Not long ago, I wrote here on the Bett Brasil blog about cyberdecks and artificial intelligence gadgets, noting how AI was beginning to move beyond screens and take on physical forms and new ways of being present in our daily lives. Now, the Codex Micro, created by OpenAI in partnership with Work Louder, embodies a possible next step: keys designed to physically interact with AI agents. Rather than predicting which device will be the future, what matters for education is recognizing these changes and fostering a range of skills and curiosity before they become a natural part of everyday life.

A keyboard that wasn't made just for typing

The Codex Micro is small, but conceptually unusual. We have before us something recognizable as a keyboard, yet some of its keys do not represent letters, numbers, or traditional shortcuts. They may represent agents. The RGB lighting indicates whether these agents are thinking, executing, waiting for an intervention, or have completed a task. There are physical controls for accepting and rejecting actions, triggering workflows, communicating with the system, and modifying interaction parameters.

This changes the meaning of the peripheral device. On a traditional keyboard, we press a key to perform an action. On the Codex Micro, we can initiate a process that will continue even after we remove our finger. The key no longer represents only what we do on the machine; it can come to represent what a machine is doing for us. It’s a small difference in appearance but a huge one in the human-computer relationship.

For nearly 150 years, we've been pressing practically the same keys

Perhaps the Codex Micro becomes more interesting when placed next to a typewriter. The QWERTY layout emerged in the 19th century and has outlived the very technology that gave rise to it. We’ve gone through the mechanical typewriter, terminals, personal computers, the internet, laptops, smartphones, cloud computing, and artificial intelligence, all while maintaining a surprisingly stable physical layout.

The mouse has also endured for decades, preserving the fundamental logic of pointing, selecting, and executing. Touchscreens have changed the gesture; graphical interfaces have changed the representation; but we continue, to a large extent, to operate machines.

AI makes a difference because the relationship begins to shift from operation to intention.

• Previously, we would select a tool and execute a sequence;
• With generative AI, we can now describe what we want;
• With agents, we can delegate entire workflows;
• With interfaces like Codex Micro, we’ve started to need to physically keep track of who is doing what, when we should step in, and how much we want to delegate.

The point here is to realize that, for the first time in a long time, the nature of what lies behind the key is changing.

The OpenAI keyboard is more interesting as a symbol than as a product

We don’t know if, ten years from now, we’ll still be using keyboards full of keys for agents. Perhaps the Codex Micro will be remembered as a pioneer. Perhaps as a curiosity. Perhaps this category of peripheral will simply disappear because voice, computer vision, wearables, or interfaces that don’t yet exist will solve the same problem better.

Talking about the future shouldn't mean making spectacular predictions. It means observing changes that are still small and asking what larger system they reveal. In that sense, the Codex Micro is almost a fossil of the future, discovered before the future even happens.

He shows that we are already trying to physically solve problems that have arisen with agent-based AI: how to visualize an agent at work, how to call on another agent, how to authorize an action, how to interrupt, how to speak, how to indicate different states, and how to divide our attention between what we do and what we delegate.

For decades, the personal computer was designed around the user running software. We are beginning to experience a form of computing in which the user also coordinates intelligences. This shift may be far more significant than the device that currently represents it.

See also:

• "Purple August" and the Cognitive Delta: How much of a student's potential does the school actually tap into?
• The issue isn't whether to use AI in research. It's knowing how to use it.
• The next educational inequality will be inequality in the questions asked

What if our students worked less as operators and more as coordinators?

This is where the small keyboard really starts to become relevant to education. We’re still debating how to teach students to use artificial intelligence, craft prompts, verify information, recognize hallucinations, and understand ethical boundaries. We need to keep doing that. But the Codex Micro suggests that perhaps we’re simultaneously preparing students for a phase that’s already beginning to change.

A student who works surrounded by law enforcement officers will need to do more than just know how to carry out procedures. They will need to develop the ability to:

• Formulate problems well enough so that they can be distributed;
• Decide what should remain under human control and what can be delegated;
• Track different processes simultaneously;
• Recognizing when a seemingly correct solution is wrong;
• Interrupt, correct, and redirect agents;
• Compare alternatives generated by different systems;
• Understand the consequences of automated decisions;
• Take responsibility for the outcome even when you did not personally carry out every step.

This even changes what we mean by competence. If a person can press a key and have an agent perform a task—such as programming, researching, organizing, or analyzing—knowing how to press the key is not the relevant skill. Competence lies in what happens before, during, and after that action: understanding the problem, establishing criteria, supervising the process, and evaluating the result.

For schools, this change is profound because, traditionally, a key part of learning has been the actual performance of tasks. When that performance can be delegated, we need to know much more precisely which human skill a given activity is intended to develop.

Show the door before it looks like a wall

There’s a reason I like Morpheus so much as an archetype when I think about education and the future. His role in *The Matrix* isn’t simply to know technologies that Neo is unaware of. It’s to make him realize that what seemed natural and definitive was just one possible configuration of reality. He’s the archetype of the mentor on the heroic journey and, specifically in this case, the protagonist’s struggle and the world torn between technology and humanity.

Perhaps we need to do something similar with our students. A student looks at a keyboard and sees a keyboard. We can help them see a 19th-century technology that has spanned different computing paradigms.

He looks at a key on the Codex Micro and sees another button. We can help him realize that that key represents a much bigger shift: we can now assign work to artificial intelligence, and we need to build interfaces to support that relationship. And then come some questions that are truly relevant to education:

• What will come after the keyboard and mouse?
• How will AI transform the physical devices we use to interact with computers?
• What new peripherals will emerge when we move beyond simply controlling machines and begin coordinating agents?
• Will the keyboard continue to be organized around letters, or will it also come to represent agents, tasks, and states?
• How might voice, gestures, sensors, and wearable devices complement or replace current interfaces?
• What skills will be needed to understand, configure, and create these new physical interfaces?
• Are we preparing students to operate today's hardware, or to imagine and build the devices of tomorrow?

Educating for the future doesn't mean putting the latest gadget on every student's desk. Nor does it mean turning teachers into futurists who try to predict which technology will dominate the next decade.

It means training people who are capable of looking at something like the Codex Micro and asking:

• Why did they do it that way?
• What changed to make this object make sense?
• What problem is he trying to solve?
• What new problems does this create?
• And how would I do it differently?

A few years ago, AI gadgets were signs that artificial intelligence was beginning to seek a presence beyond the screen. Now OpenAI is introducing a keyboard in which agents have their own keys, states, and controls. And in education, perhaps one of our responsibilities is precisely this: to recognize where certain signs are pointing and to bring our “Neos” close enough to the door so that they’ll be curious enough to discover—or build—what lies beyond it. 

About the author: