Why
Most company histories are told as a sequence of wins. This one is told as a sequence of decisions with their premises attached, and that is what makes it usable by someone building a much smaller thing. Each decision names the trade it made: short-term app revenue for an ecosystem, dilution for control, a dead logistics bet for focus, code duplication for native quality once LLMs made duplication cheap. The video's own closing lesson (20:42) is the one worth keeping: a stack or a policy should be reversed not because the tool got worse but because the cost-benefit structure underneath it changed. That is a test you can run on your own architecture every quarter, and it is the opposite of both fashion-driven rewrites and loyalty-driven inertia.
How it works
The origin: a problem the founder had every day
Lütke came through a Siemens-subsidiary apprenticeship, rotating through restaurant, accounting and warehouse work, then burnt out on financial back-end development and set out to sell snowboards online (00:54). The shop builders of 2004, Yahoo Store among them, were so closed that changing a header colour was a struggle (01:58). Rails had appeared in July 2004; he spent four months building his own shop system on it (02:16). Without a work visa in Canada, registering a business was the only legal path, and when the software solved his need completely the company pivoted from selling snowboards to selling the shop (02:33), launching in 2006 with the mission that a small business with good tools can compete with a large one (02:50).
Six decisions
- The product pivot (2006) (03:08). Stop selling boards; sell e-commerce infrastructure.
- "If you don't make money, we don't" (2006) (03:44). No fixed subscription at first, a 3 percent transaction fee only. Because the company's revenue tracked merchant success, it spent 29.8 million of the 90 million dollars it raised, turned profitable in four years and reached an IPO (04:26).
- Ecosystem before revenue (2009 on) (04:45). The app store opened with nine apps; the company gave up near-term app revenue to grow third-party developers. It wrote Liquid, a sandboxed template language so merchants could edit their own designs, and open-sourced code it had built for itself, ActiveMerchant and DelayedJob among them (05:23). Instead of marketing spend it ran a build-a-business competition with Tim Ferriss to create a pool of future merchants; one entrant, Harley Finkelstein, later became president (05:57).
- Dual-class and founder shares (2015, 2022) (06:54). With founders at 14.6 percent at the IPO, Shopify introduced 10-vote shares; in 2022 shareholders approved a founder share that keeps the founder's voting power from falling below 40 percent regardless of dilution (07:31).
- Killing a 2.7-trillion-won logistics bet (2022 to 2023) (11:56). Betting on pandemic e-commerce growth, Shopify bought the logistics company Deliverr for about 1.972 billion dollars, and it failed. Lütke said publicly that his bet was wrong, wrote the book value to zero after eleven months and sold the logistics business to Flexport for a 13 percent stake (12:14). Logistics was a side quest; to refocus on the main quest headcount went from 11,600 to 7,600, a 34 percent cut, while gross merchandise volume roughly doubled over three years (12:51).
- Fix the tool, do not replace it; then AI (2024 to 2026) (14:24). Shopify still runs one Ruby on Rails monolith of about 2.8 million lines and deploys to production more than 40 times a day. Rather than migrate to microservices or another language, it contributed to Ruby's core and improved performance by more than 90 percent (15:48). In 2025 it set the rule that before asking for people or budget you must show why AI cannot do the job, and removed spending caps on AI coding tools (16:44). In September 2026 it dropped five years of React Native and returned to native iOS and Android, because writing platform-specific code with LLMs had become cheaper than the cost of a cross-platform layer, which reversed the premise the original choice rested on (17:29).
| Decision | What it traded away | What it bought |
|---|---|---|
| 3 percent fee, no subscription | predictable revenue | incentives aligned with merchants; profitability on 30 percent of the capital raised |
| Ecosystem first | app-store revenue | developers and merchants who stayed |
| Founder shares | one-share-one-vote norms | long-horizon decisions against quarterly pressure |
| Exit logistics | 1.97 billion dollars and 4,000 jobs | focus; GMV doubled anyway |
| Keep the monolith | the fashion of microservices | 40 deploys a day and a faster Ruby for everyone |
| Native over React Native | code sharing | quality, now that LLMs write the duplicate |
The four lessons the video draws
Start from a pain you feel daily rather than from market research (19:54). Judge the business on operating income, because net income carries one-off swings such as gains and losses on equity stakes (20:10). Admit a wrong bet fast and reverse it publicly, even at a loss (20:27). Reverse a stack or a policy when the premise changes, not when the tool disappoints (20:42).
Where it lands in Jayverse
- Verex: price like decision 2. A prediction-market exchange that earns a fee on filled volume and nothing on idle accounts is already aligned with its traders; write that alignment into the pricing page and resist subscription tiers until the alignment is proven.
- Every repo: the monolith rule. Jayverse runs small TypeScript services on one stack. The Shopify lesson is that the next performance problem is more likely solved by improving the shared tooling (the Foundry scripts, the Next.js build, the Anvil devnet) than by splitting a service, and that a split needs a stated premise.
