On this day, 2024, Anthropic released Claude 3.5 Sonnet and quietly shipped a small feature alongside it called Artifacts.
Back then, it felt like you were already behind. Every week brought a new model, a new leaderboard, a new “this changes everything” headline. If you ran a business and had actual work to do, the honest reaction was less excitement and more quiet dread: how am I supposed to keep up with all of this. The experts looked confident, but most of them were guessing too.
Here is what actually happened, in plain English. Claude 3.5 Sonnet was a smarter, faster, cheaper version of Anthropic’s AI. That part was normal progress. The interesting bit was Artifacts. Until then, when you asked an AI to build something, a document, a chart, a little tool, it all came back buried inside a wall of chat text. Artifacts gave that work its own window, sitting right next to the conversation, updating live as you talked. It turned the AI from something you read into something you could watch build.
So what happened between that day and now? Two things, almost exactly a year apart. The live side-by-side canvas stopped being a novelty and became the way people expect to build with AI. And in June 2025, Anthropic won a court ruling that training its model on books counted as fair use. Notice the shift. In 2024 the question was “can it do something impressive.” By 2025 the question had moved to “is the way we built it even allowed.” Both felt enormous in the moment. Today, in 2026, both are just background hum. Building with AI in a live window is ordinary, and the legal questions are working their way through courts the way every new technology’s questions eventually do.
The takeaway is not which tool to buy. It is this: you did not need to track every model drop to understand what mattered. The few things that actually counted, you can grasp in five minutes, in hindsight, with no panic at all.
Yesterday’s breakthrough is today’s normal.
