Source
urlhttps://kwokchain.com/2020/01/22/how-to-eat-an-elephant-one-atomic-concept-at-a-time/
rawraw/highlights-eat-elephant.json

Kevin Kwok’s argument that the most consequential decision a product team ever makes isn’t features or pricing or even market — it’s the atomic concept, the primitive that everything else gets built on top of.

The argument

“The best products map to how customers think about their workflow.” Get the abstraction level right — not too high (unusable), not too low (hard to extend). Adobe’s primitives are layers, filters, artboards; they serve professional designers. Canva’s primitives are templates, drag-and-drop; they serve everyone else. Same market. Completely different products. The atomic concept did the work, not the feature list.

The corollary is sharper than most people realize: the wrong atomic concept is unfixable. You can’t bolt Canva’s atomic unit onto Photoshop without rebuilding Photoshop. The choice happens once, at origination, and it determines everything that follows. This is part of why counter-positioning is so brutal — the incumbent’s atomic concept becomes a structural liability when the world shifts underneath them.

Market transitions create openings. When customer needs shift, incumbent products stop fitting. eBay’s decentralized auction model was perfect in a scarce-supply internet. Once supply became abundant, Amazon’s centralized “price + speed” model won. The atomic concept that made eBay great became its handicap. That’s counter-positioning caused by an atomic-concept mismatch the incumbent literally cannot fix.

Changing needs are the startup’s tailwind. “Every once in a while, what was once a small use case grows into one large enough to support its own company.” The idea-maze-concept is partly about spotting these transitions before they’re obvious. Great atomic concepts crystallize how customers ought to think, not just how they currently do (simplicity-as-strategy).

The clearest modern example

ChatGPT vs. GPT-3 is the textbook case (chatgpt-pmf). Same model, two atomic concepts: GPT-3’s atomic unit was the prompt, which required engineering literacy. ChatGPT’s atomic unit was the conversation, which required English literacy. The conversation primitive unlocked a market that was 1000x larger than the prompt primitive. The model didn’t change. The atomic concept did.

Connections

chatgpt-pmf · counter-positioning · simplicity-as-strategy · idea-maze-concept · monopoly-vs-competition