<aside> <img src="/icons/circle-dashed_purple.svg" alt="/icons/circle-dashed_purple.svg" width="40px" />

"DeepSeek built everything from scratch"

</aside>

Reality

• Explanation: DeepSeek allegedly used a process called model distillation, and I say allegedly because it’s hard to prove. We’re not diving into the technical details but basically distillation means DeepSeek took ChatGPT’s outputs and trained their models on those outputs. It’s not illegal but it clearly breaks OpenAI’s terms of service.
  • A bit of trivia: Although Microsoft and OpenAI are investigating whether or not DeepSeek obtained data from ChatGPT, Microsoft in the meantime has already added R1 to its own cloud offerings. I don’t know why but I just find that funny.
• Analogy: It's like a phone manufacturer studying how the iPhone's camera processes images, then training their own camera AI to produce similar results - not by copying the code, but by analyzing thousands of iPhone photos and teaching their system to match that quality.