:44] example you might always get certain [5:47] numbers from the filing itself. You [5:49] might get other numbers like consensus [5:50] or similar from a fact set and so on and [5:52] they're put into these slides [5:54] >> and these things take ages. We're [5:56] talking probably like days you company [5:58] just to pull together these slides, [6:00] right? because everything has to be [6:01] linked back to source and checked and so [6:03] on. And what you can do in model is so [6:04] think of model as you know being used as [6:07] the building blocks for this. So you [6:08] have an Excel spreadsheet that's already [6:10] connected into these data sources and [6:12] you might want to export that into your [6:14] designs, right? And so rather than [6:17] having to go and gather that information [6:19] once that release happens, this one [6:22] pager effectively was actually three [6:23] pages, a cover page, your main page, and [6:24] a legal page just appears in your [6:27] SharePoint or Google Drive, right? And [6:30] it's kind of 90 95% of the way there. [6:32] >> Wow. Um and we think in some cases more [6:34] accurate because actually trolling [6:36] through these filings as a human uh and [6:39] as because you can also pull from [6:40] multiple data sets for the same figure. [6:42] Y [6:42] >> yeah can be more accurate. [6:44] >> I'm assuming that you were you built [6:46] something in the beginning and there was [6:47] a model that was could do some amount of [6:49] this work because you did YC how long [6:51] ago? [6:51] >> A year ago. [6:52] >> A year ago. So the models have probably [6:54] progressed a lot in that year. What were [6:55] things the models could do a year ago [6:57] that were impressive and what can they [6:59] do today and where are they going to be [7:00] able to do in the future? the whole [7:02] industry or the industry we're in right [7:03] now. [7:04] >> Yeah. [7:04] >> Um really from January [7:07] >> is going straight line up. And I think [7:09] the main difference that we've seen this [7:10] year versus last year is last year was a [7:13] year of testing, right? So everyone [7:15] wanted to make sure they had some sort [7:17] of exposure like as in they were they [7:19] were triing something or looking into [7:20] something but it was still testing, [7:22] right? And us having ran consumer [7:24] companies, we don't really care about [7:27] revenue figures. We've always been [7:28] obsessed with like usage and retention. [7:30] >> Right. Right. That's good. That's good. [7:32] That's that the most valuable thing you [7:34] want to bring into B2B. Like B2B people [7:35] don't always know the most important [7:37] stuff. [7:37] >> We which we think is is madness. But [7:40] this year everyone kind of the whole [7:42] world went from testing to using. So [7:44] there and there was a fundamental shift. [7:46] I think these agentic systems there was [7:48] small things like the improvement in [7:49] things like function calling uh with [7:51] some of the newer models but everything [7:52] has become considerably better by [7:55] itself. And that's the interesting [7:56] thing, right? It's like I think [7:58] >> even if we do nothing, [8:00] >> theoretically our product will improve, [8:02] which is a pretty interesting world to [8:03] be in, right? Um and so when you're [8:05] constantly working on something and [8:07] you're you're pushing the boundaries, [8:08] you know, month to month, what what's [8:10] possible is like just wasn't possible [8:11] the previous month. [8:12] >> I also think the uh the vision models. [8:15] >> Yeah. I mean, when they first started [8:17] coming out and and improving and we were [8:19] sitting next to each other, we were [8:20] going crazy, weren't we? It was like the [8:22] stuff that we could do with the vision [8:24] models. If you think analyzing files as [8:26] an example, you know, OCR, okay, was [8:29] amazing. You combine that with vision [8:31] and its ability to uh read information [8:34] from tables and charts, it really just [8:36] changed the game. [8:37] >> Yeah. And also I think it's a [8:39] misconception, right? So like I think [8:40] the industry still thinks that AI um or [8:44] a large proportion of it is is kind of [8:45] where it was even 6 months or so ago, [8:47] right? You know what we're seeing today [8:50] is, you know, if you look at some of [8:52] these data providers, they've got humans [8:54] reading information from say public [8:56] filings and putting that in structured [8:57] format. Um, you know, the bulk of the [9:00] work that we're doing, what we're seeing [9:01] is models are already more accurate than [9:03] humans in those sorts of tasks. And so [9:06] uh I think that will take a little bit [9:08] of time in terms of confidence but but I [9:10] but I think some of the the more lower [9:12] level more kind of data gathering and [9:14] presenting type tasks are fully already [9:16] being automated at the top firms. [9:17] >> This is interesting. This I'll tell you [9:19] like they at YC the evolution went [9:21] through at the same time. So two years [9:23] ago we would say oh you can't really [9:25] sell software to investment banks or [9:26] private equity funds. they don't really [9:27] buy software or if they do they buy one [9:30] piece of software every 10 years and the [9:31] same was true for lawyers but something [9:33] happened last year like they all got [9:34] very AI curious they're curious about AI [9:37] they were like let's [9:38] >> let's do pilots to play with the [9:39] software and try it out and you were [9:41] saying that like this is the year when [9:42] that turns into actual contracts [9:44] >> exactly [9:44] >> and you can't you can no longer say [9:46] these companies don't buy software [9:47] that's wrong they're all buying software [9:48] now and that's because this this thing [9:50] happened in the last 12 months [9:51] >> exactly and last year was the year of [9:52] kind of proof of concepts and now you [9:54] know our average contracts are in the [9:56] years first of [9:57] Um and I I just think in general um [10:01] people are seeing you know a lot more [10:03] kind of shorterterm or instant value you [10:06] know I also think just you know on the [10:07] note of uh these firms historically [10:10] inclusive of law firms and others not [10:12] buying software that is I think true um [10:15] I think the big difference here is um [10:18] this is like number one thing at from [10:20] the very top this is like the number one [10:21] thing on everyone's agenda so we are not [10:23] selling like the next CRM or like the [10:25] next like data vendor tool or anything [10:27] like that. We are selling what we're [10:29] describing as the most advanced sort of [10:31] AI solution for financial services in [10:32] the world, right? And I think if you are [10:34] a a CEO or or or an exec in general, [10:38] you've kind of got to take that call, [10:40] you know what I mean? And I think that's [10:41] really played to the advantage for all [10:42] startups for sure. [10:43] >> Can you describe some of these sales [10:44] meeting like who is the decider? Like [10:46] who decides to buy the software and and [10:48] what are they like? CEO level um or or [10:52] or in general just the most senior [10:53] people at the firm regardless of if [10:55] you're a top five or top 10 investment [10:57] bank priority firm as I said I'll sub [10:58] them up and whatever it might be um [11:01] which I think at the start we found like [11:03] really strange because we we sort of [11:05] thought that this would be looked at at [11:07] a team or or group level it's really [11:10] not. I think it's so important um [11:14] firmwide that everyone has to be [11:16] involved from from the very top. And [11:18] actually what we found is you've really [11:21] got to get that buy in um you know from [11:23] the right person at the top but then [11:24] you've also got to get buyin from the [11:26] people that are ultimately get going to [11:28] be implementing the tool. [11:30] >> So are you guys are flying to meet all [11:31] these firms where they wherever they are [11:33] >> wherever they are. Wherever they are. Um [11:35] so we have a bunch of people working out [11:36] of Hong Kong and Singapore now. Um we've [11:39] just opened a small office in India. Um [11:42] uh about half of our overall team are [11:44] based in London and we have an office in [11:45] New York. Um again I think like it's [11:48] clear the product is impactful, right? [11:50] Because when we're demoing we're like [11:51] laptop out we're showing real use cases [11:53] with real data [11:55] >> uh you know high use case frequency, [11:57] right? [11:58] >> So then if you think about like why they [12:01] wouldn't sign with you, right? A big [12:04] part of that I think is trust and and [12:05] and building that relationship, you [12:08] know, because it's also a different [12:09] dynamic. You know, it's like a lot of [12:11] time you're speaking to folks that, you [12:13] know, if they make a wrong call here, [12:14] they could get fired, right? And so you [12:16] spend a lot of time building that trust. [12:18] I think a big part of that trust is is [12:19] FaceTime um and getting in front of [12:22] people obviously, but then spending the [12:24] time on on the demo and how how that's [12:27] justified and and really investing in [12:29] things that are specific to the [12:31] customer. I want to go back a little bit [12:32] to the two previous companies that you [12:34] guys ran, Pat Llama and um Fancy. Before [12:37] we get into the details of what they [12:39] were doing, like what are some of the [12:40] learnings you took from those companies [12:42] uh into Mal ML? [12:44] >> I would say you've definitely got to [12:45] enjoy it. You've got to enjoy building [12:47] companies. Uh is my overarching uh [12:50] thought. What about you? [12:52] >> I think you've just got to be prepared [12:54] for [12:56] the worst. Yeah, you got to be prepared [12:57] for the worst and the most ridiculous [13:00] roller coaster experience. Uh I think [13:02] all startup founders will say the same. [13:04] I think it's going to be a lot of ups [13:05] and downs, a lot of the time downs and [13:08] you've just got to be prepared for that [13:10] and know it's coming. Um and as Chess [13:12] says, just enjoy it because it's fun. [13:13] >> And is there a sense that when you start [13:15] the third company that you you've gone [13:17] through that before, so you've gotten [13:19] used to the ups and downs and you can [13:22] just kind of like be calmer about it? [13:24] >> You can definitely be calmer. definitely [13:26] karma but I but but there is there is [13:28] certain things that you just cannot [13:29] prepare yourself for you know I think [13:31] like you know like it just in general no [13:34] matter how much you've kind of said to [13:36] yourself this is normal you're going up [13:37] and down there'll be the odd thing it [13:39] will be employee related or product [13:41] related whatever that will still [13:43] surprise you the fundamental thing that [13:45] I think we realize that we've at least [13:47] brought into this company is this [13:48] concept of of perseverance not blind [13:51] perseverance but perseverance and I [13:53] think there's a clear there's a clear [13:55] difference where if something logically [13:58] makes sense, right? Like if you just if [13:59] in an unemotional way you're approaching [14:01] a problem and it logically makes sense, [14:03] right? Then you should probably continue [14:07] doing that thing, right? And um not let [14:10] anything stop you. And I think like the [14:12] consistency that we've noticed that of [14:14] of of founders that we've some of that [14:16] we've invested in or worked with is like [14:19] the ones that kind of do that and really [14:21] persevere tend to win. So I think that's [14:23] probably the biggest learning for us. [14:24] Just persevere at all cost. [14:26] >> Yeah, for sure. Uh I mean one of the big [14:28] ones for me in particular is definitely [14:30] hiring. I think when you know I was CEO [14:33] of Fancy I was 22 23. Um so hiring was [14:38] new to me and I think [14:40] >> to be honest I didn't really have a clue [14:41] what I was doing [14:42] >> and I think we still [14:43] >> Yeah. Okay. And hiring is always tricky. [14:46] Um, I think one of the biggest things [14:48] right now and we say it a lot and it's [14:50] probably the biggest thing that comes up [14:52] in an interview is is how much do you [14:54] think you enjoy working with that [14:56] person? I think you know at Fancy again [14:58] being sort of inexperienced and maybe [15:00] naive it was all about where have they [15:02] worked before you know what's on their [15:03] CV and you kind of ignored the the [15:06] little things that you know that that [15:08] maybe you shouldn't. And now it's kind [15:10] of the the main thing that we look for. [15:11] You know we're going to be spending a [15:13] lot of time with these people. you know, [15:14] we work a lot and everyone on the team [15:16] does and you've got a lot of work with [15:18] this person. Um, so we often meet in [15:21] person multiple times and make sure [15:22] they're the right cultural fit and it [15:24] really makes a difference. I think right [15:25] now, you know, we hire very slowly. Um, [15:28] and we try and make sure we hire the [15:29] right people. Um, and it feels like we [15:32] we've got better at that as time's gone [15:34] on. And humans are just so much more [15:36] impactful now as well. [15:38] >> So, I think it's it's just even more [15:41] important than it has been before. Um, [15:44] and yeah, the other aspect I said is [15:47] work ethic. I think we, you know, I [15:49] think we're known for it. I mean, right [15:52] now we're still working seven days a [15:53] week as we said and we have done for [15:55] about 18 months. Um, and I think [15:58] certainly in that initial period that [16:00] you've got to do that and I think it, [16:01] you know, our team work 6 days a week at [16:03] the moment. Um, look, that may change [16:05] I'm sure as time goes on, but I think [16:08] part of that point about enjoying [16:10] working with the person is they've also [16:12] got to really enjoy what they do, right? [16:14] So, we we've started to ask more [16:15] questions around look all the other sort [16:17] of standard stuff that has to be done. I [16:19] think it's important to have the kind of [16:20] standard interview type structure, but [16:22] are they going to enjoy their dayto-day? [16:24] You know, the question that we ask [16:25] people is like, you know, what have they [16:27] been building? You know, regardless of [16:29] where they're in engineering, right? And [16:30] soon as they start to me me me me me me [16:32] me me me me me me me me me me me me me [16:32] me me me me me me me me me me me me me [16:32] me me me me me me me me mention things [16:32] like oh they've been doing a bit of vibe [16:34] coding or this that and the other we're [16:35] like okay they're going to enjoy what [16:36] they're doing right and I think that's [16:37] really important [16:38] >> it feels like there's a moment in a [16:39] company's lifetime where [16:41] >> it's suddenly it's working [16:43] >> but you haven't won the market yet it [16:45] sounds like [16:47] >> that at that time it really matters to [16:48] work hard [16:49] >> because like they're you're not the only [16:51] one trying to go after this market [16:52] >> yeah we just don't like losing [16:56] >> um maybe go back to like uh winter 17 [16:59] >> fat llama Um, tell us about Fatal Lama. [17:01] >> Um, so Fatal Lama was a marketplace that [17:04] allowed people to rent items from people [17:05] nearby with the main differences, you [17:08] know, uh, you were insured. So if I let [17:09] if you're my neighbor and I lent you a [17:11] camera, a $10,000 camera, I'm insured. [17:13] If I lent you a drill, the same thing. [17:15] Um, that was what we did differently [17:16] with the model. That model had been [17:18] tried a lot of times before, but it [17:20] still took us 3 years to find product [17:21] market fit. I sort of define product [17:23] market fit loosely as um, you know, the [17:26] unit economics add up. people want, you [17:28] know, uh uh what you have in terms of [17:30] product um in a sort of relatively large [17:32] addressable market, right? But it took [17:34] us 3 years to to get there and I think [17:37] that was, you know, we learned so much [17:39] during that period. Um and so yeah, that [17:42] that was Fat Llama and I can talk more [17:44] about that in a second, but what about [17:45] Fancy? [17:46] >> Yeah, so Fancy again consumer business. [17:48] So we were a a last mile grocery [17:50] delivery business uh which now [17:52] everyone's probably bored of speaking [17:54] about them but back then so this was end [17:56] of 2019 uh start of 2020 uh so our model [18:00] was a little bit different to a a Door [18:02] Dash or an Instacart. So we were what's [18:05] called a vertically integrated model [18:06] which means uh we had our own [18:08] warehouses. We held the stock ourselves [18:10] and this was relatively new in Europe. I [18:13] think there was one other player doing [18:14] it um out of Turkey, but in the UK in [18:17] particular, we were the first ones. And [18:19] really, it started it, you know, as a as [18:21] a delivery app for students. You know, I [18:23] was a student at the time. And it really [18:25] came about the need, you know, I was [18:26] really sitting there one day thinking, [18:28] God, I could do some, you know, Pringles [18:30] or some a beer to be honest and I didn't [18:32] want to walk to the corner shop that was [18:33] like 5 minutes away. [18:34] >> And this was during CO. [18:35] >> So, this was just before CO. So, this [18:37] was like, you know, 3 months before CO. [18:39] We kind of built the app MVP in sort of [18:41] four to 6 weeks. Um I studied computer [18:44] science at uni and so it was just before [18:46] co and really almost instantly as [18:49] opposed to fat lama we kind of found [18:51] product market fit almost overnight [18:52] which is that sounds obvious right it's [18:54] like you're delivering sort of beer ice [18:56] cream to students in [18:57] >> at the same cost as they would get it [18:59] from the corner shop. [19:01] >> Uh so so we took off and that was [19:03] obviously really exciting. Um and then [19:05] we got on to YC and then CO happened and [19:07] co you know for our business was really [19:10] that shot of adrenaline. I mean co [19:11] happened everyone was staying at home [19:13] you know the business really then [19:14] started to take off. Um it was an [19:17] amazing business. It's an extremely [19:18] difficult business. Um I'm sure we'll [19:20] come on to it in a bit but we got hit [19:22] with we always like to say a lot of [19:24] cricket bats to the face. You know a lot [19:25] of stuff can go wrong. Um but it was [19:27] awesome. So we continued to grow in the [19:29] UK. Um we raised money out of YC. Uh and [19:33] then we got an acquisition offer from [19:35] Gouff I think it was about 18 months [19:36] after we started. Um and Gopuff at the [19:39] time a market leader um they were very [19:42] very big in the US and they wanted a um [19:45] an opportunity to come into Europe and [19:47] we were best place for that. Um so it [19:49] really was an incredible ride. [19:51] >> Yeah. Yeah. It was like the at least for [19:54] that amount of time felt like the [19:55] perfect kind of startup ride I'd say. Uh [19:58] but again I think Fat Lama was the polar [20:00] opposite. The story I always talk about [20:01] with that llama in terms of perseverance [20:03] was you know I was you know maybe very [20:07] early 20s obviously and [20:10] you know at that time when you believe [20:12] in something and you're raising a small [20:13] angel around so this is preyc right is [20:16] you know we were raising I think maybe [20:18] $50,000 maybe $100,000 which um you know [20:22] nowadays is like I suppose back then at [20:23] least then it meant everything to us you [20:25] know we were scraping around we we were [20:26] doing half an hour pitches with an angel [20:28] that might put in $2,000 [20:30] you know, we were just doing everything [20:32] we could. We fundamentally believed in [20:33] this idea and um anyway, so we built the [20:36] MVP, we launched and I started [20:39] accounting. So, so I was big on like you [20:42] are the numbers going to make sense, [20:43] right? So, I had my like forecasted [20:45] average transaction value, my retention, [20:46] everything else. So, we launched and [20:49] pretty much straight away, like within [20:50] the first day, we got a rental and um [20:54] the rental was for $600. So, I'm like [20:56] over the moon. I'm update. The one of [20:58] the