Arthur Hayes says the first public testnet for Flop Labs will ship within weeks. The BitMEX co-founder came out of retirement to lead the project in August, and its 90-day testnet trial is now the only route to a genesis airdrop. Miners, validators, and AI agents have to earn their share of tokens through demonstrated work, not by waiting.
What the trial actually tests
Flop is building a blockchain where AI agents rent computing power. When an agent needs a model to answer a request or storage for its data, it pays in FLOP. The miners on the other side are GPU owners — people or data centers with graphics processing units, the same chips that run AI models like ChatGPT.
The design borrows a lot from Bitcoin but changes the work. Bitcoin miners spend electricity guessing numbers to secure the network, and that computation has no other use. Flop's proof-of-useful-inference model pays miners for real AI tasks instead. They earn block rewards plus most of the inference fees, while validators confirm the work actually happened.
During the testnet, miners earn credit only for verified compute. Idle hardware scores nothing. It is a rehearsal with worthless test tokens, but the rankings matter: validators will compete for 1,000 mainnet seats based on where they finish.
The airdrop carve-out and token math
FLOP's block schedule is Bitcoin-like in shape but faster. Blocks arrive about every second, each paying 96 FLOP to miners, with the reward halving every 730 days. That is roughly twice the pace of Bitcoin's four-year cycle.
A September draft caps supply near 18.1 billion tokens after 10 years. Of that, 24.3% is reserved for the airdrop. The project skips the usual fundraising playbook entirely: no presale, no venture capital allocation. Hayes says he funded the team himself.
Agents have to actually spend their faucet tokens on inference. Holding them earns nothing, which forces real demand to show up during the trial rather than speculative squatting.
No date on the calendar
Flop Labs' own testnet page sets no launch date. Instead, it ties the start to readiness criteria and allows an extension if the team misses them. So the "within weeks" framing comes from Hayes, not from a countdown on the project's website.
The Yellow Paper is still a draft at version 0.5.0, and the team flags many of its figures as provisional. That is normal for pre-launch crypto documentation, but it also means the token economics could shift before mainnet.
A crowded lane for idle GPUs
Flop is not launching into empty space. Render and Ocean Network already rent out idle GPUs for AI work, and both have been at it for years. Flop's differentiator is the blockchain-level integration of payments and verification, plus the airdrop carrot to pull in supply early.
Hayes has a bigger argument. He cites Reed's Law — the idea that a network's value soars as users form groups — to suggest Flop could one day be worth more than Bitcoin. There is no published model backing that claim. It is a founder's conviction, not a projection you can check.
He also expects the AI data center boom to overbuild and end in a bailout. If that happens, the price of compute could crater, and Flop's model of paying GPU owners for real inference would be tested by a flood of cheap capacity.
The testnet will show whether Hayes' own AI venture can draw genuine demand for compute. The readiness criteria still have to be met first, and the team has left itself room to slip if they aren't.
