How Shopify engineers for intensity, speed, and great hiring

Executive overview

Most engineering orgs confuse long hours with high output. Farhan Thawar, VP and Head of Engineering at Shopify, argues that intensity — more kilojoules per minute, not more minutes — is what separates great teams from merely busy ones.

The levers he has pulled at Shopify are unglamorous but concrete: pair programming, aggressive meeting deletion, six-week review cycles, weekly written updates, and a relentless drive to delete code rather than accumulate it.

The core insight: intensity is a system you design, not a personality trait you hire for — and the right system lets you ship fast, go home, and still outperform everyone working twice the hours.

Choosing the hard path

• When two options exist, the hard one wins even if it fails: you learn more, work with smarter people, and build resilience.
• The easy path that fails leaves you with nothing; the hard path that fails still compounds.
• Seek "irrational" long-term thinkers as partners — know whether you are the visionary or the person who moves the ball 1% per week.
• Write a personal decision framework for jobs before a recruiter calls; title, money, and company name are distractions from what you actually care about.
• Asking "stupid" questions in public is a learnable skill; rejection reps (retail, cold calling) build the muscle.

Pair programming as a management tool

• Pair programming is two engineers, one machine — the throughput bottleneck is elegant thinking, not keystrokes.
• Toby Lütke and Shopify's first CTO set a one-hour timer: if the problem wasn't solved, they deleted all code (keeping tests) and restarted, on the premise that the right design fits in an hour.
• At Pivotal/Xtreme Labs, 40 hours per week of pairing was the norm; at Shopify it is 4–8 hours weekly, used for focused pathfinding sprints and incident response.
• Benefits: higher happiness, faster knowledge transfer, fewer silos, near-zero multitasking.
• AI co-pilots (GitHub Copilot, Cursor + Whisper) make everyone a pair programmer by default — adding a second human on top makes it even more powerful.
• Analogy: the underhanded free throw — statistically superior, universally ignored because it looks dumb.

Creating organisational intensity

• GSD (Get Shit Done): weekly written project updates to the whole company; Parkinson's Law applied at scale — weekly cadence means weekly visible progress.
• Six-week reviews: every project, every resource plan, reviewed with Toby; short enough to remember context, long enough for real progress.
• Reviews are a pairing session, not a gotcha — leaders show up with broader context to unblock, not to audit.
• "Micromanagement" is not a dirty word at Shopify when it means working on problems together rather than inspecting output.
• Leaders share weekly threads of their own work; high-energy exec team models the intensity it expects.
• Having two or three things on fire at all times is healthy — it means the company is stretching into new territory.

Deleting meetings and reducing noise

• Once a year, Shopify deletes all recurring meetings with more than two people that are internal-only — "Meeting Armageddon."
• A two-week moratorium on creating new recurring meetings follows; forces teams to ask whether they need a cadence at all.
• Result: individual contributor meeting time dropped ~50–60%, landing at roughly 3 hours per week; managers at ~6 hours.
• Announcement-style communication moved from Slack to a feed tool (Facebook Workplace); Slack reserved for real-time collaboration.
• Fewer interruptions restored flow time and drove measurable productivity gains.

Deleting code and building infrastructure

• Code is a liability; simpler codebases are faster, more resilient, and easier to maintain.
• Shopify runs a Delete Code Club and dedicates one hack-day team per cycle specifically to finding and removing cruft — consistently finding 1 million+ lines to delete.
• Framing: every piece of work is an experiment, a feature, or infrastructure. Features consume infra; infra enables a class of features.
• Toby's test: "What has to be true so anyone could build this in one hour?" — forces investment in platform layers over point solutions.
• NFT gating example: building the platform API (2–3 months) instead of the feature itself (2–3 weeks) unlocked dozens of unforeseen use cases.
• GitHub Copilot writing 1 million lines is table stakes; the real milestone is when AI starts deleting 1 million lines.

Demo culture and high-fidelity feedback

• GSD updates include live demo links (via Shopify's internal Spin environment) or beta flags — not screenshots.
• PMs record friction logs: narrated screen recordings walking through the experience as a real merchant.
• Short-circuits status debates; teams iterate on the actual experience week over week.

Hiring: work trials over interviews

• Interviews are a weak predictor of performance — both directions (great interviewer, poor performer and vice versa).
• Preferred model: job trials. The interview process should resemble the actual job as closely as possible.
• "Life story" interview: not what you did, but why you made each career move — surfaces curiosity and range.
• Generalists outperform specialists (cf. Range by David Epstein) — look for people with varied trajectories and genuine reasons for switching.
• Internship program as structured trial: 1,000 interns planned for 2025; four months of real work product tells you more than eight hours of interviews.
• Early-career hires are more intense pairing partners; they also arrive fluent in AI tooling and a different generation's shopping behaviour.
• Post-hire: a Slack survey goes out in the first 90 days. If there's a mismatch, surface it early — it is better for the person and the company.
• 120 direct reports at Xtreme Labs worked by replacing manager functions with systems: pair programming for unblocking, product backlogs for prioritisation, weekly demos for alignment, and unscheduled rather than standing one-on-ones.

The flat org and remote-first model

• Deeper hierarchies amplify misalignment — survey data shows alignment degrades with distance from the CEO.
• Shopify targets 8–20 direct reports per manager; errs toward flat rather than deep.
• Remote-first (90–95%) with intentional IRL moments: annual Shopify Summit (talks, hack days, social events) and bursts — ad hoc 3–5 person off-sites to prototype or solve a specific problem.
• Trust battery metaphor: trust between people depletes over time at a distance; IRL events recharge it. Some leaders start new hires at 100% trust and only deplete on misalignment.

Failure corner: the React Native decision

• First week at Shopify (2019): chose a hedged platform strategy — iOS in Swift, Android in React Native — for the new point-of-sale app.
• A year later, React Native was the clear winner (Shop app, web engineers reusable, single codebase). The hedge had cost roughly 18 months of work across ~100 engineers.
• Toby's response: "Tell everyone this story." He does not come down on people for taking risks that fail; he comes down on people who don't take risks.
• The lesson Farhan drew: he already knew he would go all-in on React Native (became a core contributor, ran the working group). He should have trusted that commitment from day one and not hedged.