The S-Curve of Human Progress

The story of human civilization is not a straight line. It’s a series of S-curves — periods of explosive growth triggered by fundamental innovations, each eventually plateauing as the world absorbs the change, only for a new revolution to ignite the next curve.

What is an S-Curve?

An S-curve (logistic curve) describes how new technologies and paradigms spread through society. At first, progress is slow — only a few early adopters. Then comes the steep middle section: exponential-looking growth as the innovation reaches critical mass. Finally, saturation — the world has absorbed the change, and growth decelerates.

Every major revolution in human history follows this pattern. Let’s trace the arc.

The Agricultural Revolution (~10,000 BC)

For hundreds of thousands of years, humans lived as hunter-gatherers. World population hovered around 5–10 million people, constrained by what the land could naturally provide.

Then, in the Fertile Crescent, humans began cultivating crops and domesticating animals. This wasn’t an overnight change — it took roughly 5,000 years to spread across Eurasia, Africa, and the Americas. But its impact was transformative.

The numbers:

• World population before agriculture: ~5–10 million
• World population by 1 AD (after widespread adoption): ~250 million
• Carrying capacity of the land: increased 10–100x

The S-curve: Agriculture spread from its origins around 10,000 BC to cover most of Eurasia by 5,000 BC, reaching the rest of the world over the following millennia. The adoption curve was slow at first, accelerated as farming techniques improved and populations grew, and eventually saturated as virtually all major civilizations adopted it.

GDP per capita barely moved. The Malthusian trap meant that any productivity gains were absorbed by population growth. The agricultural revolution changed what was possible — larger settlements, division of labor, the emergence of writing, law, and organized religion — but individual material wealth remained essentially flat for millennia.

The Scientific Revolution (~1500–1700)

The scientific revolution didn’t boost GDP directly, but it might be the single most important inflection point in history. It gave humanity a technology for discovering technologies: the scientific method.

Copernicus, Galileo, Kepler, Newton — in the span of 200 years, humans went from believing the sun orbited the Earth to deriving universal laws of motion and gravitation. The key insight wasn’t any single discovery, but the process: systematic observation, hypothesis, experiment, revision.

The S-curve: Scientific thinking spread from a handful of European natural philosophers to become the dominant framework for understanding reality. By the 18th century, the Enlightenment had embedded scientific reasoning into education, governance, and commerce. The method became a platform — an accelerant for every revolution that followed.

You could argue that the scientific revolution never truly saturated. Its S-curve merged into the industrial revolution’s, forming a compound curve that’s still climbing.

The Industrial Revolution (~1760–1914)

This is where the hockey stick begins.

For most of human history, global GDP per capita was essentially flat — roughly $500–$615 per year in 1990 international dollars, whether you measured it in 1 AD or 1700 AD. The Maddison Project data makes this strikingly clear: a near-horizontal line for eighteen centuries, then a sudden, steep ascent.

The steam engine changed everything.

GDP per capita over time (Maddison Project, 1990 international dollars):

The pattern is unmistakable. Virtually no growth for 1,700 years, then a doubling in a century, then another doubling in half a century, then a tripling in fifty years.

The S-curve of steam and electricity:

US household electrification is one of the cleanest S-curves in recorded history:

• 1882: First commercial power station (Pearl Street, NYC) — effectively 0%
• 1907: ~8% of US homes electrified
• 1920: ~35%
• 1930: ~68%
• 1940: ~80% (rural areas lagging)
• 1945: ~85%
• 1956: ~98%

Slow start, steep middle, saturation. The entire transformation — from zero to near-universal — took about 75 years.

The industrial revolution reshaped everything: where people lived (urbanization surged from ~10% to over 50%), how long they lived (life expectancy in England rose from ~40 years in 1800 to ~50 by 1900), and what they could produce (manufacturing output per worker increased roughly tenfold between 1800 and 1900).

The Digital Revolution (~1970–2020s)

Computers compressed what the industrial revolution accomplished in a century into decades.

Internet adoption (global users):

Smartphone adoption (global):

Both are textbook S-curves — slow start, explosive middle, approaching saturation.

