AI
AI Makes Shipping Too Easy
Engineering used to act as a gate. Building something took months, so teams thought carefully before starting. AI removed that gate. A single developer with Claude Code can now match a whole team’s output, and the temptation is to ship every idea that surfaces. Steve Jobs said great products come from saying no to 999 things and yes to one. When shipping costs nothing, nobody says no. The Uber hyper-growth years, firefighting and constant scaling with quality as an afterthought, are now the default mode for everyone.
Automated Bug Fixes at 10 Percent and Climbing
Linear already auto-fixes 10% of incoming bugs. When a bug arrives, an agent identifies the responsible engineer, opens a PR, and lands it without any human touching the keyboard.
“10% of our bugs are automatically fixed by you know a singleshot AI instance” — Tuomas Artman
Artman expects that number to climb toward 100% over the next few years. The candid flip side: Claude Code itself shows the cost of shipping too fast. Anthropic is in a winner-takes-all race with OpenAI, and the quality gaps are visible to anyone who uses the tool daily.
Quality Wednesdays, Zero-Bug Policy, and AI Taste Limits
One view-options menu had 35 quality problems. That discovery launched Quality Wednesdays: 25 engineers, 30 minutes, each presents one fix they hunted themselves.
“we’ve probably fixed um 2,500 or 3,000 um of these small minute details uh in the application” — Tuomas Artman
The zero-bug policy works on the same logic: bugs arrive at a constant rate whether you fix them now or in three months, so fix them now. Linear spent three weeks clearing its backlog, then enforced same-day triage. AI helps locate the problem fast, but taste is another matter. In Emil’s agent-vs-human animation comparison (21:24), the agent built correct animations but got timing wrong in ways it could not perceive.
Hiring for Product Sense and the Engineer Role Shift
Linear runs a paid week-long trial with every engineering candidate. They implement a greenfield feature; sometimes it ships the same week. Candidates who decline self-select out.
Day one, every engineer gets access to Slack channels shared with all major customers. Every customer meeting is recorded and tagged, so anyone can search for what customers actually want. Artman’s prediction: within four years, engineers who move data from one place to another will be replaced by AI. The ones who remain will understand customers. They will be mini-PMs who also write code.
Q&A
What should software engineers do to grow their product sense? Get close to customers, build something for yourself first, ship it, and read Apple’s Human Interface Guidelines for UX fundamentals. ▶ 27:54
Notable Quotes
10% of our bugs are automatically fixed by you know a singleshot AI instance Tuomas Artman · ▶ 5:55
they they don’t feel they they have no taste um they they simply don’t Tuomas Artman · ▶ 20:08
we’ve probably fixed um 2,500 or 3,000 um of these small minute details uh in the application Tuomas Artman · ▶ 14:48
Key Takeaways
- Linear auto-resolves 10% of incoming bugs with a single-shot AI agent, no engineer required.
- Quality Wednesdays require each of 25 engineers to find and fix one quality issue per week.
- Artman predicts engineers will all become product-focused mini-PMs as AI absorbs pure implementation work.