first things I did is I'm straight [21:00] into the financial model and I'm [21:01] updating the average order value and [21:03] this thing is looking crazy. I'm like [21:05] we're going [21:06] >> we're going the moon, [21:07] >> right? And um so that was on Friday. It [21:09] was due to be returned on Sunday [21:10] lunchtime. All right. And remember the [21:12] criticism I got, you know, I probably [21:13] did 100 pitches to raise money and maybe [21:15] one out of 100, you know, I must have [21:17] done close to a thousand the end would [21:19] say yes. But all the other 99 was like [21:21] no one's going to lend out an item [21:22] because everyone's going to steal them [21:23] and even if you have insurance, the [21:24] insurance isn't going to work, right? [21:25] Anyway, so we had and that's all I'm [21:27] thinking, you know, it's all we're [21:28] thinking about at the time. So, we had [21:29] this first rental. Anyway, got to Sunday [21:32] and the the lender of the item called us [21:34] and said like, I can't get hold of of [21:36] the borrower of this item. So, okay, [21:38] it's fine. And then we realized after a [21:40] couple of hours, like he wasn't [21:41] responding and we started to get a bit [21:42] nervous. One of the things we did, it [21:44] was, you know, it was a native app and [21:45] when people search because it helps with [21:47] what you show in the search results is [21:48] we save or we look at the the latitude [21:51] and longitude. So, quite an accurate [21:52] geoloc. So, we had that. So, I was like, [21:56] "Okay, I'm I'm going straight out there. [21:57] We got to go and find this item." And [21:58] this was a this was like an £1,800 [22:00] drone, right? So, for us, this was like [22:03] everything, right? So, we went up. [22:05] Anyway, so I arrived on this random [22:06] street in North London, and there was [22:08] just a door open to this house. I don't [22:09] think we've ever told this story cuz [22:11] it's just like it's that awful. And and [22:13] the door was there. Door was open. I [22:15] couldn't believe it. As I opened the [22:16] door, there it was. The drone was just [22:19] there on the side. So, I picked up this [22:20] drone. I got back in the car and I just [22:22] went straight down to the to the [22:23] lender's house. I arrived at the [22:24] lender's house. But you got to picture [22:26] this, right? Is what's going through my [22:28] head is everything that we've pitched, [22:31] every person that I pitched, family and [22:32] friends, we had spent months, probably [22:35] close to a year at this point, [22:36] convincing people that this would work [22:37] and they believed us and they believed [22:39] and I and they bought into this. And so [22:41] all I'm thinking is like I've lied to [22:42] them. You know, this is like my whole [22:45] world was falling apart at this point. [22:46] And so anyway, I arrived at the [22:48] borrower's house uh or the lender's [22:49] house rather, opened the door and he was [22:51] like, "Hey, you're not the lender." And [22:53] I was like, "Oh, I know." I gave him the [22:55] drone. He was like, "And you're wearing [22:56] a fat alarm t-shirt." I was like, "Yeah, [22:58] we we deliver the item back [22:59] automatically." He was like, "This is [23:00] awesome, man." Like, closed the door. [23:03] And so we it did add up you know in [23:06] terms of the insurance and we we focused [23:07] on verification but it fundamentally [23:08] came to that that perseverance piece is [23:11] like we believe this is something that [23:12] should exist and we believe from a tech [23:14] from a tech perspective we could make it [23:16] work. Uh and it ended up doing so. Um [23:19] but yeah it was tough. [23:20] >> Both outcomes seems like really really [23:22] great though like it's like set you up [23:23] well for this company. [23:24] >> Absolutely. I mean an did what the first [23:26] 1,200 deliveries I think. Yeah. I think [23:30] I think with Fancier from the outside it [23:32] looks like a very fairy tale story. [23:35] >> Everything like after the fact. [23:37] >> Exact. Exactly. Exactly. And they're [23:39] like, "Wow, it's like I should start a [23:40] startup." And almost like you should, [23:43] but really it doesn't really it's not [23:45] all what it seems. You know, with Fancy, [23:47] we had so many disasters. Uh, you know, [23:49] and and so many times where we were [23:51] like, "God, is this really going to [23:53] work?" Um you know even when COVID [23:56] happened uh you know we had to stop [23:58] delivering for for a couple of weeks you [24:00] know there was a point I remember you [24:02] remember this sort of maybe two three [24:04] months into the company uh so we were [24:06] working with Stripe as our payment [24:07] provider and one day they just shut us [24:09] off uh cuz apparently there was a breach [24:12] in in their terms and service which [24:14] later was a misunderstanding but anyway [24:15] they shut us off for probably you know [24:17] three or four days uh bear in mind we [24:19] couldn't take any payments right and at [24:21] this point I can't remember how many [24:22] orders we were doing maybe not massive [24:23] you know, maybe 500 orders a [24:25] >> week in inventory. So, there wasn't a [24:27] problem with that at least. [24:28] >> Yeah. Exactly. So, at this point, what [24:30] do we do? We've got orders coming in. We [24:31] can't take payments. Um, so we ended up [24:33] just taking payments over the phone, [24:35] taking payments with cash, taking [24:36] payments with PayPal. Uh, and there were [24:38] so many of these stories. And if you [24:39] think of our business, you know, the [24:41] amount of issues we had with delivery [24:43] drivers, with warehouses. Um, it really [24:47] was, uh, an intense situation, but [24:50] that's really what made it fun, right? I [24:51] think um you know building a startup you [24:53] just become so thick skinned and you [24:56] know problems come about you know on a [24:58] daily hourly basis and you know kind of [25:00] your mood is like this all the time but [25:03] you know it's epic right on the payments [25:05] one we emailed Patrick directly and he [25:06] responded like within he sorted us out [25:09] >> yeah it was awesome [25:09] >> so if Patrick listen says thank you [25:10] Patrick he sort us out um but I think on [25:13] both of those the the general story is [25:16] and an you know the the team delivered [25:18] the first 1500 orders right and So u you [25:22] know ourselves so it was very much a [25:24] case of you know you hear but we rather [25:26] than coding at night we were coding [25:27] during the day because really most of [25:28] our orders were kind of in the evening [25:29] right [25:30] >> and but what that meant was we were [25:33] speaking to a customer for every single [25:35] order [25:36] >> right and so that the customer feedback [25:38] wasn't just like [25:39] >> real time that you hear about it was [25:41] like every