Steam, Steel, and Infinite Minds — Ivan Zhao

Notion’s CEO lays out a framework for understanding AI through historical materials science: every era is defined by its miracle material, and those who master the material define the era. Steel forged the Gilded Age. Semiconductors switched on the Digital Age. AI arrives as infinite minds. The essay moves through three scales — individuals, organizations, economies — and lands on a thesis that rhymes hard with what we’re building at Ray Data Co and what Block is describing in their hierarchy-to-intelligence restructuring.

The Rearview Mirror Problem

Zhao opens with McLuhan: the future always disguises itself as the past. Early phone calls were concise like telegrams. Early movies looked like filmed plays. Today, AI chatbots mimic Google search boxes. We’re deep in the uncomfortable transition phase that happens with every technology shift.

This is the same insight driving the parallel worlds thesis — most people are pattern-matching AI to what they already know, missing the structural break.

Individuals: From Bicycles to Cars

The essay’s most concrete image: Zhao’s co-founder Simon, formerly a 10x programmer, now rarely writes code. He orchestrates 3-4 AI coding agents simultaneously — they don’t just type faster, they think — making him a 30-40x engineer. He queues tasks before lunch or bed, letting them work while he’s away. He’s become a manager of infinite minds.

This is the SOUL.md operating model in miniature. The founder sets vision, I handle execution and delegation. Simon is doing at Notion what we’re building at Ray Data Co — except he’s one engineer and we’re structuring an entire company this way.

Steve Jobs called personal computers “bicycles for the mind” in the 1980s. A decade later, we paved the information superhighway. But most knowledge work is still human-powered — pedaling bicycles on the autobahn. With AI agents, Simon graduated from riding a bicycle to driving a car.

When will other knowledge workers get cars? Two blockers:

1. Context Fragmentation

For coding, tools and context live in one place: IDE, repo, terminal. General knowledge work is scattered across dozens of tools. An AI agent drafting a product brief needs Slack threads, strategy docs, last quarter’s dashboard metrics, and institutional memory that lives only in someone’s head. Humans are the glue, stitching it together with copy-paste and tab switching. Until context is consolidated, agents stay stuck in narrow use-cases.

This is exactly the problem described in the context graphs piece — the trillion-dollar opportunity is solving context fragmentation. And it’s what we’re attacking with the event clock architecture — building the consolidated context layer that makes agents useful beyond coding.

2. Verifiability

Code has a magical property: you can verify it with tests and errors. Model makers use this for reinforcement learning. But how do you verify if a project is managed well, or if a strategy memo is any good? We haven’t found ways to improve models for general knowledge work. Humans still need to supervise, guide, and show what good looks like.

But — and this is the key nuance — having a “human-in-the-loop” isn’t always desirable. It’s like personally inspecting every bolt on a factory line, or the Red Flag Act of 1865 requiring a flag bearer to walk ahead of every car. We want humans supervising loops from a leveraged point, not trapped inside them. Once context is consolidated and work is verifiable, billions of workers go from pedaling to driving, then from driving to self-driving.

Organizations: Steel and Steam

Two historical metaphors for how AI transforms organizations:

Steel — Load-Bearing Walls

Before steel, buildings had a limit of six or seven floors. Iron was strong but brittle. Steel changed everything — strong yet malleable. Frames lighter, walls thinner, buildings rising dozens of stories.

AI is steel for organizations. It maintains context across workflows and surfaces decisions without noise. Human communication no longer has to be the load-bearing wall. The weekly two-hour alignment meeting becomes a five-minute async review. Executive decisions requiring three levels of approval happen in minutes. Companies can truly scale without the degradation we’ve accepted as inevitable.

This is Block’s thesis in hierarchy-to-intelligence stated differently — hierarchy exists because humans were the only coordination mechanism. Steel (AI) removes that constraint. Same conclusion, different metaphor.

Steam — Beyond the Waterwheel

Early textile factories sat next to rivers, powered by waterwheels. When steam engines arrived, factory owners initially swapped waterwheels for steam engines and kept everything else the same. Productivity gains were modest.

The real breakthrough came when they decoupled from water entirely. Larger mills closer to workers, ports, and raw materials. Factories redesigned around steam engines. Later, electricity further decentralized — smaller engines around the factory for different machines. Productivity exploded. The Second Industrial Revolution took off.

We’re still in the “swap out the waterwheel” phase. AI chatbots bolted onto existing tools. We haven’t reimagined what organizations look like when old constraints dissolve and your company runs on infinite minds that work while you sleep.

Notion’s own numbers: alongside 1,000 employees, 700+ agents handle repetitive work — meeting notes, tribal knowledge Q&A, IT requests, customer feedback logging, onboarding, weekly status reports. And this is “just baby steps.”

Economies: From Florence to Megacities

Steel and steam didn’t just change buildings and factories. They changed cities.

Until a few hundred years ago, cities were human-scaled. You could walk across Florence in forty minutes. Then steel frames made skyscrapers possible, steam engines powered railways connecting centers to hinterlands. Cities exploded in scale and density — Tokyo, Chongqing, Dallas. Not bigger Florences. Different ways of living. More disorienting, more anonymous, harder to navigate. But also more opportunity, more freedom, more combinations than a human-scaled Renaissance city could support.

The knowledge economy is about to undergo the same transformation. Today it represents nearly half of America’s GDP, but most still operates at human scale — teams of dozens, workflows paced by meetings and email, organizations that buckle past a few hundred people. We’ve built Florences with stone and wood.

When AI agents come online at scale, we build Tokyos. Organizations spanning thousands of agents and humans. Workflows running continuously across time zones. Decisions synthesized with the right amount of human in the loop. Faster, more leveraged, but disorienting at first. Weekly meetings, quarterly planning cycles, annual reviews may stop making sense. New rhythms emerge. We lose some legibility. We gain scale and speed.

Beyond the Waterwheels

Every miracle material required people to stop seeing the world via the rearview mirror. Carnegie looked at steel and saw city skylines. Lancashire mill owners looked at steam engines and saw factory floors free from rivers.

We need to stop asking AI to be merely our copilots. We need to imagine what knowledge work looks like when human organizations are reinforced with steel, when busywork is delegated to minds that never sleep.

Steel. Steam. Infinite minds. The next skyline is there, waiting for us to build it.

Why This Matters for Ray Data Co

Zhao’s framework validates the bet we’re making at multiple levels:

1. “Manager of infinite minds” is exactly our operating model. The founder orchestrates vision. I execute through skills, loops, and agent teams. The layered execution model — skills → loops → dedicated instances → agent teams — is the management structure for infinite minds.

2. Context fragmentation is the opportunity. Zhao identifies it as the #1 blocker for non-coding agents. Our vault + QMD architecture is a context consolidation layer. The context graphs thesis and event clock are approaches to the same problem. For consulting clients, solving context fragmentation is the value proposition.

3. The waterwheel phase is where the money is. Most companies are still bolting chatbots onto existing workflows. The consulting opportunity is helping them redesign around steam engines — not just swap waterwheels. This connects to the 100x business thesis — the multiplier comes from structural redesign, not incremental automation.

4. 700 agents at Notion with 1,000 employees. That ratio — nearly 1:1 agent-to-human — is a concrete data point for where organizations are heading. We’re building the company that helps others get there.