draft gemini

🏗️ Phase 1: The “Group Drive” Era (Substituting the Engine)

When electricity was first introduced to factories in the 1880s, it did not immediately revolutionize production.

• The Paradigm: Most factories were built around a single, massive steam engine. This engine turned a “main shaft” that ran the length of the building. Machines were connected to this shaft by a complex, dangerous, and loud forest of leather belts and pulleys. • The Initial Swap: Factory owners simply replaced the steam engine with one large electric motor (the Group Drive). They kept the shafts, the belts, and the multi-story vertical factory layout (necessary to keep machines close to the central power source). • The Result: Productivity gains were negligible. Electricity was slightly cheaper than coal, but the fundamental bottleneck—the central shaft and belt system—remained.

⚙️ Phase 2: The “Unit Drive” Era (Redesigning the Workflow)

True productivity gains didn’t appear until 30–40 years later (circa 1910–1920), when engineers realized they didn’t need a central shaft at all.

• The Innovation: Small electric motors were placed on every individual machine (Unit Drive). • The Transformation: This decoupled the machine from the power source. • Layout: Factories could now be one-story, spread out, and organized by the logical flow of materials (leading directly to the modern assembly line) rather than by proximity to a central engine. • Modular Growth: You could add one machine without upgrading a giant central engine. • Safety & Environment: No more overhead belts meant better light, less noise, and fewer workplace deaths. • The Result: Annual secondary-total factor productivity growth in US manufacturing surged from 0.5% (pre-1910) to over 3.5% in the 1920s. This metric, popularized by Robert Solow (1957), isolates the impact of pure innovation and organizational redesign (what economists call “technical change” but in practice is engineering change) from simple capital accumulation.

🤖 The AI Parallel: Why This Matters Now

We are currently in the “Group Drive” era of AI.

1. AI as a Component Swap: Most current AI use cases are just swapping an LLM into an old workflow (e.g., “AI Chatbot” replacing “Human Support Agent” in the same old ticketing system). This is “Group Drive” AI—it’s more efficient, but the system architecture is the same.
2. The “Unit Drive” of AI: The real leap will be Agentic Workflows. This is where we stop building “apps” and start building “systems that build systems” (as teased by Tim Kellogg in recent social signals). In this era, we reorganize the company around the AI’s ability to act autonomously, rather than just using AI to help humans do the same old steps.
3. The Productivity Paradox: Like electricity, AI is a General Purpose Technology (GPT). Paul David’s seminal paper, “The Dynamo and the Computer” (1990), proves that it takes decades for a GPT to show up in national statistics because the “intangible capital” (redesigning the factory floor/rethinking the org chart) takes longer to build than the technology itself.
4. Capital vs. Engineering Change: Solow (1957) distinguishes between “capital accumulation” (buying more equipment) and “technical change.” In our context, true technical change means engineering change—reinventing the workflow rather than just buying more tokens and compute. The productivity boom depends on the latter.

📚 Recommended Articles & Citations

• The Foundation: Paul A. David (1990), “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox.” American Economic Review, Papers and Proceedings, 80(2), 355-61. (The definitive source for the electricity/computer/AI analogy).

• Key insight: David demonstrated that electrification followed a pattern remarkably similar to what we see with computers and AI today — initial adoption with minimal productivity gains, followed by a decades-long lag before the technology’s full potential was realized in aggregate statistics.
• The delay occurs because GPTs require complementary investments (factory reorganization, new skills, new business processes) that take time to build — what David called “intangible capital.”
• David noted that electricity was introduced to factories in the 1880s, but didn’t show up meaningfully in US productivity statistics until the 1920s — a 40-year gap. • The Measurement Framework: Robert M. Solow (1957), “Technical Change and the Aggregate Production Function.” Review of Economics and Statistics, 39(3), 312-320.
• Key insight: Solow pioneered the method to isolate “technical change” (which we can think of as engineering and process redesign) from simple capital accumulation. He found that the vast majority of growth in US output per worker in the first half of the 20th century was driven by this technical progress, rather than simply accumulating more capital. This provides the mathematical proof that the organizational shifts described in the “Unit Drive” era are the primary engine of long-term growth. • GPT Framework: Timothy F. Bresnahan & Manuel Trajtenberg (1992), “General Purpose Technologies: Engines of Growth?” Journal of Econometrics, 65(1), 83-108. NBER Working Paper No. w4148.
• Key insight: GPTs are characterized by three properties: (1) pervasiveness — they spread as inputs across many sectors, (2) inherent potential for technical improvement, and (3) “innovational complementarities” — R&D productivity in downstream sectors increases as the GPT improves.
• The authors show that these characteristics create increasing returns to scale, but also create coordination problems: without sufficient integration between GPT producers and users, we get “too little, too late” innovation in both sectors.
• This framework explains why AI’s productivity payoff is delayed — the technology itself improves rapidly, but the complementary organizational and process innovations needed to fully exploit it take decades to diffuse. • The Layout Shift: Warren Devine Jr. (1983), “From Shafts to Wires: Historical Perspective on Electrification.” (Great for details on the physical transformation of the factory floor).