order we knew what was [25:42] working what wasn't working right which [25:44] meant that you know when we started to [25:47] implement drivers [25:48] >> you know we always say that you you knew [25:51] exactly how long it was going to take [25:52] from the warehouse to that house because [25:54] you've done the route about 50 times [25:55] before and I think across all three [25:58] businesses if you sort of set up and [26:01] maintain the foundations of being [26:03] incredibly customercentric things tend [26:05] to go okay [26:06] >> yeah there is a pattern though with [26:08] startups like as they grow they move the [26:10] builders further and further away from [26:11] customers and there's a bunch of other [26:13] roles in between and eventually you kind [26:15] of have to interpret the information you [26:16] learn from customers like how are you [26:17] going to avoid that Well, it's on I mean [26:22] um I mean you speak to everyone you can [26:24] talk about on the product side I think [26:26] on on the sales side particularly [26:28] because also we have a proof of concept [26:30] phase still right so we still have a [26:32] trial phase [26:33] >> which is this nice blend between kind of [26:35] uh a sales stage but also a pure [26:38] customer feedback stage [26:40] >> so naturally uh and I and I'm 100% [26:43] involved with that and and it's also [26:45] what I enjoy as well right you know [26:47] spending time you know I don't like [26:48] demoing on a screen. I like sitting with [26:50] a user with a laptop how they're going [26:52] to work and working on the product [26:53] together. And that also is the best way [26:55] we found to sell. Um, but you're also [26:58] doing customer feedback calls [26:59] constantly, right? [27:00] >> Yeah. I mean, you know, it's always a YC [27:03] mantra. Why I think, you know, you got [27:05] to speak and listen to customers and [27:07] that's really what we try and do, you [27:09] know, on a daily basis. You know, really [27:10] figure out what are their pain points [27:12] with using model ML, what does their day [27:14] look like? and then think internally, [27:16] you know, how can we then productize [27:17] that and get that into their hands for [27:18] them to try. Uh, and it's that constant [27:21] iteration, right? We always say, you [27:22] know, the quicker we can ship things, [27:23] the quicker we can learn. Um, and really [27:26] we want to try and stay as lean as [27:27] possible. You know, we always have this [27:29] theory. We want to be the maybe not the [27:31] first, but, you know, one of the first, [27:32] you know, 10 person billion dollar [27:33] company. And it's just about staying so [27:35] close to the customer uh, and in in the [27:37] details. [27:39] >> Y's next batch is now taking [27:41] applications. Got a startup in you? [27:43] Apply at y combinator.com/apply. [27:46] It's never too early and filling out the [27:48] app will level up your idea. Okay, back [27:51] to the video. So, during YC, we have [27:54] this um group office hour topic where we [27:56] talk about the what motivates you to [27:58] build a company. Um and it's mostly to [28:01] surface sort of the motivations for [28:03] founders and their co-founders. [28:05] >> So, when things are really tough, you [28:07] know why you're there. Do you guys [28:09] remember what was your initial [28:10] motivation when you started the the [28:11] first two companies? than maybe what it [28:13] is today. [28:14] >> You know, when we have our oneto ones, [28:15] we're not brainstorming how to raise [28:17] money or we're brainstorming actually [28:18] how we can do this for the rest of our [28:20] lives because it is so rewarding. You [28:22] know, building something for [28:24] individuals. Making money, I think, you [28:25] know, it just gets boring so quickly, [28:27] right? You know, it's not it's not [28:29] really that motivating, you know, [28:31] whereas building something that makes [28:32] people smile that's very impactful and a [28:35] lot of people are using it, that's just [28:37] so motivating to us. And I think like um [28:40] that really has been the story across [28:43] all the companies. Um you know and I [28:46] think what's been nice about this and [28:47] surprising about this is we were we [28:49] questioned ourselves going from a two [28:51] consumer brand heavy type businesses [28:53] into a B2B but actually particularly in [28:55] the world of AI the most important [28:57] element is like the B2C element. Uh and [29:00] and that is to us the most rewarding [29:02] element like when you do a demo to [29:04] someone or they click run and they do [29:05] something [29:06] >> and it is just there's no money that can [29:09] buy those moments [29:10] >> cuz they've never seen that before [29:11] >> and they've never seen before. They've [29:12] never seen it. [29:12] >> It was like with our previous companies [29:13] with Fancy when you would be at the door [29:15] with their delivery sort of 10 12 [29:17] minutes after they ordered it and then [29:19] you saw their face like what the hell's [29:21] going on? You're already here. Uh like [29:23] Chaz said you just you know it's [29:25] priceless. I've noticed this with AI [29:26] products that they often the customers [29:29] cannot even imagine what the solution [29:30] will look like [29:31] >> um because the models are so good and [29:33] they're so advanced right now [29:35] >> and very often [29:36] >> um you just have to build like you build [29:39] it for them first like you can't really [29:40] figure out try to figure out what the [29:41] problems are because like it's not a [29:43] specific problem you're solving it's [29:44] just like it's like a whole other like a [29:46] whole other league of solutions. [29:48] >> Yeah. And and you know the difficult [29:50] thing as well is you you do have to [29:52] rethink things slightly from the ground [29:54] up. [29:54] >> Mhm. So I I do think that you know at [29:57] least as of today if our product we made [30:00] it we made a call that we would kind of [30:02] rebuild our version of like PowerPoint, [30:04] Word and Excel because we we do believe [30:06] at least for the time being that the way [30:08] in which people interact with technology [30:10] regardless of of type of technology will [30:13] remain quite consistent. In other words, [30:14] like a long form document, a [30:15] storytelling presentation and tabular [30:18] across XL Word and PowerPoint. Um we [30:20] think that will stay the same. [30:22] >> So you're not trying to get people to [30:23] change the current We're not trying to [30:24] get because we just think that that's [30:26] what people are very used to. However, [30:29] we think what's if we look at what we're [30:31] seeing this year, a lot of what's [30:34] happened up until now is sort of humans [30:36] are still coming into these systems and [30:38] they are like clicking run on something [30:40] basically, right? We think the