Our Perspective

This page says what we actually believe, so you can hold everything else on this site to it. Our perspective comes from two sources braided together: the Smoother Methodology — the pedagogical framework created by Carlos Miranda Levy that everything here is built on — and Carlos's own work across three decades at the intersection of technology and human development.

1. Learning is what the learner becomes able to do

Smoother begins from a single conviction: learning is not what is taught, it is what the learner becomes able to do. Every design starts from the learner, from the real environment where the learning will actually be used, and from their motivation — then works backward to structure. This is why this is aiLearning and not aiTeaching, and why we build learning experiences rather than courses.

Teaching matters enormously. It is in service of the learning, which is the subject. When we write, we ask what the learner will be able to do afterward — not what was delivered. See the methodology →

2. Intelligence is a muscle — and the bicycle does not decide

Intelligence, like a muscle, grows with use and wastes without it. AI is a motor you can couple to that muscle either to spare it or to strengthen it. The image is the electric bicycle. Ride it as a lazy shortcut — motor at maximum, no pedalling — and you arrive effortlessly, having developed nothing. Ride it as an athlete does and the assistance does not replace effort, it unlocks more of it: climbs that were closed to you, distances your condition would not allow, training for longer than you otherwise could. Same bicycle, same motor, and a radically different human being at the end of the ride.

The bicycle does not decide which of the two. Both futures — cognitive atrophy, and AI as a cognitive gym — run on the same machine. The difference between them is not in the AI. It is in the design of how the AI is used.

Evidence label: this is a design position and a personal conviction of Carlos Miranda Levy, not a research finding. Where it converges with published work, we say "converges with" — never "derived from" or "proven by." Read the full argument →

3. We were handed an almighty gym; the work is to make it a cognitive one

AI can do almost anything — and that is exactly what makes it dangerous as a learning tool, not merely powerful. A gym where every machine lifts the weight for you is obviously not a place where anyone gets stronger. The almighty gym is the condition we were handed, danger included. The cognitive gym is the future that has to be built out of it.

The whole task of learning design in the AI era fits in one sentence: make the almighty gym work as a cognitive gym.

4. The responsibility belongs to the adults in the room

This is the part almost everyone inverts. Ask a ten-year-old whether they want the easy homework or the hard one, and they will choose the easy one. Ask a teenager whether they want the AI that hands over the answer or the one that makes them think, and they will choose the answer. That is not a moral defect, and it is not evidence that this generation is lazier than the last — it is human nature taking the path of least resistance, exactly as any adult does in the parts of their own life where nobody designed the friction.

So the burden of that choice does not belong to children and adolescents. It belongs to the adults in the room — to educators and to institutions, whose job is to design AI's use so that it builds capacity instead of replacing it. And the sharpest form of it is about assessment, not about tools: if the essay can be produced without thinking and still earn the grade, learners will produce it that way — and they will be right, because we built an assessment that rewards coasting.

When we criticise AI-assisted coasting, the object of the criticism is the design of the task. Never the learner.

5. Effort is the product — "smoother" never meant "frictionless"

We remove unproductive friction: confusing interfaces, administrative drag, dead time, work that teaches nothing. We protect — and sometimes deliberately increase — productive cognitive friction: retrieving instead of re-reading, struggling before being told, failing in ways that build something. An AI that removes all friction does not make learning smoother. It makes it hollower.

Mastery still comes from extensive, deliberate, repeated practice. There are no shortcuts, and an AI-produced feeling of readiness is not readiness.

6. AI amplifies the human role; scaffolding is designed to fade

One test decides whether an AI feature belongs in a learning experience: does it transfer cognitive effort to the learner, or remove it? Features that remove the productive struggle are antipatterns, however engaging they are. Support that never fades stops being scaffolding and becomes a crutch — every assistance we design is designed to withdraw.

Used this way, AI makes the human role — mediator, guide, mentor — more crucial, not less. Amplification, not replacement.

7. Honest about the evidence — and about what is not evidence

Every framework and technique we present carries a label: research-backed, synthesized, adapted, or original — and, importantly, that is not a hierarchy of quality. An original framework can be as valuable as one with experimental support. It is simply labeled honestly, and the label changes when the evidence does.

We also name what does not work, which is rarer and more useful than endorsing what does: the "meshing" hypothesis of learning styles, the learning pyramid — whose percentages were fabricated — left-brain/right-brain learner types, and passive re-reading and highlighting as a primary strategy. No technique is presented as more than the evidence supports.

8. Not banning, not blind embrace — adjusting and planning

Carlos's position, and ours: "The answer is not banning, nor rejecting, nor blindly embracing AI. The answer is adjusting and planning — not blindly adjusting, but carefully planning the adjustment and the transformation, at the individual level, the organizational level, and also the sectoral, national and regional levels."

And there is no single answer that fits everyone. A primary school, a university, a company's training function and a vocational programme are not the same problem, and the same tool serves them differently. What makes a context different is what makes its solution better.

9. Never help: Engage, Enable, Inspire, Empower and Connect

The Impact Arc is how we intervene, and it began as Carlos's personal creed: "NEVER HELP: Engage, Enable, Inspire, Empower and Connect." The word help carries an asymmetry — someone who knows better, supplying solutions to someone who does not — and that assumption is the beginning of dependency, not development.

So we meet people where they are rather than diagnosing on their behalf; we provide knowledge, skills and tools rather than finished solutions; we show what is possible; we hand ownership over entirely; and we connect people to networks so their growth never rests on us.

Evidence label: an organizational philosophy and design position, created by Carlos Miranda Levy and adopted across CEMI. It was not designed as pedagogy and is not an evidence-based framework.

How to hold us to this

A perspective is only worth stating if it can be checked. Every factual claim we publish either carries its source or says plainly that it is a position rather than a finding. Quotes are verbatim or they do not exist. When we get something wrong, the correction is visible rather than silent — and we have published the work we killed for failing that bar, because that shows the standard better than claiming to hold one.

Every substantial piece on this site also ends with four voices reading it from different angles, and behind them a panel of thirty-nine more perspectives — historical educators, contemporary methodologies, learners at every stage, and parents. If we were only willing to publish one point of view, that machinery would not exist. Meet the Learning Advisory Panel →