Human performance as a technology problem

The gap between knowing and doing

There is a particular kind of frustration that people who care about their performance know well. You are aware of what you should be doing. You have read the material, heard the advice, made the plan. And yet, consistently, something falls between intention and execution. Not laziness: the desire is genuinely there. Something else.

The common response is to treat this as a motivation problem. If people just wanted it more, they would close the gap. So the solution is more inspiration: better goals, better content, better pep talks. This framing has produced an enormous industry, books, coaches, retreats, apps, and a fairly consistent pattern of short-term improvement followed by a return to baseline.

What if the framing itself is wrong? What if the gap between knowing and doing is primarily a systems problem, not a character problem, and specifically a problem that good technology is well-positioned to solve?

What performance systems actually do

Performance at the individual level, physical, cognitive, professional, follows recognisable patterns. Progress comes from doing the right things consistently enough that they compound. Regression comes from feedback arriving too slowly, friction on the right behaviours being slightly too high, or the person losing an accurate sense of where they actually are.

None of those failure modes require a crisis of motivation. A person can be highly motivated and still drift: because the feedback loop is broken, because the environment makes the wrong behaviour slightly easier than the right one, because their self-assessment has quietly decoupled from reality. The problem is structural.

This matters enormously for how technology can help. A system that tries to manufacture motivation, to inspire, to hype, to generate excitement about your goals, is probably working on the wrong thing. A system that tightens feedback loops, reduces friction on the right behaviours, and gives someone an honest picture of where they actually are: that is working on the right thing.

Feedback: faster, more honest, more contextualised

The single most powerful lever in any performance system is feedback speed. The further in time a consequence is from a behaviour, the harder it is to build the connection. This is well established. It is also the reason that sophisticated training programmes in any domain obsess over measurement. Not because measurement is the goal, but because without it the loop stays broken.

The interesting thing about technology's role here is not just that it can make feedback faster. It can also make it more honest and more contextualised. These are different problems.

Honesty is hard because self-reporting is unreliable. Not through deception, but through the ordinary biases that accompany self-assessment. People tend to remember the good sessions more vividly than the mediocre ones, to attribute missed targets to external circumstances, to smooth over inconsistencies when constructing their own narrative. A system that keeps an accurate record and surfaces it faithfully, without editorialising, without softening, without inflation, is providing something that human coaches and conversation partners often cannot, simply because they are social and social environments pull towards encouragement.

Contextualisation is hard for a different reason. Feedback that says "you missed" is less useful than feedback that says "you missed, and here is what you were doing differently on the days you hit." Pattern recognition across a body of historical data, surfaced at the moment someone is trying to make a decision, is a different order of intervention. That is where software has real structural advantages.

Friction and the architecture of behaviour

The second lever is friction. There is a robust and largely uncontroversial body of evidence showing that small changes to the effort required for a behaviour, in either direction, produce large effects on whether it actually happens. Making the right behaviour marginally easier, and the wrong behaviour marginally harder, is not a trivial design concern. It is often the most important one.

In performance software this shows up in questions like: how long does it actually take to log a session? Does the logging happen close to the action, or does it require a separate intentional step hours later? Does the system require the person to compose or does it allow them to respond? Is the interface something they open with purpose or something they encounter passively?

These seem like product details. They are also where the majority of real-world performance impact is made or lost. A system that requires twenty taps to record a workout will be used less than one that requires three. A system that asks open-ended questions will get worse data than one that asks specific, answerable ones. A system that surfaces the thing to do next without requiring navigation removes a decision that, accumulated across weeks, represents genuine friction on the desired behaviour.

Designing performance systems well is, in large part, an exercise in architectural empathy: understanding not just what the person wants to achieve, but where specifically the environment makes the right thing harder than it needs to be, and whether a piece of software can remove that resistance.

The accuracy problem: where you think you are vs. where you are

The third lever is calibration. People who are performing well relative to their own standard are usually aware of it: the feedback is frequent enough, and the standard is clear enough, that their internal model stays accurate. The more common case is that the standard has blurred, or the feedback has been scarce, and the person is operating on an internal model that has quietly drifted from reality.

This is not a discipline failure. It is a measurement failure. And it is quietly responsible for a large proportion of plateau and regression in performance across domains. The person is not slacking. They are working hard against a miscalibrated baseline, and the honest picture of where they actually are, relative to where they were, relative to where they intended to be, is not available to them.

A performance system that gives someone an accurate mirror is genuinely valuable. Not a harsh or punitive one: accuracy is the point, not severity. But one that returns the person to their own standard rather than allowing a slow drift away from it. The question of how to do this without producing discouragement or anxiety is a real design problem, not a small one. But the alternative, protecting the person from accurate information, consistently produces worse outcomes over time.

The CheekyGains rationale

This is the territory CheekyGains is working in. The core premise is not that most people lack the willingness to perform. It is that the systems surrounding individual performance are poorly designed, and poorly designed systems produce poor outcomes regardless of the individual's intent.

The product questions that follow from this are specific: How does someone know, honestly, where their standard currently sits? How does the system deliver feedback that is both accurate and useful rather than merely discouraging? Where are the friction points in maintaining the behaviours that matter, and can software remove them? What does accountability look like when it is built into the environment rather than extracted through willpower?

These are design problems. They are also coaching problems. And they are, increasingly, AI problems: because the ability to personalise feedback, contextualise it against someone's own history, and deliver it at the right moment in the right register is now genuinely within reach in ways it was not a few years ago.

Naira, the AI coach inside CheekyGains, is premised on exactly this: that the value of a performance coach lies not just in knowledge but in availability and context. A coach who knows your history, who can see your patterns, who is present at the moment you are actually making decisions: that is a different kind of support than one you see weekly in a fixed session. The technology does not replace the human logic of good coaching. It extends the range at which that logic can operate.

Performance is an interface problem

There is a design principle that sits underneath all of this: performance software succeeds or fails at the interface. Not in the backend, not in the algorithm, not in the product vision. At the specific points where the system meets the person at the moment they are trying to do the thing.

This is humbling because it means that smart ideas about human behaviour are not enough. The insight that feedback speed matters does not help if the feedback mechanism is slow to load. The understanding that honest calibration improves outcomes does not help if the honest data is presented in a way that triggers defensiveness rather than adjustment. The architecture of a performance system has to hold the theory and the interface in alignment, and that is genuinely difficult.

It also means that the standard for a performance product is high and specific. Not: does this give people good advice? But: does it change what someone actually does, at the moment the decision is live? That is a harder question. It is also the right one.

The mistake is to treat human performance as a motivation problem and then build a motivation product. Motivation rises and falls. Systems endure. The more durable opportunity is to build software that makes the right thing easier, that gives the person an honest picture of where they are, and that delivers feedback at the speed and specificity required for the loop to stay closed.

That is a technology problem. And it is one worth taking seriously.