You have five seconds. Have fun.

You have five seconds. Have fun.

I use that sentence to imagine a product requirement. Something worth doing occurs to you: call a friend, follow a question, go outside, make something, keep a small promise. Can you turn that passing possibility into a chosen next step before another demand takes its place?

Five seconds is my design constraint, not a measured universal attention span or a biological deadline. Human attention changes with the task and context. A consequential decision can deserve hours, weeks, or another person’s help. The interesting problem is how easily a fleeting intention can disappear even when we would have wanted to keep it.

Three transitions, three different failures¶

I think about an attention–intention–action funnel. It is a practical framework for designing help, not a claim that the brain operates as three neat boxes in sequence.

At the first transition, something never reaches consideration. At the second, an attractive stimulus gets mistaken for a priority. At the third, I know what I want but do not begin or finish. These failures require different interventions.

Another notification might help me notice a genuinely urgent matter. It might also compete with the thing I had chosen to do. A motivational message might feel good while leaving an awkward interface, an unclear next step, or a real resource constraint untouched.

Research on implementation intentions offers a modest, useful example: specifying a cue and a response can help translate a goal into behavior.¹ A plan such as “after this meeting, I will write the first paragraph” gives an intention a place to land. It is not evidence that every failure of follow-through can be solved by wanting harder.

The market for a reactive mind¶

The five-second window sits inside an economy that can profit when attention is repeatedly redirected before an intention has time to mature. My concern is that we are building a highly reactive society whose reactions become increasingly easy to anticipate and sell against. There is less room to ask: what was I trying to do before somebody made this feel urgent?

Even an ignored interruption can have a cost. In one laboratory study, phone notifications disrupted performance on an attention-demanding task without participants interacting with the phone.⁷ That supports a concern about fragmented attention. It does not prove that everybody’s underlying attention span is permanently shrinking. The practical window can still close: another interruption arrives before I recover the thought I wanted to pursue.

When I say our biology is being “harvested,” I mean that reward, stress, emotional responses, and social comparison can become inputs to commercial optimization. Dopamine, cortisol, serotonin, and testosterone belong in that discussion, but their names are not an explanation by themselves. Dopamine participates in learning from reward and expectation.⁸ Cortisol operates within a broader stress system; experiments implicate its interaction with noradrenergic activity in shifts toward habitual responding.⁹ Serotonin can influence emotional responses during social decisions, while testosterone can affect status-related behavior in ways that include generosity as well as punishment.¹⁰¹¹ These are complex, interacting systems, with effects that depend on the person and context.

A platform does not need to measure those chemicals to learn which stimulus makes someone return, compare, worry, or click. Research on social-media posting supports reward learning as one contributor to engagement.¹² Separate randomized headline tests found that negative wording increased clicks.¹³ These findings establish pieces of the mechanism. My broader argument is about what happens when businesses repeatedly select the stimuli that produce the most valuable response, while the cost of recovering our attention falls on us.

From attention to recurring revenue¶

The financial ambition is predictability: turn a passing reaction into behavior that can be modeled, monetized, and repeated. Future cash flow is never guaranteed. But a business can try to make it more forecastable by learning what acquires a customer, what keeps that customer returning, and what prompts the next purchase.

Advertising makes the objective explicit. Google’s target-ROAS bidding predicts conversion value and adjusts bids toward a desired return on ad spend.¹⁴ That does not require fear or distraction. The conflict emerges when the most profitable route to a measurable conversion competes with the person’s opportunity to reflect. A click can count as success even when the person would have preferred to keep reading, finish a conversation, or leave the screen.

Prescriptions raise a different incentive question. Recurring treatment can produce recurring revenue, and subscription models can make continued use commercially valuable. A randomized study found that patients’ requests for advertised antidepressants affected prescribing, with implications for both undertreatment and overtreatment.¹⁵ That does not establish that attention platforms manufacture illness to supply pharmaceutical customers. Long-term medication can be necessary and life-improving. The question I want to ask is whether a business rewards appropriate care and better functioning, or treats continued refills and subscription retention as sufficient evidence of success. Those measures can diverge.

Safety can become another market for reassurance. The risk I see is a person encountering so many signals of danger that engaging with the world begins to feel reckless. More monitoring, protection, or another safety subscription can then feel like the only responsible response. Real dangers warrant real investment. The test should be whether that spending reduces actual risk and restores freedom to participate, rather than leaving the customer frightened and ready to buy the next layer of reassurance.

These incentives can reinforce one another without a shared plan: capture attention, intensify a concern, sell relief, and retain the customer. I am describing a possible economic pattern, not a demonstrated causal chain connecting a feed to a prescription or a security purchase. The social risk is still worth confronting. A society that spends more of its time reacting has less room to deliberate, explore, trust, and decide what it wants to build.

