"The more a domain matters, the more design matters. When it's new, rough is fine. When there are 100,000 competitors, you need craft to be considered at all."
Each new paradigm (mobile, web) starts rough but matures fast
Design becomes table stakes as markets mature
Today, users expect craft from day one of any product
Framework
The Linear Method: Building with Taste
0
A/B tests run
1
head of product
No A/B tests: Decisions made on taste and customer understanding, not data
Minimal PMs: One head of product (Nan Yu). Ownership distributed to eng + design teams
Early shipping: Features shipped internally in week 1, then to opt-in users, then full release when crafted
Project-based teams: Small cross-functional teams assemble around features, then disperse
Deep customer context: Whole team talks to users, reads feedback, understands the problem
The founders' pactSuccess requires alignment on craft as a value. If founders don't believe it, the company can't build it. Domain matters too—Linear is in the retention business, so quality compounds every day.
Mechanics
How Craft Actually Gets Built
Week 1 ship: Design sketched, then immediately built and shipped to production (internal only)
Internal feedback: Team uses the rough feature, spots friction and polish gaps
Customer co-creation: Find 1–10 customers interested in early access, iterate alongside them (e.g., Vercel for roadmap features)
Trust over data: When someone sees a detail is wrong, we fix it. No metrics required
Taste is taught: Example: Andreas (engineer) improved the macOS submenu interaction by adding dynamic safe zones—no one asked for it, he just cared
The secret sauce
Craft happens when you remove the PM gatekeeper and give teams ownership. Engineers and designers see the whole problem and naturally optimize for quality.
The trade-off
Shipping takes longer in some dimensions (polish phase), but you ship faster in others (no reviews, no meetings). Net effect: faster quality product, not slower.
First Principles
Decision Framework
Magic + science: Talk to customers religiously (science), then use taste to decide (magic). Neither alone works.
Main quests only: Avoid side quests. Ask: "Does this help the customer? Does it help the product?" If no, it's a distraction.
Timing over perfection: Some features aren't important yet (e.g., SOC 2). Say no and move on.
Fail forward: You might make wrong choices. That's fine. Fix it and move. Data didn't make the choice for you.
YC principle (Karri's north star)Talk to customers. Build the product. Exercise. Don't do anything else.
Contrarian
The Anti-Playbook
✗Metrics drive product decisionsINSTEAD →✓ Customer understanding + taste drive decisions. Metrics answer specific questions, not the reverse.
✗PMs make the key decisionsINSTEAD →✓ Engineers and designers own the project. They see the details. They care. They ship better.
✗Ship fast, polish laterINSTEAD →✓ Ship fast internally, polish before users see it. The end result is a beautiful product, not a broken one.
✗All ideas are worth exploringINSTEAD →✓ Say no to side quests. Craft only emerges when you say no to 90% of ideas and go deep on the 10% that matter.