Attention is all you need

Claude & Gemini have been trying to beat ChatGPT by competing on engineering. Even after succeeding for a short period, building better models did not translate to getting any market share.

DeepSeek on the other hand is a PR dark horse. Everything about it is so newsworthy. It's free and 99% cheaper! It was made in China It was a side- project. None of these terms are talking about bullshit engineering jargons. Instead of the 256hz 3gb storage, they're talking about 1,000 songs in your pocket. Which is why the normal people are paying attention.

In a world where engineering has become a commodity, PR turned out to be a differentiator. It's a lesson for startups to not only spend time on their engineering but also on their lore because — Attention is all you need