Linear's Karri Saarinen: design-in-code tools risk killing conceptual exploration — the real source of innovation

Summary

Karri Saarinen, co-founder and CEO of Linear, argues that the wave of design-in-code tools risks collapsing two fundamentally distinct phases of the design process into one, and that the industry is largely missing this distinction.

Saarinen frames design as a search problem. The first phase is conceptual exploration, where designers should be free to pursue bad ideas, wrong directions, and abstract possibilities without the weight of implementation constraints. The second phase is execution design, where code-native tools genuinely add value by enabling faster iteration, edge case resolution, and refinement against real user feedback. His concern is that conflating the two phases pushes designers toward conservatism by anchoring them to the existing codebase, which biases toward solutions that "fit" rather than solutions that are best.

The analogy is architectural. Architects sketch before they price materials. Concept cars exist because the industry understands that aspirational thinking and production thinking are different cognitive modes. Saarinen's contention is that software has the same need but is culturally less disciplined about enforcing that separation.

The AI dimension compounds the risk. When designers describe desired outputs to an LLM rather than sketching or exploring, they introduce a statistically-driven intermediary that trends toward average, common patterns rather than contextually optimal or distinctive solutions. Saarinen describes this as a convergence risk: products start to look alike because the generative layer optimizes for statistical correctness, not creative specificity. He acknowledges AI can legitimately accelerate ideation by generating many variations quickly, but the trade-off must be understood consciously.

The incentive structure in technology makes this harder to resist. Shipping velocity is measurable; design quality is not. That asymmetry creates organizational pressure to treat design as a construction activity, which is precisely the framing Saarinen is pushing back on. His stated motivation is not tool-agnostic skepticism for its own sake but a belief, core to Linear's brand, that software quality is worth protecting even when it cannot be quantified.

On brand and storytelling, Saarinen treats them as the same thing. Brand is a belief system and a set of values; storytelling is how those beliefs are externalized. His critique of companies searching for "storytellers" is that the hire is meaningless without a genuine underlying story. A skilled writer is useful only if the organization already believes something worth expressing.

Linear itself declines to add seasonal UI elements, including holiday decorations, on brand consistency grounds. Saarinen notes the practical failure mode directly: a customer whose infrastructure is down during a crisis does not want to see a Santa hat in their project management tool. Linear does embed Easter eggs, including hidden arcade games, but keeps them deliberately obscure, requiring active discovery rather than passive exposure.