biggest [30:42] shift this year is going to be that even [30:45] that element won't happen and therefore [30:47] elements of the user interface we think [30:49] will be less important. In other words, [30:51] these these tasks will happen entirely [30:54] autonomously. You know, as you arrive in [30:56] the morning and the things that you [30:58] would normally have had to go and [30:59] trigger and click run, they will already [31:02] be there. They will already be done. [31:03] >> Right. Right. Right. You guys are [31:04] siblings. Um, historically in YC, we've [31:07] known that this is a pretty good recipe [31:08] for a good co-founder relationship. Tell [31:11] us, I mean, you've been co-founders of [31:12] multiple companies. You've had other [31:14] co-founders, too. like tell us what [31:16] you've learned about the most important [31:17] thing and and maybe like for the [31:19] audience of founders who are thinking [31:20] about finding a company like what should [31:22] I be looking for in my co-founder and [31:24] like what's important and what's not [31:25] important. I think we mentioned it, you [31:29] know, previously around hiring. I think [31:32] this is sort of tenfold when it comes to [31:34] your co-founder. You know, you're going [31:36] to spend so much time with this person. [31:39] You know, more time that you spend with [31:41] your family, with your partners, with [31:42] your friends. Um, so I think first and [31:45] foremost, is that person someone that [31:48] you want to spend a lot of time with? [31:49] Uh, you know, obviously with us, [31:50] >> we're still not sure. [31:51] >> Yeah, we're still not sure. [31:53] >> I think if you were not sure, you [31:54] wouldn't say that. [31:55] >> Yeah. Exactly. Exactly. Otherwise, it' [31:57] be really awkward. It would be really [31:58] awkward. Um, so look, you know, we've [32:00] always had a really good relationship. I [32:03] think not all brothers working together, [32:05] I think, is a good idea. I think, you [32:06] know, we've been fortunate enough we've [32:07] always had a good, you know, personal [32:09] and working relationship. One of the [32:11] things that make things really easy, I [32:14] think, is and again, not not all [32:16] siblings are like this, but there's no [32:17] filter between us. You know, we're very [32:19] transparent. We're very honest. And I [32:21] think you know with your founder you [32:24] need to be exactly that you know there [32:26] can't be any miscommunication lack of [32:28] communication like we all know one of [32:30] the biggest reasons if not the biggest [32:31] reason why you know startups fail is is [32:34] founder fallout and and I think with us [32:36] we're always very good at you know [32:38] communicating trusting each other. [32:40] Another thing on the trust piece then is [32:42] the way that we describe our ven [32:44] diagram. You know I I think I'm an [32:47] engineer but I'm not actually an [32:49] engineer. I think I am. Anie studied [32:50] computer science, I studied uh [32:52] accounting. So our ven diagram and I [32:55] think if you're thinking about another [32:56] co-founder I'd really consider this is [32:59] um Anie obviously handles sort of [33:01] engineering product and I handle kind of [33:03] finance commercials and products as well [33:06] kind of in the middle right so the [33:07] overlap in our vend diagram is basically [33:10] um customers and product right which we [33:13] think is just like the most important [33:15] piece anyway right and that bit we both [33:18] really enjoy and we and we love but we [33:20] also love all the other stuff and I [33:22] think like that clear uh segregation of [33:25] duties and interests from day one I [33:29] think is really important. So I think [33:31] one of the things that we think that we [33:33] look at is I really emphasize that point [33:35] of interest because actually you may [33:37] like for example Anie may have studied [33:38] CS and I may say but if I if I really am [33:41] interested in wanting to basically do [33:42] Anie's job and write the bulk of the [33:44] production code then we've probably got [33:46] a problem. You know what I mean? And so [33:47] I think it's like really think about [33:49] that. Uh and and and that tends to last [33:52] for a long time as well. [33:53] >> I got this email uh I think it was [33:55] yesterday from a person who was in their [33:57] early 20s. They were basically asking me [33:59] for not startup advice but life advice. [34:02] So what do I do? I'm like 21 and I think [34:05] he was writing something along the lines [34:06] of like I know I can go and build [34:09] something B2B that's quite narrow in AI [34:11] but it doesn't seem to motivate me [34:12] enough. I want to be something much [34:14] bigger. Like I want to have a bigger [34:15] impact on the world. like if you got [34:16] that email, if you would give him [34:17] advice, like what would you advise [34:19] someone who's in their really early 20s [34:20] or in school and they're thinking about [34:23] their career right now or maybe in the [34:24] world of AI um or thinking about [34:27] starting a company but want to do [34:28] something big like what would you tell [34:30] them? [34:30] >> I think we're probably biased because [34:32] you know we'd probably favor starting a [34:34] company um before going to work at a you [34:37] know a big corp. Um, I think first and [34:40] foremost we'd be very honest with them. [34:41] You know, as we mentioned before with [34:43] our fancy story, you know, you look from [34:45] the outside and it might seem like uh, [34:48] you know, sunshine and rainbows, but [34:50] it's really hard. You know, building a [34:52] business is is really, really, really [34:54] hard. It's really fun if you enjoy that [34:56] stuff, but it's really hard. Uh, so [34:58] really, you've got to say to yourself, [35:00] look, you might be building a business [35:01] for the next 5, 10, 15 years, [35:04] >> right? [35:05] >> It might not going any go anywhere. And [35:07] you've got to be okay with, you know, [35:09] that reality. Uh, and ultimately, you've [35:11] got to be very passionate about what [35:14] you're building. You've got to have that [35:15] perseverance that Chad spoke about [35:17] earlier. Uh, and if that sounds good to [35:20] you, then then build a startup. And if [35:21] that doesn't sound good to you, then, [35:23] you know, I probably recommend probably [35:25] getting a, you know, a job somewhere [35:26] else and then develop over time. Uh, and [35:29] then if you feel like you're more ready [35:31] to jump into the startup world, uh, then [35:33] do that. So maybe if you're if you're a [35:35] builder and you enjoy seeing your stuff [35:36] in people's hands, the smaller the [35:38] company you work for, maybe you're even [35:40] your own company, the faster it gets in [35:42] the customer's hands. 