My observation is this: it usually takes a lot of intelligence, energy, and effort to build something great.

Great products hide the hard work

It took a lot of effort to build a computer. But using a computer should be simple.

It took a lot of effort to build the iPhone or iPad. But using them feels simple. That is the beauty of a great product.

AI products should follow the same principle

When we are designing the product, planning the architecture, coding, researching, or solving hard problems, we should use the best tools available.

Use the heavy reasoning models. Use the frontier models. Use expensive people, expensive thinking, and expensive technology where it actually matters.

Because building requires intelligence.

Once the product is built, the goal should be the opposite

The product should be so simple, structured, and well-designed that the cheapest model, the fastest model, or even the least intelligent model can use it successfully.

That is where the product beauty lies.

Invest intelligence upfront

The intelligence should be invested upfront into the design, the workflow, the guardrails, the state, the tools, and the user experience. Once that is done, using the product should feel simple.

So if the product only works when the smartest model is driving it, maybe the product is not finished yet.

The goal is not to depend forever on expensive models.

The goal is to use expensive intelligence to build something so clear that simple models can operate it.

View the original post on X ↗