If our reactivity becomes someone else’s recurring revenue, reclaiming the interval between stimulus and response becomes an economic project as well as a personal one. The product opportunity is to make people more capable of engaging with the world, with less dependence on the system that helped them.

Build for completed intentions¶

Much of the digital economy is good at getting something in front of us. I am interested in a different product: help a person complete something they already endorse.

My own question emerged from noticing how many ambitions survive as conversation while ordinary actions remain unfinished. An elaborate recommendation can be another object demanding attention. The useful intervention might be much smaller: capture the thought, prepare the first reversible step, preserve the context, then get out of the way.

The business model matters. If the system earns more when I remain engaged, it has a reason to keep producing things to engage with. A product paid for useful assistance could compete on completed intentions and reduced dependence. That alignment still needs testing; a subscription does not automatically make a company benevolent.

I would want the metric to include whether I still endorse the outcome afterward. A machine can efficiently complete a task that was never worth doing. It can also turn a passing impulse into a commitment too quickly. Sometimes the most useful output is a reminder to reconsider.

Prediction is not permission¶

Neural interfaces make this distinction concrete. In research on the timing of intention, early motor-planning activity could drive an interface before the participant intended to initiate movement. Algorithm design mattered to bringing output into alignment with the person’s intended timing.²

That is a more interesting engineering problem than “read the mind faster.” A system may detect preparation, infer a likely command, or predict a preference. None of those observations is automatically authorization to act.

The distinction applies even without electrodes. An assistant might prepare an unsent message when I mention someone. Sending it changes the world for another person. Capturing an idea and making a commitment deserve different levels of confirmation.

I want prediction to reduce the cost of preparing useful help. I want control to remain with the person whose life will contain the consequences.

Where the frontier actually is¶

There are several frontiers, with different evidence and different reasons to exist.

Low-friction expression. A 2025 Nature paper describes wrist surface-electromyography models that decode gestures and handwriting and generalize across users.³ This is a way to express commands through muscle signals. It does not establish arbitrary thought reading. Its appeal is practical: input may become easier when a keyboard, camera, or spoken command is inconvenient.

Restored communication and control. A 2026 report follows one person with ALS using an implanted speech-and-cursor interface at home over nineteen months.⁴ The significance is sustained useful operation, including correction, fatigue, setup, and adaptation. A single participant cannot establish how broadly a system will work, but daily use asks a harder question than a compelling demonstration.

More natural combinations of actions. Research on simultaneous speech and gesture decoding shows why combining capabilities is itself a challenge.⁵ Training that worked for tasks performed separately did not automatically solve concurrent use. Ordinary life contains interruption and overlap. Interfaces need to survive both.

Feedback that changes performance. Closed-loop stimulation research has improved a specific cognitive-control task in an invasive clinical research setting.⁶ It suggests a possible intervention layer beyond decoding. It does not establish a consumer device that controls motivation, treats every attention problem, or tunes a person’s neurochemistry at will.

The temptation is to put all four under one exciting label. The useful approach is to ask which signal, which action, which person, and which demonstrated benefit justify each device.

Use the least complicated thing that works¶

An implant can be an extraordinary route to restoring lost function. It is a much larger intervention than most everyday interaction problems require. A button, an explicit voice command, a useful reminder, or a better sequence of screens may solve the problem first.

My investment interest is therefore in the reduction of a specific barrier to agency. Hardware novelty matters when it enables a benefit the simpler alternatives cannot deliver. The costs include fitting, calibration, support, false activation, correction, privacy, and the ability to stop using the product.

Proposed measure	Why I care
Chosen actions completed	Whether assistance becomes an outcome
Time and effort to a useful first step	Whether the interface removes friction
False activations and corrections	Whether speed creates new burdens
Goals still endorsed afterward	Whether the product respects judgment
Useful operation after the novelty fades	Whether the benefit persists
Ability to decline, revise, or leave	Whether the person remains in control

These are product questions, not validated clinical endpoints. A medical claim needs evidence appropriate to that claim. A personal productivity experiment can start smaller and record exactly what happened.

“Have fun” is part of the specification¶

I do not want a perfectly efficient system for becoming more obedient to an endless list. Fun, curiosity, friendship, and unplanned time deserve a place in the objective.

An intention can be to make a lamp, spend time with a friend, or walk without turning the walk into a performance metric. Choosing not to act can also be an exercise of agency. A product that cannot distinguish these from failure is optimizing a narrower life than I want.

My bet is that the next valuable interface will be judged partly by what people are able to do after putting it down. Can it help me keep a promise, pursue a thought, or leave the screen and enjoy an afternoon? That is a future worth building toward, five seconds at a time.