Why AI Is the Ultimate Leverage for Individuals

Background

Hyung Won Chung, a former OpenAI top researcher who previously worked on PaLM, Flan and T5X at Google Brain, gave a remote talk at Cornell University where he introduced the idea of “AI as an ultimate form of leverage.”

New Understanding: Interpreting AI Leverage

Chung offers a cognitive framework for viewing AI as a fundamentally new lever that can reshape economic models, organizational structures, and the frontier of human knowledge. The analysis proceeds from the individual level to teams and finally to humanity as a whole.

1. Common Cognitive Biases

Humans are naturally not good at perceiving slow changes that occur over the course of years. AI may be the fastest‑developing technology in history, but its development is measured in years or even decades, not minutes or hours. - Hyung Won Chung

Because people are attuned to minute‑scale variations, they tend to underestimate the long‑term, compounding impact of AI, creating a blind spot in decision‑making.

2. General Definition of Leverage

In physics, a lever lets a small input produce a large output. Applied to work, the question becomes “How do I increase my output without increasing input as much?”

"Working longer can help, but eventually you hit a physical limit. A better question is, ‘How do I increase my output without increasing input as much?’ That is leverage." - Hyung Won Chung

The essence of leverage is breaking the linear relationship between input and output.

3. Evolution of Leverage Concepts

Naval Ravikant’s leverage theory, described in *The Almanack of Naval Ravikant*, classifies historical wealth‑creation levers into three categories:

Human Labor Leverage (Permissioned) : hiring people to multiply output (e.g., building the pyramids).

Capital Leverage (Permissioned) : using money to amplify decisions (e.g., real‑estate investment with leverage).

Code & Media Leverage (Permissionless) : software and media that can be copied at near‑zero marginal cost.

"Permissioned leverage" (labor and capital) requires approval from others.

"Permissionless leverage" (code and media) only needs a computer and skill.

Naval emphasizes that the new conflict is between leveraged and non‑leveraged activities, not between rich and poor.

4. AI as Permissionless Compound Leverage

Chung argues that an AI Agent fuses labor leverage and code leverage, creating a **Permissionless Compound Leverage**:

Labor function: the agent can autonomously perform tasks ranging from customer service to data analysis.

Code scalability: as pure software, the agent can be duplicated at near‑zero marginal cost without any additional permission.

This combination makes AI the first truly permissionless labor, breaking the historical need for permission in scaling work.

Impact of AI Compound Leverage

1. Individual Level – Learning and Skill Value

Geometric drop in learning cost: Generative AI can dynamically produce customized, high‑relevance learning material, virtually eliminating barriers to acquiring new knowledge.

Re‑definition of skill value: When knowledge becomes cheap, scarcity—not complexity—determines a skill’s worth. Curiosity and intrinsic motivation become the new scarce resources.

2. Team Level – Organizational Efficiency

AI Agents as wealth‑creation engines: By merging human‑labor and code levers, agents enable “super‑individuals” and small, highly productive teams to generate multi‑billion‑dollar revenues.

Mitigating collaboration bottlenecks: Traditional labor scales poorly due to communication and coordination costs; AI agents allow lean, high‑output teams or even one‑person enterprises.

3. Human Level – Accelerating Scientific Progress

New ideas and knowledge generation: Future AI could act as a “non‑stopping research engine,” continuously producing novel hypotheses and discoveries.

Knowledge convex hull: AI connects scattered expert points into a convex hull, uncovering low‑hanging fruit and creating new knowledge.

Extension of AI Compound Leverage

1. Digital Labor Economy

VCs have poured roughly $700 million into Agent‑focused seed rounds in 2025, representing about 50 % of global VC capital. Companies such as Salesforce (AgentForce), Sierra, HarveyAI, Decagon, Glean, OpenEvidence, Devin and Windsurf are building multi‑step, cross‑platform AI agents for finance, HR, healthcare, software development, and more.

Infrastructure is scaling in parallel: GPU manufacturers, cloud providers and model vendors are announcing orders for hundreds of thousands of GPUs (e.g., xAI’s Colossus 2 with 550 k cards).

2. New Rules: Productivity, Vibe Coding, and the Future of Programmers

AI’s rapid progress on code generation is reflected in SWE‑bench scores rising from 4.4 % in 2023 to 71.7 % in 2024. Andrej Karpathy’s “Vibe Coding” concept describes a paradigm where developers specify high‑level intent and AI produces the implementation.

Consequently, the most valuable human work shifts to defining problems, setting specifications, and critically evaluating AI‑generated solutions, while routine execution becomes automated.

Summary: Using the Leverage Lens

1. Competition Compresses Leverage Upside

"It's cliché to say AI will create massive wealth. But using this leverage lens lets us interpret the noisy AI news cycle in a consistent way and spot the real opportunities." - Chung

When a leverage source becomes popular, competition erodes its margin. The lifecycle of a new lever consists of:

Discovery phase: emergence of a new lever (e.g., AI agents).

Alpha (excess‑return) phase: early adopters capture outsized gains.

Mainstream phase: tools become widely available, lowering entry barriers.

Profit‑compression phase: widespread adoption dilutes advantage.

Commoditization: the lever becomes a table‑stake capability.

2. Practical Implications

To capture the remaining upside, individuals should identify newly feasible, under‑exploited levers and apply AI agents to automate end‑to‑end workflows (e.g., an AI that ingests a PDF financial report, extracts key metrics, queries market data, and generates a structured investment analysis).

Professionals must shift from “selling time” to “productizing knowledge” by packaging unique expertise into AI‑driven services that scale without additional human effort.

Rapid, low‑cost experimentation with emerging agents, Vibe Coding tools, and AI‑generated media helps build intuition about the boundaries of new technology, enabling earlier capture of future alpha opportunities.

Overall, AI is currently transitioning from the “excess‑return” to the “mainstream” stage of its leverage lifecycle, making now a strategic moment to adopt permissionless compound leverage.

References

Cornell AI History Lecture – “AI as an ultimate form of leverage” by Hyung Won Chung (https://www.youtube.com/watch?v=CcP8db8TeKI)

The Almanack of Naval Ravikant by Eric Jorgenson