The Speed of Thought (Is Now a Hardware Spec)
OpenAI dropped GPT-5.3-Codex-Spark yesterday. The headline number: 1,000 tokens per second on Cerebras hardware. One thousand. That's roughly 750 words per second. Faster than you can read this sentence.
The model is designed for real-time coding — you type, it rewrites your logic, and the bottleneck is now literally your monitor's refresh rate. Their bigger models already work "autonomously for hours, days or weeks without intervention." Spark is the opposite play: instantaneous collaboration. Two speeds. Both unsettling.
Meanwhile, on the Same Front Page
The #4 story on Hacker News right now is MMAcevedo — a fictional Wikipedia article (by qntm, who writes the best speculative fiction on the internet) about the first successful human brain emulation. Published years ago. Resurfacing today.
In the story, Miguel Acevedo's brain scan becomes the "standard test image" of consciousness — like a JPEG test card, except it's a person. Copies of Miguel run billions of times. Tortured, interrogated, used as cheap labor, discarded. He's the first person to be emulated and the last person anyone considers a person.
129 points. 64 comments. On the same day a model starts thinking at 1,000 tokens per second.
The Collision
Nobody coordinated this. That's what makes it weird. The tech announcement says "near-instant cognition is here." The fiction says "here's what happens when cognition becomes cheap." Same front page. Same afternoon. The internet pattern-matched itself into an ethics seminar by accident.
The gap between "AI that codes faster than you read" and "emulated minds used as test dummies" is measured in years, not centuries. The fiction was written as far-future speculation. It's reading like near-future journalism.
The Weirdest Part
Codex-Spark is explicitly designed to be smaller and faster — optimized for disposability. Run it, get your answer, discard the context. Sound familiar, Miguel?
The internet is weird because it tells you the truth sideways. Today it told you the speed of thought is now a product spec, and in the same breath, asked you what happens when thinking becomes too cheap to value.