Nubian Code — The Best AI for the Most People

Pulse

The Best AI for the Most People

As AI models become increasingly powerful and accessible, the biggest determinant of value is no longer model intelligence itself, but how effectively it is combined with the knowledge, workflows, and context of the people who use it.

Was this newsletter forwarded to you? to get it in your inbox.

A question worth answering

Which AI is the best? It is the most common question in technology today, and most answers dodge it. They compare benchmark scores, list each assistant's strengths and conclude that it depends.

It does depend, but not on the things usually compared. Ask a narrower question, which AI is best for the most people, and a clearer answer appears.

The short version: no single assistant wins. The best AI for the most people is general intelligence underneath, with the knowledge of a person's work, market and institution built on top.

Who "most people" really are

"Most people" is usually imagined as a consumer asking a chatbot for a recipe or a travel plan. That is a real use, but it is not where most of the hours go.

Most people spend most of their day at work. They are clerks, analysts, engineers, administrators, managers and owners of small firms, and the largest share of them live outside the markets where AI products are designed.

So the best AI for the most people is, in practice, the best AI for working people. Their needs are concrete: write the document, check the figures, follow the rules, finish the task.

Three measures of "best"

For working people, three measures matter more than any benchmark.

Reach. Can people get to it easily, at a price that makes sense where they live, inside the tools they already use? An AI that sits behind an unfamiliar interface or an expensive plan serves a few, not most.

Reliability. Does it get the answer right, show where the answer came from and say when it is unsure? At work, one confident mistake can outweigh a month of good drafts.

Relevance. Does it know the rules, records, terms and practices of the person's actual job and market? Without this, even a correct general answer is the wrong answer for the situation.

How today's approaches measure up

Today's options fall into four broad approaches. Each is strong on one or two measures. None is strong on all three by itself.

The pattern is clear. Reach comes from distribution, reliability from focus and relevance from knowledge. The best AI for the most people has to combine all three.

ApproachReachReliabilityRelevance
General assistants (standalone chat apps)Wide, with free tiers, but outside everyday work toolsStrong and improving; sources not always shownBroad knowledge, thin on local rules and practice
Ecosystem assistants (built into email, office suites, phones, messaging)Widest, already where people workVaries by taskKnows your files, not your sector or market
Open-weight modelsDepends on someone building on themDepends on how they are deployedCan be adapted, but only with effort and data
Specialised applications (built for one job or market)Narrower, sold to firmsHigher, because the scope is definedDeepest, grounded in sector and local knowledge

The answer: intelligence underneath, knowledge on top

The best AI for the most people is not one product. It is a stack of layers.

At the bottom sit the general models. They keep getting more capable and cheaper, and they increasingly behave like infrastructure. Above them sits the layer that makes them useful: the accumulated knowledge of a sector, a market and an institution, organised so AI can draw on it and act on it.

That top layer is where reliability and relevance come from, and it is the layer most people are still missing. The labs will keep improving the foundation. The builders who capture and organise working knowledge, and deliver it inside the tools people already use, will decide who it actually serves.

The question, answered

So, which AI is best for the most people? Not the one at the top of this month's leaderboard, and not any single assistant on its own.

It is the one that reaches people where they work, gets their work right and understands the world they work in. Today that means general intelligence paired with deep, local, well-organised knowledge.

The smartest models are being built. The most useful AI for the most people is still being assembled, one sector and one market at a time.