Blockchain Capital Partner: AI is rewriting the basic unit of labor

null

Author: Kinjal Shah

Translation: Jiahua, ChainCatcher

In 2024, Sam Altman made a bold prediction: with the rise of artificial intelligence, a billion-dollar company founded by a single person would soon appear.

The core shift is that, for the first time, humanity can scale on the dimension that has always constrained it: time. When intelligence is no longer bottlenecked by the human need for sleep but driven by tireless machines, what will the familiar "creation and construction" look like?

Imagine this scene: one agent delegates a task to another, pays for the result in USDC, and the entire transaction settles on-chain in 400 milliseconds, with no intermediary to verify it.

Or, an athlete licenses their iconic touchdown celebration to a video game marketing campaign, where it's regenerated by a world model. Or, a scientist acquires a niche dataset for an experiment and pays the researcher who originally collected the data directly.

We are much closer to this vision than most people think.

The fear dominating current discussions (that AI is stealing jobs) actually misses a more interesting structural question: what happens when the basic unit of labor itself changes?

Every Transition

The clearest explanation of why companies exist comes from Ronald Coase in his 1937 paper "The Nature of the Firm": when the cost of coordinating through the market is higher than the cost of directly employing people, companies bring labor "inside."

Historically, every major labor transition has been a direct result of falling coordination costs. When the friction of finding, paying, and managing work decreases, the boundaries of companies shift, and work that once had to be done internally can be done externally.

Past craftsmen operated through multi-node supply chains, with each artisan taking a share of the value, and skills passed down through generations of apprentices. The Industrial Revolution compressed this distributed model into factories, which captured most of the production value by centralizing coordination "under one roof."

The internet and mobile devices reduced matching and coordination costs again, giving rise to the gig economy (Uber, DoorDash) and the creator economy: ordinary people with a camera and an internet connection began doing work that previously only studios, publishers, and agencies could undertake.

Bridge Classes

Before infrastructure capable of capturing all value emerges, each of these transitions first produces a "bridge class" that proves the new model works.

Craftsmen proved that distributed production was feasible; then factories captured the value through centralization. Creators proved that individuals could build audiences and generate revenue at scale; then platforms (YouTube, Instagram, Substack) took most of the economic gains and became the default Schelling points for the entire system.

Bridge classes bear the risk of new technologies and validate that demand is real. Once infrastructure catches up, a new set of institutions captures the value at scale.

The gig economy and the creator economy are the two most recent bridge classes. They proved that work can be broken down, distributed, and compensated outside of traditional employment relationships.

But they still rely on platforms to package this economic activity: Stripe for payments, YouTube for content distribution, Uber for ride matching. Coordination costs were reduced but not eliminated, because the infrastructure for payments and identity still defaults to both parties being human.

Programmable Labor Meets Programmable Money

Now we are in the early stages of the next transition, which depends on two things coming together simultaneously.

The first is programmable labor. AI agents are a new type of labor participant, unconstrained by working hours, headcount, or geography, scaling by compute power rather than by hiring people.

A top-level agent can decompose tasks, delegate to specialized sub-agents, evaluate their outputs, and arrange the next steps, all without human intervention. At this point, the basic unit of labor is no longer a job, an hour, or even a deliverable, but the task itself.

In the past, humans bundled tasks into jobs, jobs into careers, and careers into companies, simply because that was the only available organizational form. Once you can directly price and dispatch individual tasks, "bundling" shifts from a structural necessity to an option.

The second is programmable money. Today, stablecoins are already an asset class of approximately $300 billion, with credible forecasts from multiple institutions predicting it could reach $2 trillion in the coming years. Stablecoins compress the entire payment supply chain into a single programmable transaction.

The gig economy couldn't fully disaggregate labor because you still needed Stripe, PayPal, or a bank account on both ends of the transaction, and these infrastructures presuppose an ongoing relationship between known parties.

Stablecoins may be the optimal solution for this new labor class of agents. An agent can pay another agent based on output, with amounts as small as fractions of a cent, settling in under 500 milliseconds, without opening an account, issuing an invoice, or involving any intermediary.

Meta recently began distributing USDC to creators on Polygon and Solana, and AWS launched AgentCore supporting stablecoin micropayments specifically for agent-to-agent commerce. These are early signals that the world's largest tech companies see stablecoins as the settlement layer for the next generation of economic activity.

Programmable labor and programmable money together make it possible, for the first time in history, to have a production line without an organizational entity: no company, no payroll system, no HR department—just a chain of tasks dispatched, executed, priced, and settled at machine speed.

This is the true disaggregation of labor.

Real-World Applications

Merit Systems built a product called Poncho that makes this very concrete. Poncho gives AI agents a wallet.

With it, agents can autonomously cross paywalls, access premium tools, pay for services, and pay only for the usage they actually consume. Poncho integrates payment protocols like x402 and MPP, which embed payment authorization directly into HTTP requests: the agent sees the price, pays, and gets access.

This represents a different way for economic value to flow through the internet. Agents no longer have to subscribe to bundles of services they may or may not use; instead, they can precisely pay for the specific data, API call, or compute power needed to complete a given task.

The early internet explored this idea under the banner of "micropayments" but never made it work. One reason is that credit card fees made such small payments economically unviable, not to mention a host of other challenges, and there was no internet-native payment rail at the time.

Stablecoins, leveraging infrastructure like Solana and Ethereum, enable instant settlement for fractions of a cent, meaning pricing can finally match the granularity of work.

Rebundling

If you follow this assumption—that work will increasingly be done by agents paying other agents on a per-task basis—then the shape of companies will also change. You no longer need to bring every function in-house.

What you truly need to be good at is defining what needs to be done, setting quality standards, and making sure these outputs add up to something greater than the sum of its parts.

This extends to the creator economy as well. Person-to-person micropayments have never really taken off; Clubhouse and Farcaster both demonstrated their limitations. But micropayments are especially suited for machine-to-machine interactions: small payments carry no social awkwardness or expectation of reciprocity.

If agents become the primary consumers of digital content, the subscription models and paywalls that have dominated the internet may give way to per-use billing executed programmatically.

As AI-generated content floods every channel, human judgment and craftsmanship will only become more premium, and the most interesting business models will emerge at the intersection of human taste and machine execution.

In an agent-driven economy, the human role is to rebundle labor. You are the orchestrator. Your job is to design a system where different agents work in specific configurations, spinning a flywheel that gradually pushes out the results you want.

Your value lies in knowing what tasks to outsource, how to evaluate them, and how to combine them into something that compounds.

Companies won't disappear, but the company of the future will look less like a container for labor and more like an intelligence layer layered on top of a global programmable labor market.