The intuitions our species built over millennia, at the slow pace of generations, are coming apart at a dizzying speed. Our economic utility, our role in society, and our very meaning in the universe are all being challenged; and as our philosophical anchors are fracturing one by one, we drift into the unknown.

Today, AIs write, code, reason, plan, and increasingly act on our behalf; and while humans are still in the loop, it feels like our presence is fading. As AI is gaining in agency by the day, one has to wonder: can humans still control machines?

At the center of this upheaval lies a growing mismatch between human cognition and machine intelligence, AI capability is racing ahead faster than human biology can keep up.

This is what I call the bandwidth problem.

As AI becomes more capable, our ability to direct, understand, and govern it does not scale with it. We have opened Pandora’s box and we are attempting to steer what has emerged through keyboards, screens, and clicks: interfaces designed for an earlier age and unequal to the task now before them.

Unless we develop scalable ways to align humans and machines, human agency will steadily recede from the systems shaping the world. Either we align AI or we get left behind.

There are two ways to do that: alignment-by-control, which constrains the machine from the outside, and alignment-by-integration, which couples human values and intents to the machine directly. Control can hold for the next few years, but ultimately, only integration will last.

The brief interval in which humanity can still determine the terms of its integration with AI before it grows too powerful to steer is the Phase Lock Window. Once it shuts, the choice is gone and humans are left permanently out of the loop, with no say in a future run by machine intelligence.

This manifesto explores a breakout technology that offers us a path to alignment-by-integration:

Brain-computer interface.

BCI is one of the primary levers we have for aligning AI with humans, serving three simultaneous functions: it provides an architectural blueprint for an aligned intelligence (the brain), it supplies training signal so that AI learns human intents and values directly rather than inferring them from our cognitive artifacts, and it widens the bandwidth between mind and machine, and ultimately of the mind itself, so that humans can keep pace with AI.

Even though it remains widely dismissed, BCI is approaching the same inflection point AI crossed roughly a decade ago. The existence proof is already here: researchers can decode speech, reconstruct mental imagery, and detect intent from raw brain activity. Before the decade is over, thoughts will turn into AI prompts, imagined scenes will materialize on screens, and robots will respond directly to mental commands; and a few years later, that connection will become bidirectional, making it possible to write thoughts or emotions directly to one’s mind.

The question is no longer whether BCI can work, but how it will scale. The practical bet of this manifesto is that the same scaling laws that produced modern AI (better hardware, larger datasets and bigger models) apply just as well to brain decoding. By treating the brain as a vast, noisy, largely unlabeled dataset, AI can extract structure from signals that were previously too complex to decode; the ChatGPT moment of the brain is now around the corner.

The fastest path to execute this strategy will be through non-invasive BCI. Consumer devices can reach millions of people without surgery, generating the large and diverse neural datasets the field has been lacking. Consumer demand, paired with the growing value of neural data to frontier labs and robotics companies, will then finance further development, setting a flywheel in motion.

This manifesto maps that transition: how BCI will evolve over the next decade, the role it will play in the AI race, why neural data will become a critical asset for frontier labs, and how neurotechnology as a whole will help humans preserve agency and oversight over increasingly powerful machines.

BCI and AI are now on a direct collision course. AI will give BCI its scaling playbook, its models, and its capital; and in turn, BCI will give AI a new training modality and a new interface.

From this mutually reinforcing dynamic will emerge a new lever for controlling and aligning AI with humanity.

AI will scale BCI, and BCI will scale AI.

How we shape their convergence over the coming decade will define the centuries to come.