The digital revolution didn’t just increase productivity. It changed what a person could be. A teenager with a laptop now has access to more information than an entire university had in 1990. A solo developer can build what once required a team of fifty. A musician in a bedroom can reach a global audience with zero gatekeepers.

We’re near the top of this particular S-curve. Smartphone penetration in developed nations exceeds 90%. Internet access is approaching global saturation. The transformative gains of digitization — while still compounding — are flattening.

Which brings us to the question.

Zoom out far enough, and the revolutions themselves form a pattern: each S-curve sits on the shoulder of the last, steeper and faster than what came before.

Each curve is shorter and steeper than the one before. The agricultural revolution played out over millennia. The industrial revolution over about 150 years. The digital revolution in roughly 50. And the intelligence revolution is climbing faster than any of them.

The Intelligence Revolution (2020s–?)

We are living through the beginning of a new curve.

In November 2022, ChatGPT launched and reached 100 million users in two months — the fastest consumer technology adoption in history. For comparison, it took the telephone 75 years to reach 100 million users, the internet about 7 years, and Instagram 2.5 years.

But user count is only the surface. What makes the intelligence revolution structurally different from every revolution before it is the number of simultaneous compounding forces driving it forward.

1. Adoption into workflows

AI is being integrated into software development, writing, research, legal work, medicine, customer service, education, and creative work — all at once. Unlike electricity (which required rewiring factories) or the internet (which required building infrastructure from scratch), AI often slots into existing digital workflows with minimal friction.

A developer installs an AI coding assistant and becomes measurably more productive the same day. A researcher pastes a paper into an LLM and gets a summary in seconds. A lawyer uses AI to review contracts in minutes instead of hours. The adoption curve is steep because the barriers to adoption are low.

2. Better data from usage

As hundreds of millions of people interact with LLMs daily, the resulting data — queries, preferences, corrections, feedback — becomes training signal for the next generation of models. This is a feedback loop that didn’t exist in previous revolutions: the product improves the product. More users generate more data, which produces better models, which attract more users.

3. Infrastructure is being built

The tooling around AI is maturing at extraordinary speed. Retrieval-augmented generation (RAG) pipelines, agent frameworks, fine-tuning platforms, evaluation suites, specialized hardware, edge deployment tools — an entire ecosystem is being constructed to make AI more capable and more accessible. Every piece of infrastructure unlocks new use cases, which drive demand for more infrastructure.

4. AI accelerates AI research

Researchers use LLMs to write code faster, analyze papers, generate hypotheses, debug experiments, and explore solution spaces. The tools for building intelligence are themselves becoming intelligent. This creates perhaps the most powerful feedback loop: better AI → faster AI research → better AI.

When researchers can prototype in hours what used to take weeks, and when AI agents can explore architectural variations autonomously, the pace of fundamental progress accelerates in ways that are difficult to model.

5. Compounding capital investment

Global AI investment exceeded $150 billion in 2024 and is accelerating. Hyperscalers are building massive GPU clusters. Startups are funded at unprecedented valuations. This capital builds capability, which demonstrates commercial value, which attracts more capital. The investment cycle has a momentum of its own.

6. Open-source democratization

Open-weight models — Llama, Mistral, DeepSeek, Qwen, and others — mean that AI capability isn’t locked behind a single corporation. Researchers, developers, and small companies worldwide can run, fine-tune, and build on frontier-class models. This broadens the base of innovation dramatically, turning what could be a narrow technological advance into a global one.

7. Cross-domain amplification

AI isn’t just improving one field — it’s improving many fields simultaneously, and advances in one domain feed back into others. Better AI for drug discovery generates biological insights that improve medical AI. Better AI for code accelerates the development of AI infrastructure. Better AI for mathematics helps verify AI safety proofs. The intelligence revolution is not a single S-curve but a web of interlocking curves, each amplifying the others.

Where Are We on the Curve?

This is the question that matters — and the honest answer is: we don’t know.

