AI makes it easy to ship every feature request, so teams need to spend more time deciding what should exist.
2
Linear treats product quality as an ongoing engineering responsibility through Quality Wednesdays and a zero bug policy.
3
As AI handles more implementation work, software engineers will need stronger product sense and closer contact with customers.
Summary
Tuomas Artman argues that AI has removed much of the friction that once forced teams to think carefully before building. When an agent can implement almost any request, teams can fill their products with confusing or poorly designed features. At Linear, feature requests are treated as evidence of customer problems, not as specifications to copy. The team groups feedback, finds the underlying need, and designs a focused solution. Artman describes Quality Wednesdays, where every engineer finds and fixes a small product flaw, and a zero bug policy that assigns reported bugs immediately. He is direct about the limits of current agents. They can write code and tests, but they do not have human taste for timing, animation, or how an interface feels. He expects engineers to become product engineers who understand customers, choose useful problems, and build the resulting software.
Artman says agents can let teams immediately ship almost every feature request or idea. That power can produce software with a convoluted experience because engineering no longer creates the same delay that once made teams think before building. He compares the situation with Steve Jobs's advice that great products come from saying no to 999 things and yes to one. He also compares it with Uber's hypergrowth period, when the company shipped and fought fires at all costs because it was competing in a winner-takes-all market. When a small team or one person can use AI to match a larger company's feature set, Artman thinks tasteful, high-quality software becomes a way to stand apart.
Linear turns feature requests into problems to understand
Linear does not usually implement customer requests exactly as submitted. The team talks with customers, works out the actual problem, groups related requests, and looks for the root cause. It then designs a solution for that group of needs. AI can summarize large numbers of requests and suggest groupings, but Artman says it still takes human time to decide what the right thing is and to design the user experience. Linear is still moving faster in some areas. Bug fixing has accelerated, with about 10 percent of bugs automatically receiving a pull request and being landed without an engineer doing anything. Artman expects that share to rise over the next few years.
Artman uses Uber to explain why quality is hard to measure. Uber watched metrics such as revenue, trips taken, ride quality, and time from sign-up to a first trip, but revenue was the metric people focused on most. A feature such as Uber Pool could increase revenue even if it made the product harder to use. When competing products offer similar prices and functions, there may be no immediate A/B test that shows the value of better interaction details. Artman remembers an experienced iOS engineer catching a two-pixel alignment error in his first Uber pull request. Those details may not affect revenue at once, but he believes users gradually move toward the product that feels better, and the weaker product loses users over time.
Quality Wednesdays train engineers to notice small defects
At Linear, every engineer joins a weekly Quality Wednesday and brings one quality fix they found themselves. The fixes range from a one-pixel adjustment to a more efficient backend implementation. The practice began after Artman showed the team a small part of the application and asked them to inspect it closely. In one view option menu, the team found 35 problems. Artman says the company has since fixed roughly 2,500 to 3,000 small details. Engineers must search for their own issue rather than receive an assigned task. That makes them watch for quality while building unrelated features, which reduces the number of small regressions they introduce.
The zero bug policy makes reported defects immediate work
Linear's zero bug policy assigns every reported bug to someone immediately, using agents to identify who worked in the affected area. The bug becomes that person's highest priority. They can decide that an extremely rare or difficult bug should not be fixed, but they must make that decision rather than place the report in a growing backlog. Artman says bug creation happens at a roughly constant rate as teams build features. Linear spent three weeks stopping new feature work and reducing its bugs to zero. Since then, bugs are fixed within seven days, often within two or three hours. He says reported-bug emails followed by a quick fix create a very different experience for users.
Current agents can execute design instructions without having taste
Artman says current agents have no taste. They can create an animation or interface element, but they do not understand how it feels to use the product. An agent can know that one second is faster than two seconds, yet it cannot feel whether a two-second click response is frustrating. It also lacks a human sense for natural animation timing. Artman describes a design engineer's comparison between agent-created animations and versions manually adjusted for feel. The agent completed the required actions, but its easing and timing felt unnatural. Artman expects tasteful AI to be a difficult remaining problem, especially for interfaces built for a particular product and user.
Linear builds product focus into hiring and daily work
Linear hires engineers who want to build software that is high quality and beautiful. Most engineers are product engineers, although the company also has technical work such as synchronization and infrastructure scaling. Each employee completes a paid, full-week trial in which they usually build a greenfield project, product, or feature. Linear uses the week to see whether a person can drive a product from start to finish. After joining, engineers can read open customer Slack channels, listen to recorded customer meetings, and search tagged discussions about particular functionality. Artman says this gives people constant exposure to customer needs. That matters because Linear now builds for larger companies and enterprises, not only for engineers using the product themselves.
Software engineers will need to understand customers and product choices
Artman expects everyone in engineering to become product-oriented as AI takes over more implementation work. He does not know whether AI progress will continue exponentially, but he expects engineers who only move data between systems to be less needed if it does. Engineers will still need to understand what customers want, what makes a good feature, and what makes a good user experience. He describes them as mini product managers who talk with customers and then implement useful functionality. His advice is to build something for yourself, ship it, and learn from the people who use it. He also recommends Apple's Human Interface Guidelines as a practical source for learning good UX.