- Dark Horse and the Auditor: every bet gets a premise line. Canton test usage, the Base App mini app and the boundary-file pilot should each carry one sentence naming the premise they rest on, so the Deliverr test can be run: when the premise dies, the bet is written off, not defended.
- Invest: read operating income, not net income. The video's point about equity-stake swings distorting net income belongs next to the Invest items on reading company results, and it is the rule for judging any platform company in the Number readings.
Verified and unverified
Verified on 2026-09-19 from the public record: Shopify grew from Lütke's Rails snowboard shop, launched in 2006, added its app store in 2009, wrote Liquid and open-sourced ActiveMerchant and DelayedJob, listed in 2015 with multiple-voting shares, adopted a founder share in 2022, bought Deliverr in 2022 for about 2.1 billion dollars in cash and stock and sold its logistics arm to Flexport in 2023 for an equity stake, and cut roughly 20 percent of staff in 2023 after 10 percent in 2022; it is known for running a large Rails monolith and for Lütke's 2025 memo requiring teams to show why AI cannot do a task before asking for resources. Taken from the video and not independently checked: the 3-percent-fee-only early model, the 29.8-of-90-million and four-years-to-profit figures, the 14.6 percent founder stake, the 34 percent cumulative cut and GMV doubling, the 2.8-million-line and 40-deploys figures, the more-than-90-percent Ruby improvement, and the September 2026 React Native reversal, which is the newest claim and should be checked against Shopify's engineering blog before being repeated.
Sources: 헤이제임스, 플랫폼 하나 만들어줘 — 1인 개발자가 시작해 수백조 기업이 된 쇼피파이 | 20년 동안 내린 결정 6가지, YouTube · Shopify investor relations and engineering blog for the founder share, Deliverr and Flexport transactions · related items: Tech #62 (agentic engineering writes the boundaries), Invest 800 to 809 (chart basics), Life #2 (the agent is a good sergeant).
Key expressions
| Expression | 뜻 · 쓰이는 자리 |
|---|---|
| apprenticeship | 도제 과정(현장을 돌며 배우는 직업 훈련) · 독일식 교육 제도를 말할 때. "a German apprentice-trained developer" |
| burnt out (on) | ~에 번아웃된 · 일에 지쳐 의욕을 잃은 상태. "burnt out on financial back-end work" |
| pivot | 피벗(사업 방향 전환) · 스타트업 표준 어휘. "pivot from snowboards to shop software" |
| transaction fee | 거래 수수료 · 구독료(subscription)와 대비되는 과금 방식. "a 3 percent transaction fee and no subscription" |
| aligned with | ~와 일치한(이해가 같은 방향인) · 인센티브 설계의 핵심 표현. "incentives aligned with merchants" |
| ecosystem | 생태계(플랫폼 위의 서드파티 개발자·앱·상인) · 플랫폼 전략 어휘. "grow an ecosystem before revenue" |
| third-party developer | 서드파티 개발자(플랫폼 소유자가 아닌 외부 개발자) · 앱스토어 문맥. "grow third-party developers" |
| sandboxed | 샌드박스에 갇힌(격리되어 해를 못 끼치는) · Liquid 같은 안전한 템플릿 언어를 설명할 때. "a sandboxed template language" |
| dual-class shares | 복수의결권 주식(의결권이 다른 두 종류의 주식) · 지배구조 용어. "Dual-class and founder shares" |
| dilution | 희석(신주 발행으로 기존 지분 비율이 줄어드는 것) · 자본 조달의 부작용. "regardless of dilution" |
| write off / write … to zero | 상각하다 / 장부가를 0으로 만들다 · 실패한 투자의 회계 처리. "wrote the book value to zero" |
| book value | 장부가(회계상 기록된 가치) · 시장가와 대비. "wrote the book value to zero after eleven months" |
| side quest / main quest | 사이드 퀘스트 / 메인 퀘스트(게임 용어를 빌린 부업/본업) · 뤼트케의 표현. "Logistics was a side quest" |
| headcount | 인원 수(직원 수) · 감원·채용 문맥의 표준 단어. "headcount went from 11,600 to 7,600" |
| GMV | Gross Merchandise Volume(총거래액, 플랫폼을 거친 거래 총액) · 이커머스 핵심 지표. "gross merchandise volume roughly doubled" |
| monolith | 모놀리스(하나의 배포 단위로 된 큰 코드베이스) · 마이크로서비스와 대비. "one Ruby on Rails monolith" |
| microservices | 마이크로서비스(작은 서비스들로 쪼갠 구조) · 유행이 된 아키텍처. "migrate to microservices" |
| premise | 전제(결정이 딛고 있는 가정) · 이 항목의 핵심 단어. "reversed the premise the original choice rested on" |
| operating income vs net income | 영업이익 대 순이익 · 본업 건전성은 영업이익으로 본다는 규칙. "Judge the business on operating income" |
| one-off | 일회성의(반복되지 않는) · 회계·실적 읽기에서. "one-off swings such as gains and losses on equity stakes" |
| cost-benefit structure | 비용-편익 구조 · 결정을 뒤집을 조건을 말할 때. "the cost-benefit structure underneath it changed" |