The Frontier and the Rest of Us

We have spent much of the current AI conversation looking upward, asking how intelligent the models can become, how far the frontier can move, whether artificial intelligence can discover new science, design better systems, write better software, reason across disciplines or eventually outperform the best humans in areas we once believed belonged almost exclusively to human expertise. These are important questions, and they are part of what led me elsewhere to another possibility: that the frontier may not exist entirely inside the model, but somewhere in the interaction between a highly capable model and a human capable of bringing the right insight, judgment, context or question into that interaction. Yet many of the examples that make this possibility interesting still involve brilliant people, scientists, researchers, engineers and builders who already know enough to recognise what matters, push the model in useful directions and understand when something unusual has appeared, which leaves another question underneath the entire discussion: what happens when frontier intelligence meets someone entirely ordinary?

Most people who eventually use increasingly capable AI systems will not be scientists trying to cure disease, engineers designing new architectures or entrepreneurs attempting to build the next major company. They will simply be people trying to understand something, solve something, save time, make better decisions or get more from knowledge they already possess. Think about the warehouse worker who knows exactly why the same part of the morning shift always becomes chaotic, the shop owner who notices that certain products disappear faster toward the end of every month but has never thought about demand forecasting, the mechanic who has spent twenty years recognising sounds and behaviours in engines without ever formalising that knowledge, or the teacher who knows that some students begin to struggle long before their grades reveal it. None of these observations need to be brilliant in the way we normally use the word; they may simply be ordinary pieces of knowledge accumulated from being somewhere, doing something and paying attention for long enough, and that brings us to the question at the centre of this paper: what can frontier intelligence do with material that appears ordinary?

This changes the way we should think about access to AI because much of the conversation still assumes that people must rise toward the technology: learn prompting, learn coding, understand agents, models and APIs, then become increasingly technical so they can use increasingly powerful technical systems. There will always be value in understanding your tools, but mature technologies usually move in the opposite direction because successful technology gradually absorbs its own complexity. You do not need to understand telecommunications engineering before making a phone call, network protocols before sending a message or the internal architecture of a search engine before asking where the nearest restaurant is, so why should a farmer need to understand machine learning before intelligence can help connect weather, soil, pricing and crop decisions, or a small business owner study software architecture before describing a workflow that wastes three hours every week? The more interesting future may not be one where everybody learns the machinery of AI, but one where the machinery becomes capable enough to understand what people already know, what they are trying to achieve and where they repeatedly get stuck.

This is where the common man becomes far more interesting than the phrase initially suggests, because ordinary people are not empty vessels waiting for intelligence to arrive. They carry context, experience, habits, frustrations, local knowledge and small pieces of expertise that often never become formal knowledge because there has historically been no practical machinery for turning them into systems. Much of what people know never becomes a report, dataset, academic paper, software product or even a written note; it remains inside routines and observations such as, “we always do this first because something goes wrong later if we don't,” “customers normally ask for this after that happens,” or “this machine starts sounding different before it fails.” The person may not understand the technical reason behind the pattern and may not even recognise that the pattern is valuable, but they know it because they have lived beside it, and frontier intelligence may give us a way to discover how much useful knowledge has always been sitting inside ordinary experience without the tools to properly examine, organise or extend it.

This is very different from simply giving everyone better answers, because answers are only one part of what becomes possible. Imagine instead that a sufficiently capable system begins asking better questions about your own experience, organising observations you have never connected, testing assumptions you have carried for years and helping you turn practical knowledge into something that can be examined, repeated or built upon. You might arrive with no technical vocabulary at all and simply say, “Every Friday this part of my work becomes a mess,” but that sentence already gives the system somewhere to begin. Why Friday? Which tasks change, what information arrives late, which decisions keep repeating, what can be predicted or automated, and what knowledge currently exists only inside one person's head? What you previously experienced as an irritation can gradually become visible as a system, and once you can see the system, you may discover that you were carrying far more useful information than you realised.

This is where my thinking around Fred Studio becomes relevant, not simply as a place for building AI products but as a possible interface between frontier capability and ordinary human problems. Most people will not arrive thinking in applications, architectures or agents; they will arrive thinking, “this keeps happening,” “this takes too long,” or “I know something is wrong here but I do not know how to explain it.” That can be enough to begin. The first task is to understand the person, their work, recurring decisions, environment and the knowledge they may have stopped noticing because it has become ordinary, then determine what intelligence can be built around that problem.

The wider point goes beyond any particular studio or product. If frontier intelligence combined with exceptional human insight can produce exceptional outcomes, we should also be interested in what happens when the human on the other side is not exceptional at all. Frontier AI with a frontier scientist may produce new science, with a brilliant entrepreneur it may create new companies and with an experienced professional it may uncover stronger solutions inside an established field, but frontier AI with an average person carrying an ordinary problem still leaves us with a question mark, and that question mark may tell us more about the social impact of AI than many benchmark scores will.

