If you place AI on humanity's historical timeline, we are somewhere deep in the stone age. That might sound strange for the month when Claude Fable was released. But that is exactly the point. We are building tools of stone, and nobody knows what comes after.
The stones
The analogy holds longer than I expected. Everything is carved from the same stone, the transformer models. First we started saving the best stones so we could reuse them. That is saved prompts. Then we shaped the stone into something that looks like a real tool. That is skills. Now we already have a whole arsenal, everything from bows to spears. Chatbots, agents, image generators.
The stone age was in many ways the most important phase in human development. Starting to use stones and tools is often counted as the first step toward everything that came after. But the next step needed some special combination that nobody could see in advance. What comes after you have built every tool that can be built from stone? I don't think anyone knows yet what the AI bronze age looks like.
The questions nobody has time to ask
In the middle of a gold rush, some questions never get asked at all. Vicki Boykis wrote about local models in June and asked, roughly: if performance and price actually constrain us, what architectural tradeoffs do we need to make? (Boykis 2026) It is a question almost nobody has asked during the token gold rush. Everyone is striking stones as fast as they can. Nobody asks which tools we actually need.
The bronze age
Around the 2010s, companies started to seriously use their technology and market position to make money. That would explain why many people feel that technology after 2015 improved more slowly, and why everything filled up with ads and microtransactions.
Mirror that onto AI. What does the world look like when the AI companies start using their position the same way? How do they operate in ten years?
I don't know what comes after the stone age.
— Niklas