20VC: $1BN ARR in 18 Months; The Untold Story of Higgsfield | Spending $4M Per Month on Models | Why Moats in AI are BS | Scaling a Content Team to 150 People with Alex Mashrabov

这期播客节目采访了 Hicksfield 的创始人 Alex，他讲述了自己从哈萨克斯坦的一名竞技程序员，成长为一家预计年收入将达到 10 亿美元的公司的领导者的历程。Alex 分享说，他的父母灌输了一种信念，认为美国是技术发展的重要之地，这促使他在 19 岁时成为全球前三的竞技程序员。 他的第一个创业公司 AI Factory 建立了用于语言翻译的预 Transformer 架构，后来他以 1.66 亿美元的价格将其卖给了 Snap。这次收购让他得以搬到美国，实现了他的梦想。Alex 指出，虽然旧金山提供了精英管理和多样性，但他发现硅谷的投资者群体“极其趋同”。 在 Snap 领导生成式 AI 之后，Alex 认识到社交媒体广告领域存在一个巨大的未满足需求：公司难以跟上内容生产的速度。他创立了 Hicksfield 来解决这个问题，押注于视频 AI。该公司最初烧掉了其 1600 万美元种子资金中的 1000 多万美元，未能找到产品与市场的契合点。在去年 3 月 31 日推出产品后，通过转向专注于产品主导增长（PLG）和客户反馈（特别是对 AI 中摄像头控制的需求），Hicksfield 立即取得了成功。 自那时起，Hicksfield 仅在 18 个月内实现了 10 亿美元的年化收入，超越了 Coursor 达到这一里程碑所需的 24 个月时间。Alex 通过将过去四周的收入乘以 13、按比例计算年度合同，并且只计入实时收入来计算这一数字。他指出，超过 50% 的收入来自企业，其中很大一部分来自有抱负的创作者和自由职业者。一个客户显著地从每月 99 美元扩展到每年 600 万美元的交易。Hicksfield 的企业部门在第 12 个月实现了超过 300% 的净收入留存率（NRR），尽管在那些未能立即理解其价值的消费者中，初期流失率达到 30%。 该公司的市场进入策略严重依赖内容和分发，没有付费广告。他们内部雇佣了 150 多名创意专业人士。Alex 最初致力于构建专有模型，但意识到追逐基准测试是一种“企业心理战”，常常导致被操纵的结果。他发现专注于客户驱动的解决方案，并利用经过特定数据后训练的开源模型，可以提供更好的利润（开源模型超过 80%，闭源模型为 20-30%）和更高的客户成本效益，从而形成一种“代币经济学”方法，将客户工作负载路由到最有效的模型。 Hicksfield 内部模型使用量每月超过 400 万美元，个人工程师和创意人员每月花费超过 10,000 美元，有些甚至一周花费 30,000 美元。Alex 预计“10 倍工程师”和创意人员的个人支出将增加。虽然最初预计 AI 会取代法律和客户支持等角色，但他现在认为 AI 正在赋能这些团队，AI 负责第一线防御，而人类代理则管理由于产品快速迭代而增加的复杂性。 谈到护城河，Alex 认为它们存在于提供切实的成果（例如推动销售的 AI 广告）和网络效应中。Hicksfield 培育了一个分享和基于开源项目进行建设的社区，旨在创造“滚雪球效应”。Alex 认为公司的估值可能被低估了，因为它不是硅谷的内部人士，暗示一家以这种速度增长的 10 亿美元收入公司可能价值 250 亿美元或更多。 他的管理风格受到 Jensen Huang 和 Elon Musk 等领导者的影响，强调“不废话”、注重细节的方法，而非传统的企业管理原则。他专注于招聘最优秀的人才并赋能他们。根据他在亚洲和哈萨克斯坦的经历，他强调了祖国丰富的人才储备，尤其是在数学和物理领域，以及 15% 个人所得税的优势，这有助于公司 300 多名员工的多样性。 Alex 承认做出了巨大的个人牺牲，每周在 Hicksfield 工作 80-90 小时，受到对技术的深刻信念和“巨大的错失恐惧”的驱动。他预测大多数社交媒体内容将很快由 AI 生成，Hicksfield 的内部预测预计未来 12 个月收入将达到 45 亿美元，尽管他个人认为，由于 AI 在营销和电影内容中的日益普及，这一数字可能超过 100 亿美元。他的最终目标是让 Hicksfield 成为一家比 AppLovin 和 Shopify 更大的上市公司。

The podcast features Alex, founder of Hicksfield, discussing his journey from a competitive programmer in Kazakhstan to leading a company projected to hit $1 billion in annual revenue. Alex shares how his parents instilled a belief that the US was where technology mattered, leading him to become a top-three competitive programmer globally by age 19. His first venture, AI Factory, built pre-transformer architecture for language translation, which he later sold to Snap for $166 million. This acquisition allowed him to move to the US, fulfilling his dream. Alex notes that while San Francisco offered meritocracy and diversity, he found the Silicon Valley investor landscape to be "extremely consensus-driven." After leading Generative AI at Snap, Alex recognized a significant unmet need in social media advertising: companies struggled to keep up with the pace of content production. He founded Hicksfield to address this, betting on video AI. The company initially burned over $10 million of its $16 million seed funding, failing to find product-market fit. A pivot to focus on product-led growth (PLG) and customer feedback (specifically the need for camera control in AI) led to immediate success after launching their product on March 31st last year. Hicksfield has since achieved an