It would be easy to turn this into another promise that giving ordinary people extraordinary intelligence will automatically create extraordinary outcomes, but access to intelligence does not erase every difference between people. Put two people inside the same library and they will not necessarily leave with the same understanding, just as two people can receive the same tools and use them at completely different levels. If AI makes knowledge, explanation, technical execution and problem solving dramatically easier to reach, then curiosity, judgment, initiative, taste and persistence may all matter more, along with the ability to recognise a useful answer, the willingness to test something instead of merely discussing it and the judgment to know which problems deserve attention when the cost of attempting a solution begins to fall.

That creates a less comfortable version of the AI democratisation story because the common man may gain access to intelligence previous generations could never have imagined while still discovering that intelligence was only one part of what held him back. “I cannot build this because I cannot code” becomes weaker when the system can increasingly help with the coding, while “I cannot research this” changes when research can be assisted and “I do not understand statistics” becomes less final when statistics can be explained and performed interactively. Even “I cannot afford a consultant” begins to change when parts of consulting expertise become available through software, and as these familiar barriers weaken one after another, we arrive at something we have rarely been able to examine properly at scale: how much of a person's limitation came from capabilities they did not possess, and how much came from what they chose to do with the capabilities available to them?

Human capability has always been expensive and unevenly distributed because expertise required years of training, advice required access, building required teams, research required time and software required technical knowledge, while money, language, geography and networks created additional distance between ordinary people and what they might have been capable of creating. Frontier AI may begin removing pieces of that distance without removing human responsibility for what happens next, and this is why the future of the common man may not simply be a story about empowerment, it may also become a story about exposure. When increasingly powerful intelligence becomes easier to reach, we may see more clearly what people actually do when capability is placed in front of them. Some will build, some will learn and some will solve problems nobody previously considered important enough to solve, while others may use the same intelligence primarily for convenience or entertainment, so access can become widespread without outcomes becoming remotely equal.

The point is not to decide the answer before the evidence arrives. Instead, consider what becomes possible when the person on the other side of the frontier brings nothing extraordinary except years of experience with a problem. Could the system discover that the problem is more valuable than the person carrying it realised, uncover patterns hidden inside informal experience or help turn personal knowledge into something transferable? Someone who never considered themselves an innovator may suddenly be able to build a useful tool because the technical distance between recognising a problem and attempting a solution has collapsed, while an observation carried almost casually for ten years could become the beginning of a business, a better workflow or a solution for thousands of people dealing with the same thing. If these outcomes begin appearing, they would tell us something important about what intelligence can unlock, and if they do not appear nearly as often as expected, that would tell us something equally important about where the real limits of human and AI collaboration sit.

The common man therefore gives us a particularly useful test because he removes the easiest explanation. If someone extraordinary produces something extraordinary with AI, we can always point toward the person, but if ordinary people begin producing work that previously required teams, understanding problems that once required specialists or turning small pieces of lived knowledge into systems with real value, then something more fundamental may be happening. Intelligence would no longer merely be making experts more powerful; it would be changing the minimum capability required to participate, and that could become one of the most important economic consequences of AI because the company may not be the only unit receiving more capability. The individual may gradually acquire pieces of what previously belonged mainly to organisations: research, analysis, design, programming, translation, planning, administration and strategy.

These systems do not need to become perfect replacements for specialists before this begins to matter, because even partial access changes what one person can attempt. Someone who previously needed five different people before an idea could move may suddenly be able to travel much further alone, and the interesting person at the frontier may no longer always be the scientist, engineer or specialist who already knows what to ask. It may also be someone carrying nothing more sophisticated than a problem they understand because they have lived beside it for years.

We have spent enormous effort asking how extraordinary artificial intelligence can become, building larger models, more capable agents, faster infrastructure and increasingly powerful systems while watching the frontier continue moving upward, but another movement deserves our attention because the frontier may also be travelling in the opposite direction, toward the person who has never written code, the worker who has never thought of himself as an innovator, the small business owner who has never hired a consultant, and the teacher, mechanic, farmer, nurse, parent or shopkeeper who knows something useful but has never had the machinery to do very much with it. The Frontier and the Rest of Us is ultimately about that meeting point, where increasingly extraordinary intelligence begins reaching increasingly ordinary lives and we start discovering what was limited by access to capability, what was already sitting inside human experience, and what still depends on the person receiving it.

That is also where this paper meets the larger investigation that follows. If the frontier is beginning to reach the rest of us, then understanding access is only the beginning, because the more difficult question is what actually happens when human knowledge, judgment, curiosity, experience and intent begin working continuously with intelligence that is becoming more capable than any tool we have previously had. The next papers will follow that interaction rather than assume its conclusion, watching the evidence, the experiments, the failures and the unexpected outcomes as they arrive, because the frontier may not only be something humanity is building somewhere ahead of us.

It may also be something we are about to discover between us.

— Teff