Why Context is More Important Than Code in the AI Era

We are obsessed with speed. Every week, a new language model, a new agent, or a framework emerges, promising a revolution. It seems that if we just integrate them into our workflow, we will start shipping products faster and more efficiently. But Karri Saarinen, co-founder of Linear, suggests we stop and ask an uncomfortable question. Do our products actually get better because we write code faster? Spoiler: no.

The volume of shipped software is not the product. Customers do not buy lines of code, the number of closed tickets, or impressive development velocity charts. They buy a solution to their problem. And in the pursuit of efficiency, we risk losing the very essence of product work.

The Trap of Outsourcing Thinking

Historically, companies solved the scale problem in one way. They hired more people, created highly specialized roles, and implemented rigid processes. This is how "software factories" were born. Today, we are simply replacing some of these people with AI agents, but we retain the same fatal flaw in thinking. We are outsourcing not only the routine but also the thinking process itself.

Here lies the main trap of our time. Product creation has always consisted of two inseparable parts: the artifact itself and the knowledge gained. When you puzzle over the architecture or argue about interface details, you are not just working. You are learning to understand the problem, the customer's taste, and the boundaries of technology. When you automate everything, you break the connection between action and learning. The team stops understanding why a solution works and loses that very intuition which distinguishes a good product from a great one.

Product intuition is not an innate gift or a mystical revelation. It is simply compressed experience. It is the brain's ability to instantly recognize patterns because you have seen the context in which they arise thousands of times. If AI takes away your routine, it gives you the scarcest resource—time. But the question is where you direct that time. If you spend it on generating an even greater volume of meaningless features, you will lose. If you direct it toward deepening context, you will gain an unlimited competitive advantage.

Practices for Preserving the Learning Cycle

How exactly can we preserve the learning cycle and avoid turning into a soulless assembly line? At Linear, they use specific practices that shift the focus from mechanical execution to quality and deep understanding.

• Context aggregation through AI. Instead of drowning in thousands of support tickets, Saarinen set up an agent that sends him a daily summary. This agent analyzes exactly what customers are saying about their AI workflows. This is not saving time for the sake of idleness. It is filtering the noise to focus on the core signal.
• Quality Wednesday. Once a week, every team member is required to find one micro-bug in the product—a weird animation, a typo, or a clunky hover—and fix it. The goal is not the fix itself. The goal is training the eye. When the team collectively discusses these findings, quality standards become tangible rather than abstract rules from a wiki.
• Feature Roast. Before launch, a new feature goes through collective, honest, and uncompromising criticism. It is a safe space where one can be skeptical and ask uncomfortable questions. If a feature confuses the internal team, it is guaranteed to confuse the user as well. It is a cheap way to get feedback before receiving it from the real market.

The Essence of Hiring and Leadership in the AI Era

This approach fundamentally changes the very essence of hiring and leadership. You do not hire a person so they can close tasks faster. You hire them for the trajectory they set for the company. Do they have taste? Are they capable of forming the right judgment? In an era when machines write code, the only thing left for a human is the ability to ask the right questions and possess a deep understanding of context.

Tools will only get smarter. Automation will swallow up more and more routine tasks, and that is wonderful. But your task as a product leader is not to build a factory for producing code. Your task is to create an environment where context is accumulated, transferred, and transformed into the flawless judgment of the entire team.

Code is just a side effect. Context is the product.