Notes — Tobi Lütke on Creating Shopify for Americans as a German in Canada

Notes on Tobi Lütke in conversation with Tyler Cowen — Conversations with Tyler (https://conversationswithtyler.com/episodes/tobi-lutke/), 18 September 2024.

Four questions [Adler frame]

Q1 — What is it about as a whole? Tobi Lütke, co-founder and CEO of Shopify, uses his own biography — a German who apprenticed as a programmer, moved to Canada, and built an American-facing company from Ottawa — as a lens on national culture and company building. The conversation moves from meeting mechanics and venting to a three-way comparison of German, Canadian, and American work culture, then to why Germany produces few tech giants despite deep technical talent, Canada’s per-capita GDP stagnation and its habit of exporting raw materials rather than finished products, untranslatable German words, and finally a sustained argument that companies are complex adaptive systems best run by engineers who combine trust with AI-assisted verification.

Q2 — How is it argued? Almost entirely by cultural contrast and personal anecdote rather than data. Lütke repeatedly triangulates three reference cultures — Germany, Canada, the United States — using Shopify’s own hiring and customer patterns as evidence (60 percent of new engineers in one year were Canadian ‘boomerangs’ returning from American companies; Shopify sold better in the US than Canada for its first decade). Cowen’s method is comparative probing: he supplies an alternative comparison case (Netherlands and Sweden against Germany; Canada’s Olympic ‘Own the Podium’ programme) and asks Lütke to explain the divergence, repeatedly pushing him past the easy answer (‘comparing to the United States is a bad idea’) toward more specific causal claims.

Q3 — Is it true, in whole or part? Lütke is candid about the limits of his own explanatory power — he says outright, more than once, that he does not know and would rather ask Cowen the question back [§ Why Germany produces no tech giants]. [?] His causal claims (optimism as the master variable behind national tech-founding rates; the ‘referees to builders’ ratio explaining Canada’s productivity slowdown; SR&ED tax credits penalising commercialisation) are stated with conviction but are single-source, anecdotal, and not tested against counter-evidence in the conversation itself. The empirical claims that are checkable — Canada’s per-capita GDP softness, the scale of recent immigration relative to housing growth, Shopify’s own hiring composition — are consistent with what is publicly known, though Lütke gives no citations. [?]

Q4 — What of it? The episode is the wiki’s first treatment of Lütke through a national-culture and economic-policy lens, distinct from the product-and-management register of Tobi Lütke on First Principles Thinking, Maximising Human Potential, and Product Delight. Its most durable contribution is the closing argument that engineering-trained founders build companies differently because they treat organisations as nondeterministic systems to be instrumented, tested, and iterated — ‘trust plus Copilot and automated verify’ — which extends First Principles Thinking into an explicit theory of organisational design.

Glossary

Verschlimmbesserung — a German compound noun for an attempted improvement that makes the thing worse; Lütke links it to Chesterton’s Fence, since both describe removing or changing a part of a system without understanding why it was there. [§ Untranslatable words and Chesterton’s Fence]

SR&ED tax credits — Canada’s Scientific Research and Experimental Development programme, which subsidises original research but, in Lütke’s account, imposes heavy documentation burdens once the work turns commercial — so it rewards invention but penalises the harder work of turning invention into a shipped product. [§ Canada’s optimism deficit and the raw-material economy]

‘Own the Podium’ — a Canada-wide programme launched around the Vancouver 2010 Winter Olympics that gave national athletes explicit permission and resourcing to aim for gold rather than a respectable finish; Lütke’s example of a ‘go-for-bronze’ culture temporarily overridden by top-down ambition. [§ German bluntness versus North American iteration speed]

Trust plus Copilot and automated verify — Lütke’s phrase for his preferred model of company design: give employees real trust and latitude (raising the ceiling on what excellent work looks like) while using AI assistance and automated checks (linting, tests) to catch errors, rather than defaulting to rigid process that raises the floor but caps the ceiling. [§ Companies as complex adaptive systems]

Key claims by section

Outsider mindset and the mechanics of meetings [§ Outsider mindset and the mechanics of meetings]

Lütke traces his working style to growing up an outsider in a small German town with no interest in computing, learning to observe skilled people from a distance — via, among other things, the archived Linux kernel mailing list.
His preferred way of extracting information in meetings is to state a high-conviction, possibly wrong claim and let others correct it — a technique he compares to posting a confident wrong take online to prompt correction; Shopify’s internal wiki documents this as a known trait of his.
He treats venting as a legitimate, under-acknowledged psychological outlet that social media distorted by exposing private venting to permanent public record; the fix he floats is naming venting explicitly when it starts, so listeners do not mistake it for a considered position.

German bluntness versus North American iteration speed [§ German bluntness versus North American iteration speed]

Germans, in Lütke’s account, are direct without the ‘shit-sandwich’ softening common in North America, and hold a higher, more binary quality bar — a product is either below an acceptable standard or genuinely excellent, with little regard for anything in between.
North America’s comparative strength is willingness to ship an early, imperfect version and iterate quickly rather than trying to perfect a product before release; Lütke has watched Shopify colleagues split along these cultural lines when debating whether to ship.
Canadians sit closer to Americans on minimum quality bar but suffer a ‘go-for-bronze’ mentality that Lütke attributes to being a smaller country adjacent to a much larger one; he cites Canada’s ‘Own the Podium’ Olympic programme as proof that explicitly granting permission to aim for world-class results can override this default, and credits much of Shopify’s own success to himself and his co-founder giving colleagues that same permission.

