This is a founder writing a philosophy-of-craft essay in the form of a blog post, and the form matters. Karri Saarinen (Linear's co-founder and CEO) isn't announcing a feature or selling a product here — he's staking out a position on how software should be made, and more pointedly, on what's being lost as AI agents start generating changes independently across codebases. The core argument is compact: good products carry the perspective of their makers. Two teams with identical inputs will build different things because they understand the problem differently. That understanding develops through experience, customer conversations, and accumulated judgment — what Linear calls "intuition over process or data." Saarinen is careful to define intuition not as gut feeling but as compressed expertise: the residue of paying attention over time. The sharpest section targets what happens when individuals work with agents to produce changes independently, "each operating a small software factory." Each change might pass its own review, but without shared perspective the product becomes an average of its outputs. Saarinen's metaphor — Frankenstein's monster versus a carefully shaped sculpture — is blunt and effective. The additive logic of capability-stacking replaces the subtractive discipline of deciding what a product should not do. This is where the essay earns its keep. The observation that "many products are starting to feel messy" — including from successful companies — is doing real diagnostic work. Saarinen identifies a specific failure mode: products that are collections of abilities without a unified point of view. The technology can do more, so more gets added, and the core problem erodes under the weight of features nobody shaped into coherence. What the essay doesn't do is offer a structural solution. The prescription is essentially "develop a perspective and follow it further," which is true but operates at the level of founder exhortation, not organizational design. How do you maintain shared perspective at 50 engineers? At 200? When half the code is agent-generated? These are the harder questions, and the essay gestures at them without engaging. The writing itself is clean and unpretentious — no jargon, no frameworks, no numbered lists. It reads like someone thinking out loud with confidence, which is the hardest register to hit in founder writing. The argument flows from observation to principle to concern to hope in a single unbroken arc. If you've used Linear, you can feel this philosophy in the product's tight, opinionated interface. The essay's real audience isn't Linear's customers — it's other founders and product leads who feel the same unease about agent-assisted development but haven't articulated it. Saarinen is naming something the industry is experiencing but hasn't framed clearly: the risk that AI doesn't just accelerate software production but dissolves the shared taste that held products together.
