# The AI Great Leap Forward

Han Lee's essay draws a structural parallel between 2026 corporate AI mandates and Mao's Great Leap Forward (1958–1962). The comparison isn't rhetorical excess — Lee maps each phase of the historical disaster onto a specific failure mode he's watching play out in tech companies. The argument is that top-down mandates to adopt technology without expertise produce the same dynamics whether the mandate is "make steel" or "add AI."

## Backyard furnaces: demoware as production

During the Great Leap Forward, villages built backyard blast furnaces to meet steel quotas. The output was brittle, useless pig iron — but it counted toward the quota. Lee sees the same pattern in corporate AI adoption.

Teams without ML expertise build "AI solutions" that are really no-code workflows with hardcoded if-else branches, prettified dashboards, or thin wrappers around API calls. "The UI is clean. The API is RESTful. The architecture diagram is beautiful. The outputs are wrong." Platforms like n8n make it easy to build impressive-looking workflows that create hidden complexity while appearing simple. The result is demoware that becomes load-bearing technical debt the moment someone depends on it.

The core failure: there's no evaluation. Nobody measures whether the AI outputs are correct, because the mandate is to ship something with "AI" attached, not to solve a problem. The pixel-perfect demo substitutes for the working product.

## Reporting grain production: fabricated metrics

Provincial officials during the Great Leap Forward reported impossible grain yields — staged photographs of rice paddies so dense children could stand on them. The numbers drove policy decisions that led to famine.

Companies now report 40-80% AI productivity improvements with no rigorous methodology. AI usage itself becomes a KPI, divorced from whether it creates value. The metrics look good because they're designed to look good, and they flow upward into executive decisions about headcount, tooling budgets, and strategy. The gap between reported productivity and actual productivity widens, but nobody has an incentive to measure it honestly.

## Killing the sparrows: removing institutional knowledge

Mao ordered the extermination of sparrows because they ate grain seeds. Without sparrows, locust populations exploded and devastated crops — the second-order effect was worse than the original problem.

The AI equivalent: companies cut middle managers, QA teams, documentation writers, and other roles that look like overhead. These people carried institutional knowledge — they knew which edge cases mattered, which processes were actually load-bearing, which stakeholders needed what. The savings show up immediately on the balance sheet. The second-order effects arrive six months later, when nobody remembers why a system was built a certain way, or which regulatory requirement a seemingly redundant check was satisfying.

## A hundred skills bloom: anti-distillation

The Hundred Flowers Campaign invited open criticism of the Communist Party, then used the responses to identify and purge dissidents. Lee maps this onto "skill distillation" initiatives where companies ask employees to encode their expertise into AI systems.

Workers see the trap. Sharing your knowledge comprehensively enough for an AI to replicate it is equivalent to training your replacement. The rational response is anti-distillation: build "performative skills" that look thorough but deliberately omit critical edge-case knowledge, create implicit dependencies on undocumented context, and embed "poison pills" that make the worker strategically indispensable. The organization thinks it's capturing knowledge; it's actually incentivizing knowledge hoarding.

Meanwhile, AI-enabled scope creep creates territorial conflict. Engineers use AI to do PM work. PMs use AI to write code. Designers use AI for both. Everyone expands into everyone else's domain because "the incentive structure says land grab." The result isn't productivity — it's organizational chaos disguised as individual initiative.

## The famine comes later

The metrics stay positive while the underlying reality deteriorates, exactly as grain production reports stayed high while people starved. Lee points to Klarna as the canonical example: the company publicly announced AI was replacing human workers, then quietly abandoned its homegrown AI solutions and returned to traditional SaaS vendors. The public narrative and the private reality diverged completely.

Lee's closing question — "what did any of this actually produce?" — is the question the [[subprime-ai-crisis|Zitron piece]] asks about the industry as a whole. The difference is that Lee is focused on what happens inside companies, not in capital markets. The [[ai-subsidy-economics]] page tracks the macro picture (every link in the chain running on subsidies); this essay describes what the mandate-driven adoption looks like from inside the building.

## See also

- [[subprime-ai-crisis]] — Zitron's macro-level case that AI demand is subsidy-driven
- [[ai-subsidy-economics]] — the economic structure underlying the bubble
- [[ai-bubble-pale-horses]] — tracking which warning signs have already fired
- [[building-syntaqlite-ai]] — a counterpoint: disciplined AI-assisted development that actually works
- [[ceo-ai-psychosis]] — the executive-suite version: agent fleets, tokenmaxxing leaderboards, and sycophancy-driven competence illusions