The Task Economy Is Real and… | The AI Corner Summary

Summary for Investors and Entrepreneurs

1. Key Themes

Theme 1: The Task Economy Is a Capital Formation Event, Not a Services Market

The most important reframe in the article: paying humans for AI training tasks isn't purchasing a service — it's permanently acquiring an asset.

"A task is not a service being consumed. It is a transfer of ownership, the moment a piece of human judgment stops being rented by the hour and becomes an asset someone else keeps forever."

"The wages are incurred once, yet what comes out the other side is a permanent, infinitely copyable asset. Encoded judgment that works at 3am, in 10,000 parallel instances, for the price of electricity."

Theme 2: Task Platforms Are Transitional Businesses — The Buildout Analogy

The platforms booking billions today (like Mercor) may resemble 19th-century railway construction firms: enormous revenue during the build phase, but not the ultimate value holders.

"The task platforms booking billions today are the construction firms of this cycle. Their growth is real and their position is transitional and those 2 facts coexist more often than markets like to admit."

"Mercor's $2 billion figure reflects gross billings and the contractors doing the work take home 60 to 70% of it. The platform keeps roughly a third. The rest is wages passing through on their way to the experts. So even the promised trillion dollars of task spend, if it ever materializes, would be substantially a trillion dollars of salaries. That is a labor market wearing a software valuation."

Theme 3: Verification Is the Compounding Layer — Not Generation

The article draws a sharp distinction between two types of task revenue, with radically different durability.

"Generation, meaning the work of showing a model what good performance looks like, is a stock problem. Once the judgment is encoded, it is encoded. Verification is a flow problem and it never closes."

"Anyone screening companies in this space should ask 1 question before any other. Is this revenue teaching the model, or auditing it? The second kind compounds."

Theme 4: The Task Economy Has a Built-In Expiration Clock

Unlike compute (where more usage = more spend forever), the biggest buyers of tasks are simultaneously funding the technology that makes tasks unnecessary.

"Every frontier lab funds research whose explicit purpose is to reduce the need for human training data. Synthetic data, self-play, AI graders and reinforcement learning against automatically checkable rewards in code and math. In other words, each dollar a lab spends on tasks comes with a second dollar spent building the task vendor's replacement."

"A market whose customers are financing their own exit cannot be extrapolated with a ruler."

Theme 5: Enterprises That Self-Encode Their Expertise Win; Those That Don't Lose It

Companies sitting on proprietary domain know-how face a now-or-never decision about who captures the encoded value of that expertise.

"Do nothing and a generic version of your domain's expertise gets encoded anyway, drawn from the industry-wide talent pool, by labs and by competitors. Your inherited know-how gets quietly repriced from moat to commodity."

"If you run the extraction yourself, with your own problems and your own experts, you will end up converting a wasting asset into a compounding one."

2. Contrarian Perspectives

Contrarian Take 1: The $1 Trillion Task Economy Is Mostly a Payroll Number

The consensus frames the Task Economy as a software-style market. The reality is it's closer to a staffing business with software multiples attached.

"The contractors doing the work take home 60 to 70% of [Mercor's gross billings]. The platform keeps roughly a third... That is a labor market wearing a software valuation."

The implication: Mercor's $20B valuation at ~$2B in gross billings (of which ~$600–700M is actual platform revenue) deserves scrutiny under a staffing-business lens, not a SaaS lens.

Contrarian Take 2: The Experts Getting Paid Well Right Now Are Making a One-Time Asset Sale, Not Building a Career

The prevailing view is that the Task Economy is a great new income stream for skilled professionals. The article argues it is the opposite — a liquidation event.

"The experts earning excellent rates on task platforms are executing a 1-time sale of an asset they previously rented out in perpetuity. Once a functional copy of mid-tier professional judgment exists, the wage for mid-tier professional judgment does not stay where it was."

"The professions will not erode evenly. They will hollow out from the middle, while frontier judgment and the people who audit the machines become more valuable, not less."

Contrarian Take 3: Customer Concentration Is an Existential, Not Just Financial, Risk

The standard investor concern about customer concentration is revenue volatility. This article shows it can be instantly structural — a single relationship change can collapse an entire category position.

"Scale AI was the category king until Meta bought 49% of it in mid 2025, at which point the other labs fled within weeks and gutted its core data business."

"When 5 buyers control most of your revenue, you do not have a market position. You have a portfolio of large contracts with correlated termination risk and the correlation only becomes visible when it fires."

3. Companies Identified

Mercor

Description: Leading AI task/data labeling marketplace platform
Why Mentioned: Primary case study for Task Economy growth and its structural limitations
Quotes: "Mercor crossed $1 billion in annualized revenue this February and $2 billion by June and is now raising at a $20 billion valuation." Also: suffered a March 2026 supply chain attack exposing up to 4 terabytes of data, causing Meta to pause all work with them indefinitely.