annualized revenue of $1 billion in just 18 months, surpassing Coursor's 24-month timeline to that milestone. Alex calculates this by taking the last four weeks' revenue and multiplying by 13, prorating annual contracts, and only including live revenue. He notes that over 50% of revenue comes from businesses, with a significant portion from aspirational creators and freelancers. One customer notably scaled from $99/month to a $6 million annual deal. Hicksfield's business segment boasts a Net Revenue Retention (NRR) of over 300% at month 12, despite initial 30% churn among consumers who don't immediately grasp the value. The company's go-to-market strategy relies heavily on content and distribution, with no paid advertising. They employ over 150 creative professionals in-house. Alex initially pursued building proprietary models but realized chasing benchmarks was a "corporate psyop" that often led to gamed results. He found that focusing on customer-driven solutions and leveraging open-source models post-trained on specific data offered better margins (over 80% for open vs. 20-30% for closed) and cost efficiency for customers, leading to a "tokenomics" approach where they route customer workloads to the most efficient model. Internal model usage at Hicksfield exceeds $4 million per month, with individual engineers and creatives spending over $10,000 monthly, some even $30,000 in a week. Alex foresees this individual spend increasing for "10X engineers" and creatives. While initially expecting AI to replace roles like legal and customer support, he now sees it empowering these teams, with AI handling first-line defense while human agents manage increasing complexity due to rapid product velocity. Discussing moats, Alex believes they exist in delivering tangible outcomes (like AI ads that drive sales) and network effects. Hicksfield fosters a community that shares and builds upon open-source projects, aiming to create a "snowball effect." Alex sees the company's valuation as potentially discounted because it's not a Silicon Valley insider, suggesting a $1 billion revenue company growing at this rate could be worth $25 billion or more. His management style is influenced by leaders like Jensen Huang and Elon Musk, emphasizing a "no-bullshit", detail-oriented approach over traditional corporate management principles. He focuses on hiring the best talent and empowering them. From his experiences in Asia and Kazakhstan, he highlights the dense talent pool in his home country, particularly in math and physics, and the benefits of a 15% personal income tax, contributing to the diversity of the company's 300+ employees. Alex admits to significant personal sacrifices, spending 80-90 hours a week on Hicksfield, driven by a deep conviction in the technology and a "huge fear of missing out." He predicts that most social media content will soon be AI-generated, and Hicksfield's internal projections anticipate $4.5 billion in revenue in the next 12 months, though he personally believes it could exceed $10 billion due to increasing AI adoption in marketing and cinematic content. His ultimate ambition is for Hicksfield to become a public company larger than AppLovin and Shopify.