In my last piece, I argued that AI is fundamentally an arbitrage-collapsing technology. Every market is a supply/demand matching problem. Between buyer and seller sits a long chain of intermediaries who exist purely to resolve informational arbitrage. AI collapses that arbitrage the same way market makers do in capital markets: it finds the spread, closes it, until price converges toward marginal cost.
That's the macro thesis. But there's a sharper, more concrete way to see what's actually happening — and it comes from an unlikely place: DeFi.
The DeFi Intents Revolution
There's been a profound architectural shift in how DeFi transactions are conceived. The old model was imperative: "Swap this token, on this DEX, through this route, at this slippage." The user had to understand the mechanism. One wrong parameter and you paid it.
The new model is declarative. You express an intent.
"I want to end up with X amount of Token B, at the best possible price, by any means necessary."
You sign the intent and broadcast it. A competitive market of solvers — autonomous agents who specialize in execution — race to fulfill your intent optimally. The best solver wins and gets a fee. Protocols like CoW Protocol, UniswapX, and Anoma are built entirely on this architecture.
The key innovation isn't the technology. It's the separation of "what I want" from "how it gets done." The user never touches the mechanism.
Every Business is an Intent
Now zoom out. Every real-world and online business opportunity at its core is also just an expression of intent:
"I need a developer to build a landing page by Friday under $500."
"I want to reach 10,000 decision-makers in HR who fit this ICP."
"I want to hire the best candidate for a growth role in a Series A fintech."
In each case, the intent is clear. The desired outcome is articulable. What has been expensive and slow is the solver layer: the recruitment agency, the media buyer, the growth consultant. They are the solvers, and they've been extracting rents because becoming a good solver required expensive human capital — relationships, expertise, time.
AI is now the universal solver. And critically, just like in DeFi, when you have many competing solvers who can all read the same data and execute on the same rails, the solver margin collapses to near zero. The rent extraction disappears.
The Three-Layer Structure
When you see AI through the intents lens, a clean three-layer architecture emerges for every market that gets disrupted by it.
Users, businesses, and increasingly other AI agents express desired outcomes in natural language or structured goals. No need to specify the "how." This is where the value to the end user lives — and where the defensible moats will form.
Autonomous agents compete to fulfill intents. They have access to APIs, databases, execution rails, and capital. The best-fit solver wins. Margins here are structurally thin — just like MEV bots in DeFi, the edge lives in milliseconds and marginal efficiency gains. This layer commoditizes fast.
The underlying assets and services being matched: SaaS tools, labor marketplaces, ad inventory, logistics networks, financial instruments. These become increasingly commoditized as solvers arbitrage across all of them simultaneously.
Distribution as the Concrete Case
Let me bring this out of the abstract with the market I know best: distribution.
The conventional assumption is that more reach equals more value. It doesn't. Blind amplification creates noise, not signal. It fails because it ignores the most important variable in any market transaction: relevance.
Redefining "Creator"
When I talk about creators, I mean anyone who holds distribution — which is far broader than the influencer framing most people default to. Community managers who hold trust within a specific niche. Newsletter writers with domain-specific readership. Forum moderators, Discord admins, bloggers with organic traffic, podcast hosts with loyal listener bases.
What unites them isn't follower count. It's relevance + trust within a domain. A 2,000-person community of CTOs is infinitely more valuable to a B2B SaaS product than a 200,000-person general tech audience.
The inefficiency today? Builders can't find them. Creators don't know the opportunity exists. That's a pure discovery gap — and discovery gaps are exactly what AI closes.
The Practical Model
A builder launches a Community Affiliation Campaign with defined reward mechanics: X% of revenue, token allocation, or cash for driving measurable growth.
Not just any creator — one whose audience matches the product's ideal user profile. Domain-specific. High-relevance. Trust-aligned. The solver does the work that previously required an agency and three weeks.
Promote this product, drive results, earn a share of the upside. Not a flat fee for posting — an outcome-linked earning opportunity. Skin in the game changes the nature of the promotion entirely.
A builder who needed distribution found it. A creator who had distribution monetized it. Both sides captured value they couldn't find alone. The intermediary layer extracted nothing.
What Survives the Compression?
In DeFi intents, the breakthrough wasn't smarter users — it was abstracting the user away from execution entirely. The user doesn't need to understand AMMs, liquidity pools, or MEV. They just state what they want.
AI does exactly this for the real economy. The solver layer gets commoditized. The intermediary layer gets disintermediated. So where does durable value accumulate?
The same things that hold value in efficient financial markets:
The intents themselves — people who understand what to ask for and how to frame the right goals.
Proprietary supply — data, relationships, and audience trust that solvers can't easily access or replicate.
The trust layer — reputation, identity, and skin in the game that makes intents credible and outcomes accountable.
The solver will be commoditized. The intents — and the trust behind them — will be the scarce resource.
DeFi taught us that separating intent from execution unlocks massive efficiency gains. AI is about to do the same for every market in the real world.
The question for every business isn't "how do I compete with AI?" It's: which layer am I building in, and do I understand what survives?