When Thinking Matters More Than Coding: What AI Really Changes

A forecast from Dario Amodei, founder of Anthropic, the company behind Claude. The very text-based neural network that today doesn’t just “answer questions,” but assembles agents, writes applications, and completes tasks that previously took weeks of work by developer teams.

If we set aside the marketing and speak honestly, we are already at a point where programming is no longer a niche craft for the initiated. At Anthropic, this is felt from the inside. Amodei himself says that some engineers openly state: “I don’t code at all anymore. I formulate the task, the model writes the code, and I review it, edit it, and handle everything else.”

And these are not interns. These are engineers who until recently built architectures themselves and debugged overnight.

According to him, we are about 6–12 months away from the point when AI can perform most of a programmer’s work completely from start to finish. Not just isolated functions or pieces of logic, but the entire cycle: from idea and structure to implementation, testing, and revisions. And then a truly uncomfortable question arises — how quickly will this cycle close? How much time will pass between “I need an idea” and “the product is ready and works”?

Judging by current dynamics, the answer will be uncomfortable for the old system. Very little.

In fact, even now, AI can create almost any digital product: a website, a service, an app, an internal tool, or a startup prototype. Those who for years postponed ideas saying “I’m not a programmer” or “I need an expensive developer” can officially cross out that excuse. The time has truly come.

The key point here is that you no longer need to be able to code or deeply understand technologies. That is no longer an entry ticket. The only thing required is the ability to think and clearly understand what you want to achieve. Formulate the task, see the goal, set constraints, understand for whom and why you are doing it.

AI executes excellently. But there is one fundamental point that many forget. AI can do almost everything except one thing — it does not make decisions for you.

This is where the new dividing line emerges.

A very accurate metaphor is a cook versus a head chef. AI is the perfect cook. It chops, fries, boils, repeats a recipe flawlessly, never tires, and never argues. It can cook a thousand dishes in a row without losing quality.

But who decides which dish to make? For which client? At what price? In which restaurant and for which audience? That’s the head chef’s decision.

The same applies to AI. It can implement almost any idea. But it will not tell you which idea is worth doing and which is not. It will not take responsibility for strategy, positioning, risks, or consequences.

That’s why today, winners are not those learning to “write prompts,” but those learning to properly set tasks for neural networks and make decisions. To see the system as a whole, not just individual buttons and lines of code.

This skill is already highly paid for today. And it will become key in the coming years. Code is depreciating. Thinking is appreciating.