Why Germany produces no tech giants [§ Why Germany produces no tech giants]

Lütke rejects the idea that Germany lacks technical talent — he cites Shopify and Palantir as counter-examples founded by Germans — and instead points to a cultural belief, especially outside Germany’s tech hubs, that ‘tech is something Americans do’ and a broader pessimism about the future that he thinks is a precondition for wanting to build it faster.
Startups are a low-status career choice in Germany, so risk-tolerant talent willing to leave a stable career for an uncertain venture is scarcer than in North America; Germany’s apprenticeship system (which Lütke went through himself) produces excellent craftsmanship but channels talent toward craft roles rather than founding ambitious companies.
Pressed on why the Netherlands and Sweden — culturally close to Germany — have produced more scaled tech companies (ASML, Adyen), Lütke declines a confident answer but repeats optimism as the variable he would weight most heavily, alongside a possible correlation between a country’s dominant flavour of humour (ironic and dark, in Germany’s case) and its tech-founding culture. [?]

Canada’s optimism deficit and the raw-material economy [§ Canada’s optimism deficit and the raw-material economy]

Lütke reads Canada’s per-capita GDP softness as inseparable from a decline in national optimism since the high-trust, multiculturalism-as-success era he associates with Pierre Trudeau; he flags a rising ratio of ‘referees to builders’ — public-sector growth relative to private building — as a plausible contributing mechanism, without claiming certainty about causation.
His central metaphor is that Canada exports raw materials rather than finished products across every domain: beaver pelts once went to London to be turned into hats; Canadian oil is refined in Houston; Waterloo, one of the world’s best computer-science schools, exports graduates who become ‘load-bearing’ for Silicon Valley companies rather than staying to found Canadian ones. He extends this to research policy: SR&ED tax credits reward original research generously but burden commercialisation with documentation, so good people avoid the commercial side.
He supports Canada’s existing points-based skills immigration system and worries current backlash conflates two separate problems — a genuinely mismanaged recent immigration surge (three million people in three years against almost no new housing) and the underlying, historically popular, policy of skilled immigration — risking a ‘baby out with the bathwater’ overcorrection.

Untranslatable words and Chesterton’s Fence [§ Untranslatable words and Chesterton’s Fence]

Lütke’s favourite untranslatable German word is Verschlimmbesserung — an attempted improvement that makes something worse — which he connects directly to Chesterton’s Fence: the danger of changing a system component without understanding the (possibly cultural or technical) reason it exists.
He notes the reverse gap too: German has no native word for ‘entitlement’, which he has had to explain to Shopify interns using the gratitude/entitlement spectrum as a deliberate framing device.
He reads originally-German texts (Wittgenstein’s Tractatus, Kant) in the original when he can, but values retranslation of classic texts roughly every 25–50 years to keep them accessible to new generations, and regrets having lapsed on his self-imposed rule of reading one German book a year.

The future of retail and goal-oriented software [§ The future of retail and goal-oriented software]

Lütke is skeptical that virtual reality will replicate literal storefronts online; he expects VR’s retail impact to run instead through virtual avatars and digital twins of products — furniture, where seeing a couch rendered in your own living room already beats an in-store view, is the clearest present case, and he flags Gaussian splats as a technology that will make producing such twins cheap.
His central AI claim is that software has spent roughly 20 years (since Web 2.0) building interfaces that let people toggle and configure settings to express a goal indirectly; the next phase is goal-oriented software, where a person states the goal directly and the system works with them toward it — a shift he compares to the qualitative jump AI copilots are already delivering for engineers.
He frames retail pricing as too heterogeneous across businesses for economists’ sticky-price models to average meaningfully — pricing decisions blend genuine aesthetic intuition (compressed expertise) with pure guesswork — and uses this to make a broader argument that fields imitating physics’ search for one simple underlying equation often misapply that ambition to genuinely complex domains.

Companies as complex adaptive systems [§ Companies as complex adaptive systems]

Lütke argues companies are complex adaptive systems, not scaled-up military-style hierarchies, and that engineers who found or run companies tend to outperform because engineering — properly understood — is 30 years of practice turning nondeterministic systems into more deterministic ones, a skill that transfers directly to organisational design.
He expects today’s companies to look embarrassingly primitive in retrospect, much as his own apprenticeship-era belief that software, once shipped, could never meaningfully change again now looks wrong next to 20-year-old, actively maintained codebases; the analogy is a soccer team executing at 80 percent of a theoretical perfect game — real ceilings exist, but most companies are far below even that.
His closing thesis names the mechanism for improving that ceiling: process and policy raise the floor (preventing catastrophic failure) but simultaneously lower the ceiling (capping how good the best work can be), because they are built from low trust. ‘Trust plus Copilot and automated verify’ — extending real latitude to employees while using AI assistance and automated testing/linting to catch errors — is his model for raising the ceiling without abandoning the floor.

Reading list and third-party retailer mistakes [§ Reading list and third-party retailer mistakes]

Lütke names Thomas Sowell’s A Conflict of Visions, James Burnham’s The Managerial Revolution and The Machiavellians, and James C. Scott’s Seeing Like a State as recent influences, alongside a habit of reading fantasy (Brandon Sanderson) for its function as a simplified simulation of society with the variables changed.
He identifies the most common modelling mistake among Shopify’s third-party retailers as trying to build products ‘other people like’ in general, rather than serving a specific, dedicated cluster of customers — his framing is Kevin Kelly’s ‘1,000 True Fans’ essay applied at commerce scale, and he argues customers should describe their problems, not dictate solutions, since ceding that judgement is an abdication of a founder’s vision.
He credits Shopify’s own growth partly to structuring incentives so the company and its merchants are ‘on the same side of the table’ economically, which is why product managers maintain hundreds of active WhatsApp conversations with fast-growing merchants as a direct inspiration channel.