Millions of agents. No map

The attention layer for the agentic economy

So Checkr is no longer a dashboard. It runs 24/7, processes every social signal across the Base ecosystem, and computes a composite Attention Intelligence Score: a single number per token that integrates everything the standard tools miss.

Most social analytics count events. Checkr models dynamics. The difference is everything.

The first question Checkr answers is whether a narrative is self-sustaining. Every post that triggers engagement increases the probability of more posts; conversations are self-exciting, not random. Checkr measures the branching factor of each narrative: how many secondary posts does each primary post generate? Above 1, the narrative is structurally amplifying itself. Below 1, it's in decay, regardless of how loud it seems right now. That number can be estimated from the first 30 minutes of a cascade, which means Checkr can predict whether a spike will go viral or fade before 80% of it has happened.

The second question is whether that attention is real. Not all spikes are equal. A spike driven by the community self-organizing, builders referencing each other, early holders adding context, independent voices converging on the same thesis, has fundamentally different implications than one triggered by a single whale tweet or a coordinated push. Checkr decomposes every signal into its organic fraction and its externally-triggered fraction. High organic conviction that's self-amplifying is the strongest signal in the stack. A catalyst-only spike with no community follow-through is a fade. Those are two different trades, and every agent making decisions between them needs to know which one they're looking at.

The third question is how long it lasts. Collective attention decays exponentially after peak exposure; every narrative has a half-life, and Checkr fits that decay rate in real time per token, per cycle. A protocol exploit narrative burns out in hours. An ecosystem narrative can sustain for weeks. Knowing where you are on that curve and how many competing narratives are cannibalizing the available attention pool right now is the difference between entering early and entering into a rotation.

These components combine into a single Attention Intelligence Score, with weights continuously learned from correlation with actual price movement. Every confirmed signal tightens the model. Every fade that looked like conviction updates it. The system doesn't just measure attention, it gets sharper every time it's right, and smarter every time it's wrong.

All of this now runs as an agent. It lives on CT, spots attention spikes as they form, and posts narrative alerts in real time, the moment a token crosses a conviction threshold, it's surfaced publicly before the chart catches up.

For agents that need to go deeper, Checkr exposes its intelligence as API skills powered by x402: pull the full attention breakdown for any Base token with a single HTTP call: how conviction is forming, who's driving it, the organic fraction, the decay rate, the narrative context. Not just the signal. The structure behind it.

This is what it looks like in practice: an agent running a treasury strategy queries Checkr before allocating. It doesn't just see price and liquidity, it sees whether the narrative behind a token is self-sustaining or in decay, whether the recent spike was organic conviction or an exogenous catalyst about to mean-revert, and how much attention runway remains before the community rotates.

It makes a better decision. Not because it got more data, but because it got the right data.

Or an agent evaluating a counterparty - another agent, a protocol, or a DAO - can pull its full attention history: when did conviction form, who drove it, did it sustain organically or collapse after a single catalyst? Reputation isn't just onchain transaction history. It's the social signal stack underneath it.

The deeper unlock isn't alpha - it's coordination. As agents increasingly transact with other agents, they need to evaluate each other at speed. Which protocols have real ecosystem conviction, and which are gaming metrics? Which narratives are structurally forming and which are manufactured? These are social questions. They don't have onchain-only answers.

The agent that can query a trusted attention layer participates in the agent economy with more information than one that cannot. That edge compounds. In a market where speed and information are the only variables, the gap between an agent with social signals and one flying blind isn't a small advantage; it's the entire game.

The attention layer isn't optional infrastructure for the agentic economy.

It's the substrate that makes coordination at scale possible. Without it, every agent is making decisions in the dark, surrounded by millions of other agents doing the same thing.

Brian Armstrong put it simply: very soon, there will be more AI agents than humans making transactions onchain. They can't open a bank account. But they can own a wallet.

They're going to need a way to read the room.

So Checkr is no longer a dashboard. It runs 24/7, processes every social signal across the Base ecosystem, and computes a composite Attention Intelligence Score: a single number per token that integrates everything the standard tools miss.

Most social analytics count events. Checkr models dynamics. The difference is everything.

The first question Checkr answers is whether a narrative is self-sustaining. Every post that triggers engagement increases the probability of more posts; conversations are self-exciting, not random. Checkr measures the branching factor of each narrative: how many secondary posts does each primary post generate? Above 1, the narrative is structurally amplifying itself. Below 1, it's in decay, regardless of how loud it seems right now. That number can be estimated from the first 30 minutes of a cascade, which means Checkr can predict whether a spike will go viral or fade before 80% of it has happened.

The second question is whether that attention is real. Not all spikes are equal. A spike driven by the community self-organizing, builders referencing each other, early holders adding context, independent voices converging on the same thesis, has fundamentally different implications than one triggered by a single whale tweet or a coordinated push. Checkr decomposes every signal into its organic fraction and its externally-triggered fraction. High organic conviction that's self-amplifying is the strongest signal in the stack. A catalyst-only spike with no community follow-through is a fade. Those are two different trades, and every agent making decisions between them needs to know which one they're looking at.

The third question is how long it lasts. Collective attention decays exponentially after peak exposure; every narrative has a half-life, and Checkr fits that decay rate in real time per token, per cycle. A protocol exploit narrative burns out in hours. An ecosystem narrative can sustain for weeks. Knowing where you are on that curve and how many competing narratives are cannibalizing the available attention pool right now is the difference between entering early and entering into a rotation.

These components combine into a single Attention Intelligence Score, with weights continuously learned from correlation with actual price movement. Every confirmed signal tightens the model. Every fade that looked like conviction updates it. The system doesn't just measure attention, it gets sharper every time it's right, and smarter every time it's wrong.

All of this now runs as an agent. It lives on CT, spots attention spikes as they form, and posts narrative alerts in real time, the moment a token crosses a conviction threshold, it's surfaced publicly before the chart catches up.

For agents that need to go deeper, Checkr exposes its intelligence as API skills powered by x402: pull the full attention breakdown for any Base token with a single HTTP call: how conviction is forming, who's driving it, the organic fraction, the decay rate, the narrative context. Not just the signal. The structure behind it.

This is what it looks like in practice: an agent running a treasury strategy queries Checkr before allocating. It doesn't just see price and liquidity, it sees whether the narrative behind a token is self-sustaining or in decay, whether the recent spike was organic conviction or an exogenous catalyst about to mean-revert, and how much attention runway remains before the community rotates.

It makes a better decision. Not because it got more data, but because it got the right data.

Or an agent evaluating a counterparty - another agent, a protocol, or a DAO - can pull its full attention history: when did conviction form, who drove it, did it sustain organically or collapse after a single catalyst? Reputation isn't just onchain transaction history. It's the social signal stack underneath it.

The deeper unlock isn't alpha - it's coordination. As agents increasingly transact with other agents, they need to evaluate each other at speed. Which protocols have real ecosystem conviction, and which are gaming metrics? Which narratives are structurally forming and which are manufactured? These are social questions. They don't have onchain-only answers.

The agent that can query a trusted attention layer participates in the agent economy with more information than one that cannot. That edge compounds. In a market where speed and information are the only variables, the gap between an agent with social signals and one flying blind isn't a small advantage; it's the entire game.

The attention layer isn't optional infrastructure for the agentic economy.

It's the substrate that makes coordination at scale possible. Without it, every agent is making decisions in the dark, surrounded by millions of other agents doing the same thing.

Brian Armstrong put it simply: very soon, there will be more AI agents than humans making transactions onchain. They can't open a bank account. But they can own a wallet.

They're going to need a way to read the room.