Farhan Thawar on Intensity, Pair Programming, and Engineering at Shopify
Source: Lenny’s Podcast Speaker: Farhan Thawar Link: Episode
Overview
Farhan Thawar, VP and Head of Engineering at Shopify, covers three interlocking themes: choosing the hard path, manufacturing intensity without adding hours, and hiring for potential over credentials. The conversation is grounded in specific Shopify practices — Meetingageddon, the Delete Code Club, six-week reviews with Toby Lütke, the OK1/OK2 approval process — and in Farhan’s own career story, from Atari-coding teenager to a sequence of engineering leadership roles culminating at Shopify. The episode is unusually concrete: most claims are illustrated with a specific experiment, a headcount number, or a policy name.
Key ideas
- Choose the hard path. When a path is hard, failure still produces learning and exposure to talented people. The easy alternative — sending ten résumés a day instead of building a Shopify app — yields nothing when it fails. The asymmetry makes the hard path the rational default even before accounting for upside.
- Pair programming as a management tool. Two people at one computer produces less code but better code and accelerated knowledge transfer. Shopify runs roughly 4–8 hours of pairing per week per engineer (classified as a “pathfinding” company on the pairing spectrum). Toby Lütke’s rule: if you cannot delete and restart a piece of work within one hour, it is not well understood. AI copilots are framed as the permanent on-demand pair.
- Intensity without extra hours. More kilojoules per minute, not more minutes. Shopify’s Meetingageddon deletes all recurring meetings with three or more internal participants annually; a two-week moratorium follows; IC meeting load dropped from roughly 5–6 hours to roughly 3 hours per week. Other tools: the GSD (Get Stuff Done) weekly project update, six-week reviews with the CEO, and the OK1/OK2 two-step approval process that forces cross-functional alignment early.
- Infrastructure-first thinking. Engineering work is bucketed into experiments, features, and infrastructure. For infrastructure, the right question is: “What can you build so anyone could build this in one hour?” The Delete Code Club operationalises the inverse: hack days that routinely surface a million or more lines of code to remove. Code is a liability, not an asset.
- Hiring for potential, not credentials. The life story interview (why you made each transition, not what you did) surfaces motivation and self-awareness. Job trials replace résumé screening. The 1,000-intern programme in 2025 is a scaled work trial. Canonical hires: an ML engineer met at a coffee shop with no prior software job, still at Shopify; a waitress who became an HR director.
Related
- Evan Spiegel on Consumer Social, the Distribution Problem, and Building Snap — Loon Shots dual-org model; similar tension between small creative unit and large execution org
- Eoghan McCabe on Intercom, AI, and the Future of Customer Support — engineering culture and pace at a high-intensity tech company
- Brian Chesky on Airbnb and Product-Led Growth — founder-led intensity and deep involvement in product decisions
- Ami Vora on Product Management and Working with Engineers — engineering–PM collaboration dynamics