Opinion

For a long time the comforting story about AI went like this: sure, it can crunch numbers and play chess, but it will never do the human things, the creative things, the intuitive things. That was the line, repeated confidently, for years. I want to gently point out that the line keeps moving, and it is moving in a direction that deserves more honesty than it usually gets.

The AI safety conversation is obsessed with the ceiling: the smartest model, the frontier, the hypothetical superintelligence. I think we are watching the wrong number. The change that will actually reshape the world this decade is not that the best model got smarter. It is that a good-enough model got almost free.

Agentic AI is the phrase every vendor deck and every board meeting is chasing this year. It is also, for most of the companies rushing to deploy it, a mistake they have not noticed yet. I say this as someone who thinks agents are genuinely useful. That is exactly why the current stampede worries me.

There was no dramatic moment, no single announcement, no headline that said it plainly. But somewhere in 2026, without a parade, open-weight models stopped being the scrappy underdog and became the default sensible choice for a huge slice of real work. It was a quiet war, and the open side won more of it than anyone expected.

I want to make a small, cranky request on behalf of clear thinking everywhere: stop calling everything an agent. The word has been stretched so far that it now means anything from a genuinely autonomous system to a chatbot that calls one API. When a word means everything, it means nothing, and the fuzziness is not an accident. It is marketing.

Every few weeks someone asks me if AI is a bubble, usually hoping I will pick a team. Yes or no, hype or real, tulips or telephones. I find the question a little boring, because the honest answer is that it is obviously both, and the useful conversation starts only after you accept that.

Read enough launch posts and you notice something strange: every new model is state of the art. All of them. Simultaneously. That cannot be true, and yet each chart is technically honest. Welcome to the benchmark wars, where nobody is exactly lying and almost everyone is being misled, gently, on purpose.

I am instinctively allergic to tech regulation written by people who have never shipped anything. A lot of the EU AI Act fits that description. It is late, convoluted, and parts of it will age badly. And yet, reading through what actually takes effect in 2026, I keep landing somewhere uncomfortable: most of it is the kind of thing the industry should have done on its own and did not.

There is a huge industry teaching people how to use AI: prompt courses, tool roundups, productivity threads without end. Almost nobody teaches the skill that is quietly becoming more valuable than any of them: knowing when not to use it. In 2026, the people doing the best work are not the ones using AI the most. They are the ones who know when to close the tab.

Every executive survey says AI budgets are going up. A quieter set of numbers says most of that money is not paying off. Only about a quarter of enterprise AI initiatives deliver the ROI they promised. That gap is the most interesting, and least discussed, story in corporate AI, because the reason for it is almost never the thing everyone blames.

Open almost any app you used happily a year ago and there is a new little sparkle icon in the corner, promising an AI assistant you did not ask for and probably will not use. The bolt-on chatbot is the defining product decision of the year, and I think it is quietly making a lot of software worse. I say this as someone who loves the underlying technology.

The most important AI story of the next year will not be a model release. It will be a court ruling. The lawsuits over what these models were trained on, the New York Times against OpenAI, Getty against Stability, are entering decisive phases, and the question they answer will quietly reshape the entire industry. I have a side, and I want to explain it honestly.

Every model launches with a splashy chart and a breathless thread. Almost nobody reads the boring document that ships alongside it, the model card, with its dull sections on training data, limitations, and known failure modes. I read them, every time, and I think it is one of the highest-value habits you can build in this field. Here is why.

There are enough AI blogs. Most of them read like a press release with the serial numbers filed off. This one is trying to be the thing I actually wanted to read: written by someone who uses these tools every day, likes them more than is probably healthy, and still reads the fine print.