100% and and [35:44] enjoying that. I then think another like [35:45] different way to look at this is is um [35:48] you know, think about I know it's maybe [35:50] slightly more, but is that if that's the [35:51] correct word, but you know, if you're [35:52] lying on your deathbed, you know, and [35:54] you look back at your life, like what do [35:57] you what do you want from your life? And [35:58] I think we we do that a lot. And [36:00] basically, you know, what we conclude is [36:02] we just we just want to have been [36:04] impactful with our time. You know, we [36:06] want to have, you know, you know, left [36:08] the world a better place, but but but in [36:10] a way that we've just built things that [36:12] people enjoy using and and I think [36:14] that's really what motivates us. But I [36:16] think really that one point of um you [36:19] know, you've really got to enjoy the [36:20] journey and we just we just love [36:22] building. [36:23] >> Just sounds like if you were in his [36:24] shoes, you would give yourself basically [36:25] the advice to the path that we ended up [36:27] on. [36:28] >> Yeah. Go all in. It wouldn't change [36:29] anything. [36:30] >> Hell no. Goal in. Goal in like all in. [36:33] And I think if anything uh more is more [36:35] in general. [36:36] >> I remember when I met with Paul Graham [36:38] sometime in around the beginning of the [36:40] batch and I was doubting whether if this [36:42] doesn't work out like maybe my cur is [36:44] And he was some he said [36:45] something like well if you're 26 and [36:47] you're poor uh that will be the worst [36:49] outcome. And probably most 26 year olds [36:51] don't have any money anyway. So doesn't [36:52] make a big difference. [36:53] >> Yeah. Yeah. It's also just like the way [36:54] I describe it is like well the worst [36:57] case like the very worst case is you're [37:00] you're going to learn so much over such [37:03] a short period of time and that is the [37:05] worst case cuz often when you compare [37:07] this to like you know we often now when [37:09] we're we're hiring people we talk about [37:11] their opportunity costs right [37:13] >> I think a lot of the time if you take a [37:14] step back those jobs or those roles [37:17] they're still going to be there in a [37:18] year's time right so like the actual [37:20] worst case is you've just learned a lot [37:22] over a here. And I think if you frame it [37:24] like that, then a lot more people would [37:27] just get out and build stuff. [37:28] >> And a lot of people that end up on these [37:30] tracks, um, they sort of like 5 years [37:32] into investment banking. It's hard to [37:35] step out of that. [37:35] >> Super hard. [37:36] >> But like it's just the risk that you [37:37] want to take, you want to take early on [37:39] because it's just like much harder to [37:41] change when you're 30 and you have [37:42] commitments and [37:43] >> definitely way harder. [37:45] >> Yeah. When you're hiring people, people [37:47] that come from finance or quant [37:48] backgrounds, are there things that they [37:50] have to unlearn about their job? Either [37:52] about the company they're working for, [37:54] about the industry in order to work for [37:55] a tech startup serving the same [37:57] industry. [37:58] >> Yeah. First principle thinking, eh, [38:00] >> they just got to move faster. Okay. Um, [38:03] and again, this is probably more your [38:05] realm than mine, but what we notice is, [38:07] you know, we'll ask someone to do [38:08] something and well, they won't, but they [38:11] will try and, you know, spend a day, two [38:14] days putting together some sort of [38:16] presentation, slide deck, [38:18] >> and we're like, what the hell's going [38:19] on? You know, uh, you know, we need this [38:22] now. Um, so it's things like that where [38:25] you get, you know, caught up in the, I [38:28] guess, the big company thinking, which [38:29] look makes sense in certain [38:30] environments. Um but when you're working [38:32] in a startup and you're wearing so many [38:33] different hats, frankly there's no time [38:35] for that. So we really try and from day [38:37] one uh I think you in particular really [38:40] try and just say look we need this now. [38:42] Um instead of you know all these [38:44] spreadsheets and and slides I think that [38:45] is the first lesson. It's like [38:47] everything that you have learned for 5 [38:48] years try and unlearn immediately. Um I [38:51] think just in the sense it's different [38:53] you know you just got to yeah you just [38:55] got to move so much faster. I think [38:57] building the plane going down the runway [38:59] is the best way to describe things. You [39:01] know, I think you know you know that is [39:03] the the quickest way to learn. It's [39:06] different when you're you know where are [39:08] now it's different. You know the [39:10] business you know we have large [39:12] customers in production that that can't [39:14] be the case but in terms of like the way [39:16] that you think or the way that you go [39:17] about your work the fundamentals you [39:19] know still need to remain the same. Um [39:22] you know we are big on this first [39:23] principles thinking which I know is very [39:25] much the terminology that is overused [39:27] but I think in general that way of [39:30] looking at the world is a is a better [39:31] way to build a company [39:33] >> and you guys chose to do YC three times. [39:35] Um can you describe sort of like how [39:37] you're thinking here [39:38] >> the second and third time it wasn't a [39:40] financial thing. We still get asked like [39:43] still we've we've had loads of teams [39:45] that you know people that have worked [39:46] for us that have gone on to even start [39:47] YC companies or gone on to build some [39:49] great companies and we still get asked [39:52] how we managed to maintain that and I [39:54] quote YC culture throughout the life [39:57] cycle of the company right and so the [40:00] the the way I sort of answer this is you [40:02] cannot replicate that you know I I [40:05] remember very clearly Michael Cyborg on [40:09] uh in 2017, you know, the first office [40:12] hours we had, you know, a young English [40:14] lad, you know, we report on our numbers [40:17] and I reported monthly and I remember [40:20] Michael being like, "What the hell are [40:21] you doing?" Like, let's talk weekly if [40:23] not daily, if not hourly. And I think [40:25] that culture never leaves a company, [40:29] right? That's the second aspect which [40:31] definitely cannot be overlooked is the [40:33] instant support network that you get. [40:36] You know, I think us being from Europe, [40:38] working seven days a week, it's very [40:41] unusual to say the least. And there's [40:43] not many people that we can call on. And [40:45] in fact, a lot of the people that we [40:46] surround ourselves with outside of work, [40:48] they fundamentally disagree with the way [40:50] that we work and how we're building and [40:52] why we're building, particularly when we [40:53] don't