Most people interacting with AI are still picking a model and hoping they picked right. Perplexity Computer launched in February and took a different approach. It runs on 19 AI models simultaneously, including Claude, GPT, Gemini, and Grok, and routes your task automatically based on what the task actually requires. You don't choose the tool. You describe the problem. The system decides which model handles it best and assembles the result.

That matters more for adoption than it might seem. A big reason AI is hard to pick up isn't the technology itself. It's the overhead: which model do I use, for what, when does it fail? Perplexity Computer removes that layer. The parallel here is what happened when Anthropic moved from Claude Code, a developer CLI tool, to Claude Cowork, a desktop interface built for non-developers. The capability didn't change. The surface did. Perplexity Computer is doing something similar: taking a complicated routing decision and making it invisible. At $200 a month, it's priced for professionals and businesses who want the full range of current AI without the maintenance cost of figuring out which model to deploy for each task.

Most people writing about AI pick a lane. They're optimists or they're doomsayers, and they stay there. Dario Amodei, CEO of Anthropic, went on 60 Minutes to say AI might compress a decade of scientific progress into a few years. That is not a careful statement. That is one of the most sweeping claims a tech CEO has made in public, attached to a name that carries real credibility in the field.

What makes Amodei worth watching is that he wrote both essays. In 2024, he published Machines of Loving Grace, an optimistic case that AI could help defeat most disease and lift billions out of poverty within a decade. Eighteen months later, he published The Adolescence of Technology, and the register had shifted. Not a reversal. More like a researcher returning to a hypothesis after seeing new data. The second essay grapples seriously with how the same capabilities that could accelerate medicine could concentrate power in dangerous ways. He's not catastrophizing. He's updating.

That 18-month gap is the real signal. The first essay was written by someone confident in the trajectory. The second was written by someone who had watched the trajectory accelerate faster than expected, and started asking harder questions about where it ends. Both are worth reading, in order.