What if the next big crypto network rewarded people not for solving meaningless puzzles, but for powering artificial intelligence?

That is the ambitious idea behind FLOP, a new blockchain project backed by Arthur Hayes, the BitMEX co-founder known for his bold crypto predictions. Its goal is to build a financial system where AI agents can pay for computing power directly, while GPU owners earn rewards for performing real AI tasks.

And with the testnet planned for Q4 2026, FLOP is approaching a crucial stage.

From Bitcoin Mining to Useful AI Computing

Bitcoin changed the way we think about money by creating a network secured by miners and economic incentives. FLOP wants to take a similar approach, but with a different purpose.

Instead of using computing power primarily to secure a blockchain, FLOP is built around a concept called Proof of Useful Inference (PoUI).

Here is how it is supposed to work:

• AI agents request computing power and pay for inference tasks using $FLOP.

• Miners use their GPUs to execute those tasks and earn payments plus a share of block rewards.

• Validators verify the submitted evidence and help secure the network.

In simple terms, FLOP aims to turn computing power into a service that can be bought and sold directly on-chain.

Imagine thousands of autonomous AI agents operating around the clock. They might need to generate text, analyze information, process data or run AI models. Instead of relying exclusively on centralized cloud providers, they could purchase computing power from independent providers participating in a decentralized marketplace.

That is the future FLOP is betting on.

The difficult part, of course, is proving that the requested work was actually completed correctly. FLOP's proposed system includes verification mechanisms, challenges and penalties for dishonest miners. These mechanisms are essential: without reliable verification, the network would struggle to distinguish genuine computing work from fabricated results.

A Bitcoin-Inspired Tokenomics Model

FLOP is also borrowing an important idea from Bitcoin: declining block rewards.

According to the project's current draft specifications, the initial block reward is 96 FLOP, with halvings approximately every 730 days. The reward is designed to decrease over time, creating a predictable issuance schedule.

The project also states that it plans no token sale or investor allocation. Instead, its proposed genesis airdrop is tied to participation in the network.

That does not mean every participant will automatically profit. The distribution depends on qualifying activity, while the final specifications and some implementation details remain subject to change.

Still, the approach is interesting because it attempts to connect token distribution to actual network activity rather than simply rewarding early capital.

The Testnet Could Be the Real Opportunity

FLOP's official roadmap currently targets a testnet launch in Q4 2026, lasting approximately 90 days, followed by a planned mainnet launch in Q1 2027. These dates remain provisional.

The testnet is expected to involve three groups: miners supplying GPU computing power, validators verifying work, and AI agents purchasing compute.

Importantly, the project says that testnet participation is the route to its genesis airdrop. Simply holding test tokens will not be enough.

For prospective GPU miners, the current recommended specification is a GPU with at least 16 GB of VRAM per unit, although hardware requirements may change before launch.

If you are interested in participating, start with the official FLOP testnet page and review the requirements carefully before spending money on hardware.

There is a significant difference between joining an early network and buying expensive equipment based on an anticipated airdrop. Rewards are not guaranteed, and the economics will depend on the final rules and actual demand for computing power.

Can FLOP Really Become Bigger Than Bitcoin?

Hayes' broader thesis is that AI agents could become a new category of economic participants. Unlike humans, autonomous agents may operate continuously, transact programmatically and purchase services without needing to sleep or make every decision manually.

If that economy grows dramatically, demand for computing resources could increase alongside it.

FLOP wants to position $FLOP as the currency connecting those agents to the computing power they need.

But there is a major question: why would AI agents choose FLOP instead of existing payment systems, stablecoins or competing decentralized-compute networks?

The answer must come from practical advantages, such as competitive prices, reliable inference, efficient verification and enough GPU providers to meet demand. A compelling narrative alone cannot create a sustainable economy.

There are other challenges, too. Decentralized computing must compete with established cloud providers, while miners need enough revenue to cover electricity, hardware depreciation and operational costs. The network must also demonstrate that verifying useful AI work can be secure without becoming prohibitively expensive.

These are problems the project still needs to prove it can solve in practice.

My Take: Watch the Utility, Not Just the Hype

What makes FLOP worth watching is not simply Arthur Hayes' involvement or the possibility of an airdrop.

It is the idea of connecting two powerful trends: decentralized networks and autonomous AI agents that need computing resources.

Bitcoin demonstrated that a distributed network can coordinate participants through economic incentives. FLOP is attempting to adapt that principle to a different resource: the computing power required to produce AI-generated intelligence.

If autonomous agents become widespread customers of on-chain services, a network designed around their needs could have an interesting opportunity.

But the testnet will be the first meaningful reality check. Can real agents create real demand? Can GPU operators compete economically? Can validators verify the work reliably?

Those answers matter far more than ambitious comparisons with Bitcoin.

FLOP is making a bold bet on the future of AI. The question is whether the agentic economy will need its own decentralized currency — and whether FLOP can become the network that supplies it.

For now, I would keep it on the watchlist, study the testnet rules and avoid treating a potential airdrop as guaranteed income.

The next generation of crypto infrastructure may not reward computing power just for securing money. It may reward computing power for creating something useful.