Building Linear: craft, focus, and opinionated software

Executive overview

Most software teams treat quality as a post-launch polish pass. Linear treats it as a structural constraint from day one — baked into hiring, product reviews, and how projects are scoped.

Karri Saarinen (co-founder and CEO of Linear) argues that opinionated, well-crafted software is not a luxury but a competitive baseline: the bar rises permanently with each new entrant in a category. Linear ships to internal users first, then a small beta cohort, then general release — using craft reviews rather than A/B tests or metrics to decide when something is ready.

The core insight: smaller teams of high-caliber, product-minded people outperform larger specialised ones — but only if the culture genuinely values craft and everyone stays close to customers.

Craft and quality

• Craft starts with hiring people who care about it; it cannot be mandated top-down.
• Founder disagreement at Airbnb, Facebook, and Amazon on quality standards is a company-killer — align early.
• Linear's review process: demo with project lead, then Karri or a co-founder uses the feature directly in different states before release.
• Paper cuts and friction destroy collaboration tools — quality is a business requirement, not aesthetics.
• Releasing janky internal builds early is not contradictory to craft; it enables faster iteration before general release.
• Ownership without a PM layer means engineers notice and fix micro-details (e.g., the diagonal "safe zone" on sub-menus) because they live in the product.

Opinionated software and the Linear method

• Opinionated software gives users good defaults so they don't waste time configuring the tool instead of doing the work.
• Flexible software forces every team to invent their own workflow — inconsistency across the org is the cost.
• Design for someone, not for everyone; generalised solutions are sub-optimal for all use cases.
• Cycles (Linear's equivalent of sprints) run on an automated schedule — no admin overhead, just "these are the things we're working on this period."
• Cycles exist to defend focus: when something urgent arrives, the team has a documented baseline to explain trade-offs.
• No feature-level metrics goals; the success criterion is "customers agree the problem is solved."
• Company-level metrics (e.g., weekly active users) exist, but no metric drives individual feature decisions.

Product management without PMs

• Linear has one PM (head of product); engineering and design own project scoping, communication, and trade-offs.
• Project leads rotate by assignment, not by seniority or title.
• The cost: engineers must be hired for product sensibility, not just technical skill.
• The benefit: people with broader scope ship higher-quality work and fewer things fall through role gaps.
• Interview for product thinking — ask candidates to go deeper on past decisions, not just describe them.
• Paid work trial (a few days, co-scheduled to reduce friction): candidates get Slack and Notion access, tackle a vague problem in the real codebase, and present their approach.

The magic and science of building product

• "Science" at Linear = talking to users constantly, shared Slack channels with customers, founders answering support questions.
• "Magic" = when everyone in the company builds deep product and customer empathy, intuition replaces data-driven decision-making.
• No A/B tests; using data to make a choice for you can produce the right metric and the wrong product.
• Working with 1–10 design partners per feature (e.g., Vercel on roadmap) surfaces how companies actually operate, which is hard to imagine in the abstract.

Focus and avoiding side quests

• The YC heuristic: talk to customers, build the product, exercise — if you're doing anything else, question it.
• Frame decisions as main quest vs. side quest: does this progress building the product and serving customers?
• Timing matters: decline things that aren't wrong, just not right yet (e.g., SOC 2, podcast appearances).
• Never more than doubled headcount in a year; slow hiring preserves culture and reduces coordination overhead.

Growth and product-market fit

• April 2019: company founded, waitlist launched on Twitter with a short survey (tools used, company size, intent).
• May 2019: internal dogfooding; June 2019: first cohort invitations (~10 per week, growing slowly).
• Cohort-based beta: invite a small group, fix the bugs they surface, then invite the next — avoids flooding early software with the same issues simultaneously.
• June 2020: public launch with pricing; nearly all private-beta companies converted to paid.
• PMF is a spectrum, not a binary — Linear first targeted early-stage startups, then expanded to larger companies.
• When a segment shows strong pull (crypto companies, then AI companies), double down before expanding.
• Linear's moat is retention and trust: companies that adopt early should stay forever, which requires sustained quality.

Hiring philosophy

• Never more than doubled headcount in a year.
• Every role requires breadth beyond the job title — engineers who think like PMs, marketers who think like storytellers.
• Narrow specialists create "not my job" gaps; generalists with depth don't.
• Paid work trial is the final step: vague problem statement, real codebase, short time frame — tests how candidates think under ambiguity, not just what they know.
• Most candidates, including those currently employed, find ways to schedule it; almost none have declined.