Consider the recent history. When GPT-4 launched in March 2023, it felt like the ceiling had risen permanently. Then, by late 2024, a narrative emerged that large language models were “plateauing” — that scaling laws were hitting diminishing returns, that we’d reached the limits of the transformer architecture, that the next leap required fundamentally new approaches.

Then coding agents arrived and shattered that narrative. Not by being bigger models, but by being better systems — demonstrating that how you deploy intelligence matters as much as the raw capability. Agents that can read, write, execute, and reason across entire codebases turned “impressive chatbot” into “productive colleague.” Reasoning models showed that inference-time compute could unlock capabilities that no one predicted from the training curves alone.

This is a pattern that repeats in every S-curve. The middle section often looks like a plateau if you zoom in too close. Electricity had decades of experimentation before factories were redesigned around it and productivity exploded. The internet had the dot-com crash of 2000 — a period where many declared the technology overhyped — before it became the backbone of civilization. The smartphone had years of clunky PDAs and stylus-based interfaces before the iPhone triggered the steep part of the curve.

The Honest Conclusion

We might be at position A — still in the early phase, with the most dramatic transformation yet to come. We might be at B — right at the inflection point, living through the period of maximum change. We might even be at C — closer to the ceiling than we think.

But consider the evidence: every previous revolution had one or two compounding forces. The agricultural revolution had better crops feeding more people who could farm more land. The industrial revolution had machines producing goods that funded more machines.

The intelligence revolution has at least seven simultaneous compounding forces, and it’s being adopted faster than any technology in human history.

Here’s what we can say with confidence:

Previous revolutions reshaped civilization over decades to centuries. The agricultural revolution played out over millennia. The industrial revolution over roughly 150 years. The digital revolution over about 50 years. Each cycle is faster than the last.

If the intelligence revolution follows this pattern of acceleration — and there’s every reason to think it will, given the number of compounding forces — then we may be living through the steepest part of the steepest S-curve humanity has ever produced. The changes happening in 2025 and 2026 might look, from the vantage point of 2050, like the weeks in 1995 when the world suddenly “got” the internet. A moment of transformation so rapid that people living through it couldn’t fully grasp it.

We cannot predict the future. But we can observe that every metric — adoption speed, investment growth, capability improvement, cross-domain impact — is moving faster than any comparable period in history. The world might be in the middle of a more rapid transformation than it has ever seen, including the recent technological revolution that put a computer in every pocket.

There’s one thing we can say with confidence: this is a very wild time to be alive.

Sources

1. Maddison Project Database (2020). Bolt, J. and van Zanden, J.L., “Maddison style estimates of the evolution of the world economy. A new 2020 update.” GDP per capita figures in 1990 international Geary-Khamis dollars. rug.nl/ggdc/historicaldevelopment/maddison/
2. World population estimates. United Nations, Department of Economic and Social Affairs, World Population Prospects 2024. Pre-agricultural population estimates from Kremer, M. (1993), “Population Growth and Technological Change: One Million B.C. to 1990,” The Quarterly Journal of Economics, 108(3), 681–716.
3. US household electrification. US Census Bureau, Historical Statistics of the United States, Series S 108–119. Rural electrification data from the US Energy Information Administration (EIA).
4. Internet adoption data. International Telecommunication Union (ITU), ICT Statistics. itu.int/en/ITU-D/Statistics
5. Smartphone adoption. GSMA Intelligence, The Mobile Economy 2024; Statista, Number of smartphone users worldwide from 2016 to 2029.
6. ChatGPT adoption. Hu, K. (2023), “ChatGPT sets record for fastest-growing user base,” Reuters, February 2, 2023.
7. AI investment figures. Stanford University, AI Index Report 2025, Institute for Human-Centered Artificial Intelligence (HAI).
8. Life expectancy data. Roser, M., Ortiz-Ospina, E., and Ritchie, H. (2013), “Life Expectancy,” Our World in Data. ourworldindata.org/life-expectancy
9. S-curve / logistic growth model. Modis, T. (2002), “Forecasting the growth of complexity and change,” Technological Forecasting and Social Change, 69(4), 377–404.