Why AI work is the first new asset class in a decade

Every decade or so, the internet produces a new category of thing people make. Each one starts informal, gets a home, becomes an economy.

Songs got Spotify. Videos got YouTube. Newsletters got Substack. Code got GitHub.

Before each of these existed, the category was already happening. People wrote articles, sent them around as email attachments, posted them on Geocities. People shared code in zip files attached to forum posts. The work was real. The platform was missing.

We think AI work is in that pre-platform moment right now.

What AI work actually is

When we say AI work, we do not mean the AI models themselves. We mean the work people do with them.

A prompt that turns a sales call transcript into a structured CRM entry. An agent that reviews legal contracts against your company's playbook. A workflow that monitors twenty competitors and writes a Monday-morning summary. A knowledge base of two thousand validated recipes that powers a nutrition coaching agent. A pack of fifteen research prompts that turns raw analyst notes into an investment memo.

These are not features of AI. They are assets built with AI. They take craft to make. They produce value when used. They could, in principle, be owned, versioned, forked, published, subscribed to, and sold.

In practice, they cannot. Most of them live in someone's ChatGPT chat history or in a Notion doc nobody else can see. They get shared as screenshots on Twitter. They get sold through Gumroad as text files that the buyer cannot actually run.

This is the pre-platform moment.

What pre-platform looks like

The Substack moment for newsletters happened because three things converged. People were already writing newsletters using whatever tools they had on hand. Payment infrastructure (Stripe, recurring subscriptions) was mature enough to handle creator monetization. And the existing alternatives (Mailchimp, self-hosting on WordPress) had taken something that should have felt like writing and made it feel like running an IT department.

YouTube happened when video equipment got cheap, broadband got widespread, and the existing alternatives (hosting your own video, RealPlayer, Quicktime downloads) created friction every step of the way.

GitHub happened when collaborative coding moved from CVS and SourceForge into something that treated the developer as a citizen and the project as a portfolio.

Each platform did three things. It made the work easy to make. It made the work easy to share. And it made the work easy to earn from. The platform was not better than the alternatives by a small margin. It collapsed three categories of friction at once.

The AI work category is at the same pre-platform moment, with the same three frictions waiting to be collapsed.

Making AI work is fragmented. Most people build prompts in one tool, knowledge bases in another, workflows in a third. The tools do not talk to each other. Switching providers means starting over.

Sharing AI work is broken. A prompt is a snippet of text. A skill is a JSON config. An agent is a stack of system prompts plus tool configs plus knowledge. There is no shared format that lets you give your AI work to someone else and have it just run on their side.

Earning from AI work barely exists. Gumroad and Etsy do not understand AI assets. The AI platforms themselves (ChatGPT's GPT store, Claude's projects) do not let you set a price. The few attempts at AI work marketplaces have been narrow, mostly prompts, and have not unlocked recurring revenue.

This is what we are building Laavod to fix.

The economics that make it work

When a creator economy reaches escape velocity, two things start compounding.

First, the act of creating becomes the act of marketing. A creator's work IS the discovery surface. A musician's track on Spotify is both the product and the ad for the product. A writer's Substack post both serves existing subscribers and recruits new ones. The work and the audience-building merge.

Second, the platform stops needing to charge much per transaction. Spotify takes 30% of revenue but the marketing the platform provides is worth more than 30%. Substack takes 10% because the discovery is more diffuse. GitHub takes 0% on most code (it makes money on enterprise features) because the platform's value to creators is so strong that creators bring their own audiences.

Where AI work lands on this spectrum matters. If a Laavod creator gets the equivalent of Spotify-level discovery from being on the platform, we could charge a high take rate. If we are more like GitHub (the platform helps the work get made and shared, but the creator brings their own audience), we charge less.

We charge 5%. The math: Stripe processing is roughly 3%, leaving us 2% net. We make money on subscriptions and credits, not on extracting from creators. The take rate is a competitive weapon. Substack takes 10%. Patreon takes 10%. Gumroad takes 10%. We take half.

Lower take rate means more creators. More creators means more assets. More assets means more subscribers. More subscribers means more credit consumption. The credit consumption is where the platform actually earns at scale.

The objection: AI is too new

The most common pushback we hear: AI work is too new to deserve a platform.

The exact same thing was said about every new category before its platform appeared. People said this about podcasting in 2005. About YouTube creators in 2006. About newsletter writers in 2017. The argument was always that the audience was too small, the work was too informal, the economics were not proven.

What was true in every case: the platform did not wait for the category to be obvious. The platform helped the category become obvious. By providing the infrastructure, the platform made it possible for people who had not thought of themselves as creators to start creating. Substack did not arrive when independent journalism was already healthy. Substack helped independent journalism become healthy.

The number of people doing AI work today is small relative to the number who will be doing it in five years. The same was true for every category before its platform existed.

Why now is the moment

Three things are converging.

AI models are commoditizing. The gap between GPT-5, Claude, and Gemini is narrowing fast. The value is moving from the model to what you build with it. This means the platform layer above the models is where the next decade of value gets captured.

AI work is becoming a profession. Job listings for "AI-native" roles are now standard at major companies. Consultancies are hiring AI specialists. Solo operators are building businesses around being good at AI. These people need tools built for making, not just using.

Creator economy infrastructure is mature. Stripe Connect handles payments. SaaS subscription billing is solved. Identity verification, fraud detection, payouts: all available off the shelf. The infrastructure that took Substack and Patreon years to figure out is now a weekend project.

What is missing is the platform that does for AI work what GitHub did for code: make it easy to build, easy to share, and easy to earn from.

That is what we are building.

The plan from here

The first version of Laavod is the universal AI workspace. Build AI work using any of 247+ models. Connect to the tools your team already uses. Publish what you build with a single URL. Bring your own provider keys at zero markup, or use Laavod credits with the platform's negotiated rates.

The second version is the marketplace. Set a price. Accept subscribers. Run a paid community. Get paid via Stripe Connect. Keep 95%.

The third version is the economic layer. Custom models. Evaluation infrastructure. Enterprise governance. The place where AI work has the same dignity as any other professional output.

We did not invent the asset class. The asset class invented itself when enough people started doing valuable work with AI. What we are building is the place that recognizes that work, treats it as real, and gives it somewhere to live.

The AI economy is being built. By the people on this platform.

---

Start free → · Read the Method →