The Price of initiative just collapsed

From libraries to large models: why the next divide is between people who try things and people who don’t.

I was in New York last weekend and went to The Morgan Library & Museum — not because I needed “a break from screens” or for some virtue signalling on Instagram, but because I love it. Libraries are one of the few inventions that still feel unequivocally on the side of human flourishing.

The Morgan is a hymn to the book — to reading as a form of agency. It reminds you that civilisation is, in part, a long chain of people trying to pass ideas forward without mangling them. It also makes you briefly understand the appeal of being a robber baron — not for the monocles or moral flexibility, but for the means to build something that says books mattered enough to deserve architecture.

As you leave, you turn and meet the Gutenberg Bible. The Morgan is the only institution on earth that holds three copies — which, by any measure, is a flex.

We tend to tell the story of printing as if the press arrived and — boom — knowledge spread. That’s true but incomplete. What struck me most standing there wasn’t just the leap forward; it was the bottleneck that remained. Printing made books. It didn’t make readers.

Even if you could get your hands on a printed Bible, you still needed literacy — and not literacy as we know it: with public schools, clear fonts, and the assumption that the words were for you. You needed the language, the time, and the permission to learn. The technology was astonishing. The interface, for most people, was not.

The future had arrived — and then it queued.

It’s that queue that followed me out of the building. Gutenberg printed his Bible around 1455. England did not even begin legislating for a national system of elementary education until 1870 — four centuries later, after long arguments over who should learn and what they should be allowed to read. The machine was fast. Everything else — institutions, access, the will to distribute — was slow. Those who could read surged ahead; those who could not were quietly kept in place.

Time is the thread I keep returning to, because it’s what makes today harder. We do not have four centuries to close the gap between technology and society. We may not even have four years. New releases from Anthropic, OpenAI and newcomer Openclaw have blown apart what’s possible in the last 3 weeks.

Whenever someone says, “AI is changing everything,” people’s eyes glaze over — and they’re right to let them. Most commentary is either breathless or moralising. Neither helps. It’s all either: “It’s the end of work!” or “It’s just autocomplete!” — and both miss the real question: what has become cheap, fast, and widely available that wasn’t before?

You only really notice the shift when you give these systems a job you’d normally hand to a competent colleague. Not party tricks, not poems — real work with consequences. The change isn’t that the machines have become mystical; it’s that they’ve crossed the threshold of usefulness. They can now hold context long enough to complete messy, multi-step work — coherent, continuous, close enough to done.

I tested that in a small way. In thirty minutes with my iPad, I went from “I want an app that does seating plans for parties” to a working version — guest names stored, tables generated, usable interface. I am not a developer. The AI did the building. I did the thinking and the deciding.

My main lesson? The price of initiative is collapsing.

When the cost of trying falls, the number of attempts rises. More prototypes, more internal tools, more strange little experiments that used to die in the “too hard / too expensive / not worth it” stage. The optimistic version of the future is a broadening of agency: more people able to act on their ideas, not just talk about them.

The less comfortable version is that advantage flows to those who adapt fastest — not because they’re smarter or better, but because they compound earlier. They redesign workflows first. They learn what to trust and what to check. They shrink the time between insight and execution. The laggards don’t get a soft landing; they wake up to find the baseline has moved beneath their feet.

The impact won’t arrive as one big shock. It will seep in through a thousand local compressions: a team of ten becoming a team of seven; the “first draft” role evaporating; the entry-level rung thinning out; the work that trained people being automated before new ways to learn exist. Think of the analyst who spent two days building a first draft from messy notes and spreadsheets — that task still exists, but it now takes an hour and a check rather than a career apprenticeship.

Time is warping. There are now multiple clocks running at once — none synchronised.

There is the technology clock, which moves in lurches — step changes every few weeks.

There is the economic clock, which lags and then moves all at once. A large organisation can look at all this and shrug — not yet material, not yet embedded. Meanwhile a startup is being built right now that assumes these capabilities as the default: cheap synthesis, rapid prototyping, workflows stitched together because the bottleneck isn’t implementation anymore — it is only deciding what to build. Even if big companies do not feel it this quarter, the competitor is being assembled anyway.

There is a corporate clock, running inside a small number of companies that own the data centres, the GPUs, and the foundation models. They decide which capabilities are released, on what terms, and in which languages — and their incentives do not automatically match the interests of everyone who will have to live with the consequences.

There is the political clock, which runs on headlines. It struggles to sustain attention on anything structural when the next crisis arrives before lunch.

And there is the human clock — the one that matters most — which is the time it takes an actual person to learn, to change their working life, to feel secure enough to take a risk.

We need a government that can hold two truths at once: the immediate, visible urgencies (cost of living, housing, NHS waits) and the structural shifts already transforming what “work” and “opportunity” mean. These aren’t competing priorities — they’re the same priority seen from two distances.

If I were designing AI policy, I’d make one practical promise: not to “reskill Britain” in the abstract, but to help people stay effective as the baseline moves. Policy has one real job here — to expand agency and buy time.

Here are four starting moves.

First, access. Put secure, supervised AI tools in people’s hands — in JobCentres, libraries, and community hubs — with guardrails and trained support. If a 50‑year‑old logistics coordinator in Sunderland can’t safely use these tools to rebuild her CV or test a business idea, we’ve already decided who this technology is for.

Second, visibility. Build a public dashboard tracking which tasks are being automated, in which sectors, and how quickly. A transition you can’t see can’t be governed.

Third, shared upside. When public contracts become cheaper because AI did the middle, the public should share the benefit — through better services, lower costs, or reinvestment in skills. Write it into procurement.

Fourth, agency in schools. Not lectures about AI, but making with it: build something, present it locally, reflect on what was human and what was machine. That changes how young people see both the tools and themselves.

Gutenberg’s revolution wasn’t just speed; it changed what was scarce. The bottleneck moved from copying to comprehension. Society had to reorganise around a new constraint.

It’s happening again. The question is no longer “Can this be built?” but “Who realises it already can?”

Capability is moving faster than comprehension, and when the bottleneck moves, power moves with it — towards the people and firms who see it first and have the resources to act on it.

So the task now is not to restrain the technology, but to widen the circle of people who can see it clearly enough to use it — and to shape it — on their own terms, in workplaces, in communities, and in the public sector

.