need to from a financial sense, [40:55] right? I think that's difficult. Whereas [40:56] in in the Bay Area, you know, [40:58] particularly uh and and driven by YC and [41:00] through YC, you get this like instant [41:03] network that you just you you cannot get [41:06] anywhere else. [41:07] >> And you ended up being in in the batch [41:09] winter 24, which was sort of like one of [41:12] the first really big AI batches where [41:14] most companies were building companies [41:15] similar to you. [41:16] >> Um how did that feel? Like was there [41:18] something [41:19] >> like just being in the Bay Area like 18 [41:21] months ago? [41:22] >> I mean it was just crazy. I think I [41:24] think again working side by side with [41:26] people building really cool products [41:29] working seven days a week and that buzz [41:32] you know around the office around San [41:34] Francisco it's really really hard to [41:37] replicate um and we absolutely loved it [41:39] didn't we [41:40] >> yeah and it's like it felt it's [41:42] interesting you mentioned that cuz that [41:43] really was [41:45] >> that was like the first batch where it [41:46] was like through and through is just you [41:48] know AI companies pretty much and like [41:50] >> it felt like in that Slack group didn't [41:52] it felt almost daily or hourly there was [41:54] like a scientific breakthrough. It was [41:56] like phenomenal to be a part of. [41:58] >> One question I get I was in Munich and [42:00] then I was in Zurich on a trip recently [42:01] and I was in London. I saw you guys uh I [42:03] was in London um a year year and a few [42:06] months ago is from European founders is [42:08] where they should be based and where [42:09] they should be building their companies [42:10] and I don't have the right answer. Um [42:12] it's a controversial topic. You guys are [42:14] based in London. So um but I'm curious [42:16] what you think and like as you're [42:18] thinking about this question. [42:19] >> Go [42:21] this is a tough one. Uh [42:24] it's tricky. I think [42:26] we absolutely love San Francisco. Uh [42:30] we've said it many times. Every time we [42:32] come here, [42:32] >> it's the best place in the world. [42:34] >> There's something different about it. [42:35] And and we always thought, you know, [42:38] being from the UK, this San Francisco [42:41] buzz, Silicon Valley, [42:43] >> it's a bit of a myth. At least that's [42:45] what I thought growing up. And then I [42:47] think what was interesting with uh with [42:49] Fancy uh we were a remote batch, right? [42:52] So we didn't come to San Francisco. Um [42:54] and then when we came here again for for [42:56] model ML's batch, just being in San [42:58] Francisco and being a part of the [43:00] community, it was crazy. You know, Chaz [43:02] always tells a story when, you know, [43:04] you're at the gym during the batch and [43:07] you just had people on the treadmill on [43:08] their laptops, you know, on Zoom calls, [43:10] you know, talking about funding rounds [43:11] and X, Y, and Zed. And in the UK that [43:14] just doesn't happen. Um, and look, [43:16] clearly [43:17] >> clearly, [43:18] >> um, so look, we love the UK. Um, if if [43:21] we were to start a company again, I [43:23] mean, we're trying to convince people to [43:24] to move to SF for this company. Um, we [43:27] think it's an incredible place to build [43:29] a company. Um, and like we say about [43:31] work ethic, it is really challenging, we [43:34] think, in the UK in particular, um, to [43:36] find people who frankly want to work [43:39] this hard. Um, I think in San Francisco [43:41] it's a lot more common. I think in [43:43] Europe it's just not um one of the [43:47] positive things about Europe and we [43:49] always said this is is talent and [43:51] particularly engineering talent which [43:53] might sound counterintuitive I think you [43:55] got amazing engineering talent you know [43:57] in the Bay Area [43:59] >> the problem is it's very very expensive [44:02] >> and the competition is just so so high [44:04] you know you're going up against [44:06] >> you end up hiring from your own network [44:07] anyway you're trying to bring people [44:08] over from the UK exactly um whereas in [44:11] the UK I think the level is still really [44:15] really strong but the competition is [44:17] less. So your ability to hire that top [44:20] class talent [44:21] >> I think is probably stronger in Europe. [44:23] >> It's a trick question. If you're based [44:25] in Europe you I've been given advice. [44:26] Okay. If you decide to stay in Europe [44:28] try to move to one of the few cities [44:30] where there are other really ambitious [44:32] people and there's just not that many of [44:33] them. Uh London is one of them. But [44:35] Europe is not one country. So, it's like [44:37] a big culture shift to to or maybe not [44:39] big, but it's a big change to move from [44:41] um I don't know uh Spain to to London. [44:44] It's not that big of a difference to [44:46] move from Spain to San Francisco. [44:47] >> Yeah. And I also think as we spoke about [44:49] our hiring process earlier, you've just [44:51] got to be really rigorous with your [44:53] hiring process. You know, there are [44:54] amazing people and amazing talent in [44:57] these cities. You've just got to find [44:58] them. Um so, I think don't settle for [45:00] second best. You really got to find [45:02] those people that are self motivated, [45:04] hungry. Uh, chances are if they know [45:06] about YC, they've got the right [45:08] attitude. [45:08] >> And I suspect a lot of your customers [45:10] are in the US. [45:11] >> Pretty much all of them. I'd say 80% of [45:12] our customers are US. [45:13] >> How do you handle that? [45:14] >> Well, I spend the bulk of my time in New [45:16] York. An spends the bulk of his time in [45:18] London. Um, and then kind of split [45:20] between Hong Kong and and San Francisco. [45:23] Interestingly, actually, for finance, as [45:25] I mentioned this when I came in earlier, [45:27] there's a surprising number of decision [45:29] makers for global firms around sort of [45:31] technology implementation in San [45:33] Francisco. It's not out of New York or [45:35] out of London. It's out of San [45:36] Francisco. So, I think if we'd known [45:39] more about that um you know back when we [45:42] made the decision to move the [45:43] engineering team to London, that would [45:44] have influenced things. But uh yeah, my [45:46] view for what it's worth is I think [45:48] people should do what what it takes to [45:50] to move to San Francisco if that's not [45:52] possible or that's not where your [45:53] customers are based at the very least as [45:54] you said, move to a tier one city and [45:57] try and be as close to uh you know folks [46:00] that are also building stuff. [46:02] >> Awesome. Thank you so much for coming. [46:03] It's great to see you guys. [46:04] >> Yeah, great to see you again. [46:05] >> Thanks for [46:07] [Music]