Tokens Per Second and why human reading speeds are about to become the bottleneck.
Tokens are roughly word fragments. TPS (Tokens Per Second) dictates how fast text streams onto your screen. The average human reads at roughly 4-5 words per second. Modern LLM APIs (like GPT-4o or Groq) frequently achieve 100 to 500 TPS.
At these speeds, the UI bottleneck is no longer the AI, but how fast the browser can render text, and how fast the human brain can parse it.