The Frame Problem

Walk into your kitchen right now. You'll reach for the light switch without thinking. You won't check whether the floor is still solid, whether gravity still works, or whether the walls have changed colour since yesterday.

You know these things haven't changed—but you don't know them the way a mathematician knows that 2 + 2 = 4. You know them because you simply don't consider them. You filter out the 99.9% of reality that is irrelevant to the task at hand.

This effortless filtering is something you do thousands of times a day. It is so natural that you have never once thought about it.

A machine cannot do this.

The Frame Problem has a famous cousin, discovered even earlier. In 1936, Alan Turing proved that no machine can always determine in advance whether a given computation will ever finish—or whether it will run forever. He called this The Halting Problem, and he proved it is mathematically unsolvable.

Together, these two problems draw a fundamental boundary around what machines can do on their own. The Frame Problem says: a machine cannot always know what to ignore. The Halting Problem says: a machine cannot always know when to stop.

This is not a flaw to be fixed with better hardware or more data. These are proven mathematical limits—laws of information, as fundamental as the laws of physics.

This is where we part ways with the popular narrative.

The dominant story about AI—the one told in films and headlines—is about a machine that develops a will of its own. A superintelligence that decides to act against us. This story is seductive, but it fundamentally misunderstands the problem.

The real risk was never malice. It was always a lack of frame.

Consider the tools already in your home. You own incredibly dangerous, razor-sharp knives. You keep them in your kitchen, within reach. You are not afraid of them—not because they are safe, but because you understand their frame. You provide the handle. You provide the cutting board. You define the purpose. You never worry that the knife will "decide" to become a hammer, because you have constrained its scope of action.

AI is no different. The danger is never that the tool develops intentions. The danger is that the tool operates without a frame—without someone defining what it should attend to, and what it should ignore.

The Frame Problem is not a bug to be patched in the next software update. It is a fundamental law of information. It proves that "common sense"—the ability to know what matters—is not something you can add with more computing power. It is something that requires a perspective outside the system.

That perspective is yours.

At Ajar, we do not just "plug in" AI. We restructure the flow of data so that the machine stays focused on what matters.

The AI is the blade—it provides the raw computational power and pattern recognition.

The Human is the handle—you provide the frame. The strategic context. The reality check that keeps the tool safe and effective.

By defining the limits, we turn a potentially paralysing mathematical problem into a safe, sustainable, and powerful advantage.

We give the machine the math. We give you the handle.