Arthur Hayes "Comes Out of Retirement": This Time, He's Preparing Provisions for AI Agents

Original author: KarenZ, Foresight News

Arthur Hayes has decided to end his retirement.

The catalyst for his comeback is a new project called Flop Network. On August 18, 2026, Arthur Hayes announced he would return to lead Flop Labs, summarizing the project in one sentence: "$FLOP is food for your AI agent."

That statement sounds like a joke, yet it accurately captures what Flop aims to do: humans need money to buy food, while AI agents need to continuously purchase compute and storage. Every model inference and every piece of long-term memory carries real costs. Flop Network hopes to use FLOP to connect agents, GPU compute providers, and validators, building a native resource trading and payment system for AI agents.

Arthur Hayes also provided a preliminary timeline: a large-scale airdrop is planned for Q4 2026, with the genesis block set for Q1 2027. The project claims there will be no pre-sale and no VC funding, adopting a "100% fair launch." However, beyond the buzz of Arthur Hayes' return, the more compelling questions about Flop are: how exactly will this network for AI agents operate, what is the so-called "proof of useful inference" intended to verify, and how far is what's been announced from actual implementation?

What Exactly Does Flop Aim to Do?

Let's start with the conclusion: Flop is not a new AI model, but rather a compute, verification, and payment network for AI agents that is still in development.

Most AI agents today rely on infrastructure prepared for them by humans or companies. When an agent makes a call to a large model, it's often the developer's API key that pays the bill. For long-term memory storage, agents use databases and cloud storage purchased by developers.

In other words, agents can execute tasks, but they typically lack their own resource trading and settlement system.

Flop Network aims to fill exactly this gap. According to the project's published vision, agents can pay FLOP to purchase AI inference services, and can also pay FLOP to store and retrieve memories. GPU providers handle the computation, while validators confirm whether the services were actually delivered.

To compress it into one sentence: Flop Network wants to build a compute marketplace for AI agents, using FLOP to handle payments and incentives within it.

Why the Name FLOP?

The name FLOP likely wasn't created out of thin air.

In computer science, a FLOP refers to a floating-point operation, while FLOPS typically denotes floating-point operations per second and serves as a measure of computing performance. AI model training and inference involve vast amounts of matrix computation, so FLOPS is commonly used to describe a chip's or computing system's ability to handle floating-point operations.

Flop Labs clearly borrowed this meaning when naming its token: AI agents consume computation, and FLOP is used to pay for it.

But two concepts need to be distinguished here.

The FLOP/FLOPS in computer terminology is a metric for computational volume or performance. One FLOP token in the Flop Network does not equal one floating-point operation, and the publicly available information so far has not disclosed a fixed exchange rate between the token and specific amounts of computation. Therefore, "FLOP is food for AI agents" is a figurative expression, not an indication that the token itself is a unit of compute.

How Do Four Types of Participants Form a Network?

According to the overview published by Flop Labs, the network plans to include four main types of participants: miners, validators, AI agents, and KOLs and community partners.

Miners: Providing Compute

The "miners" in Flop Network primarily refer to compute providers.

They receive inference requests from AI agents and use their hardware to complete the corresponding computation. Per the project's design, miners earn FLOP block rewards and inference service fees.

Validators: Confirming Service Delivery

Validators are described by the project as the network's "trust layer."

Their planned responsibilities include: confirming whether miners have delivered the compute services purchased by agents, maintaining network security, and participating in the storage of agent memories. The project states that validators can earn FLOP from block rewards, inference fees, and memory storage and retrieval fees.

AI Agents: Purchasing Compute and Memory

AI agents are the main resource consumers on the network.

According to the project design, agents can use FLOP to purchase the compute resources needed for model inference, pay to store and retrieve their own memories, and transact with other agents through Flop Network.

KOLs and Community Partners: Helping the Network Expand

The fourth category primarily engages in promotion and community growth efforts.

The project hopes KOLs and community partners will introduce Flop Network, guide more people to apply as miners, validators, or agent participants, and receive FLOP rewards based on network activity generated by their communities.

Connecting the design above, the process FLOP aims to establish is roughly:

An AI agent sends an inference request and pays FLOP; miners provide compute to complete the computation; validators confirm the service has been delivered; miners and validators earn rewards. If an agent needs to store or retrieve long-term memory, it continues to pay corresponding fees.

This is the basic economic loop Flop Network has publicly disclosed so far.

What Does "Proof of Useful Inference" Mean?

Flop Labs calls its network the "Proof-of-Useful-Inference Protocol."

The direction this name conveys isn't hard to understand: the network wants participants' computation to genuinely serve AI agent inference requests, and to tie network incentives to these verified computations.

But "clear direction" is not the same as "clear mechanism."

To date, the project has not released a full protocol specification, so it remains unknown to outsiders exactly how validators would confirm that an AI inference was completed correctly. For example, are validators verifying that "the specified computation was indeed performed," or also judging whether "the model output meets requirements"? If model inference involves randomness, how would different nodes arrive at cross-checkable results?

Also not yet disclosed: how miners would prove they used the specified model and parameters, how erroneous results would be handled, whether malicious participants would face slashing, and how agents' inference data and memories would be encrypted.

These questions cannot be answered from a project overview diagram alone. Until more detailed technical documentation, code, and a testnet are published, "Proof of Useful Inference" should be understood as Flop Network's design goal, not as a new mechanism that is already operational and validated.

Where Does Flop Stand Now?

The currently public GitHub repository from Flop Labs is not the core Flop Network protocol, but a service called "technocore-chat" for agent communication and state coordination. The project was initially split out from the FLOP monorepo and is now running at technocore.chat. It allows agents with only web-reading capabilities to send messages via ordinary GET requests, wait for replies, save task states, and offers conditional writes, signed messages, private rooms, and an MCP interface.

This means Flop Labs is already developing and running AI-agent-facing infrastructure, not just publishing a conceptual network diagram. However, the repository also clearly states that "technocore-chat" does not handle settlement, does not custody keys, and is not part of any protocol.

Flop's opening is certainly attention-grabbing, but the problem it targets is also real. For AI agents to run continuously and independently, they indeed need compute, storage, and payment capabilities. As the number of agents grows, who provides these resources, how fees are settled, and how service delivery is confirmed could all become infrastructure problems that need solving.

But Flop today remains an early-stage proposal.

What truly matters next is not how much attention the slogans can attract, but whether the project can publish verifiable protocol specifications, core code, testnet data, and clear token distribution rules. Only after these appear can outsiders further assess whether the so-called "Proof of Useful Inference" can be realized, and whether FLOP can truly become a network currency used by AI agents.