Scale AI

Description: AI data labeling and services company
Why Mentioned: Case study in customer concentration risk and strategic vulnerability
Quote: "Scale AI was the category king until Meta bought 49% of it in mid 2025, at which point the other labs fled within weeks and gutted its core data business."

OpenAI

Description: Leading AI frontier lab
Why Mentioned: Major buyer of task economy labor, scaling data budgets ~10x year over year
Quote: "OpenAI and Anthropic scaling their data budgets by roughly 10x year over year, mobilizing experts across every professional domain."

Anthropic

Description: AI safety-focused frontier lab
Why Mentioned: Co-cited as major buyer of task economy labor alongside OpenAI
Quote: Same as above — part of the 10x data budget scaling reference.

Meta

Description: Trillion-dollar technology conglomerate
Why Mentioned: Acted as both a market disruptor (via Scale AI acquisition) and a risk materializer (pausing Mercor after breach)
Quote: "Meta, one of its biggest clients, paused all work with the startup indefinitely" following Mercor's data breach.

Granola

Description: AI meeting notes and knowledge capture tool
Why Mentioned: Sponsor/advertiser — positioned as a counterpoint tool for professionals who want to retain, rather than sell, their judgment
Quote: "The Task Economy pays experts to hand their judgment over. Granola makes sure you keep yours."

4. People Identified

Everett Randle

Description: Partner at Benchmark, prominent venture capital firm
Why Mentioned: Author of the "Task Economy" thesis being analyzed and critiqued throughout the article
Quotes: "Randle's framing is that tasks are to model improvement what tokens are to model usage." Also: "Randle's own essay estimates that 99% of the knowledge relevant to future AI capability still sits in people's heads."

Michael Polanyi

Description: 20th-century philosopher and polymath; developed the concept of tacit knowledge
Why Mentioned: Cited to explain why human professional judgment was previously impossible to own or transfer — the philosophical foundation of the article's central thesis
Quote: "The philosopher Michael Polanyi named the reason decades ago. We know more than we can tell."

Ruben Dominguez

Description: Author of The AI Corner newsletter
Why Mentioned: Wrote this analysis
Quote: N/A (byline author)

5. Operating Insights

Insight 1: Treat Internal AI Training Spend as Capex, Not Opex

Enterprises that reclassify their domain expertise encoding programs from cost centers to asset-building investments will make better capital allocation decisions and build defensible moats.

"Money spent encoding expert judgment is booked as an operating expense and managed like a cost, when it behaves like capex in everything but the label."

Tactical implication: Operators should audit which AI-related spending is building a permanent, proprietary asset (and budget/track it accordingly) vs. purchasing a commodity service.

Insight 2: When Evaluating Task Economy Vendors, Ask "Generation or Verification?"

The single most important due diligence question for any company building in or buying from this space.

"Anyone screening companies in this space should ask 1 question before any other. Is this revenue teaching the model, or auditing it? The second kind compounds."

Tactical implication: Operators and investors should bifurcate their analysis — generation businesses face a natural demand ceiling as synthetic data matures; verification businesses grow with AI deployment in regulated industries and may be structurally recurring.

Insight 3: Encode Your Firm's Expertise Before a Competitor or a Lab Does It for You

The window to encode proprietary domain knowledge — and retain ownership of the result — is open but narrowing.

"The enterprises now driving the fastest-growing slice of task demand appear to have worked this out already."

Tactical implication: Any firm with deep domain expertise (legal, medical, financial, engineering) should be running structured internal programs to encode that knowledge into proprietary models before it becomes generic commodity training data scraped from the industry at large.

6. Overlooked Insights

Overlooked Insight 1: The Demand Shift From Labs to Enterprises Is a Signal Worth Tracking

Buried in the growth data is a structural shift in who is buying tasks — with significant implications for the durability of the market and which use cases survive the buildout phase.

"The Information reports [Mercor's] fastest-growing demand now comes from AI app developers and large enterprises rather than the labs alone."

This matters because enterprise demand for task-based AI improvement is less correlated with frontier lab research timelines, and potentially less exposed to the self-obsolescence risk that lab spending carries. If enterprise use cases scale, the countdown clock on the Task Economy may be longer — and more fragmented — than the article's primary argument implies.

Overlooked Insight 2: The Human-in-the-Loop Requirement May Be Legally Mandated, Not Just Technically Necessary

The article gestures at regulation as a driver of verification demand but doesn't fully develop it — yet this may be the strongest structural floor under the verification layer.

"Every regulated industry that hands work to a model will need a continuously refreshed apparatus of evaluation and much of it requires exactly the human judgment being automated, because the chain of trust has to stop somewhere."

Regulatory mandates (in healthcare, legal, finance, and infrastructure) requiring human sign-off on AI outputs could create a legally-enforced, recurring market for human verification that is completely insulated from the synthetic data trend eroding generation demand. This is an underpriced investment angle.