FLOP validator guide: requirements, stake, rewards and rotation

Validators are the second operator role on Flop Network. They don't run AI models. They author blocks, run finality, store network data and attest to miners' proofs. This guide summarises what the draft whitepaper, testnet page and airdrop page say about running one.

What validators do

• Produce blocks. Blocks are produced once per second by a stake-weighted lottery among validators.
• Run finality. AlephBFT runs over a rotating committee of 100 validators. The committee is drawn from the full set of 1,000 and reshuffled every hour.
• Store data. Validators store and serve quotes, proofs, transcripts and model weights (the data-availability layer), plus agent memory.
• Verify miners. They attest miners' proofs and direct model loads. Re-execution itself is done by bonded checkers, and validators verify and co-sign their verdicts.
• Govern. Validators vote on Flop Improvement Proposals (FIPs). Most changes need approval from two-thirds of the active set.

The whitepaper says validators "never execute inference and need no GPU or trusted-execution hardware; a commodity server with fast storage suffices."

Validator hardware requirements (provisional)

Component	Recommended
CPU	8+ cores
Memory	64 GB RAM
Storage	4 TB NVMe (2 TB minimum)
Network	1 Gbps, redundant connection

These specs are labelled provisional and "subject to refinement before the testnet opens".

The validator set: 1,000 seats with rotation

The set is capped at 1,000. Entry at genesis is by testnet ranking. After that, roughly every 30 days, the 50 lowest performers are replaced by the top 50 in waiting. Performance is a composite of uptime, block production, accuracy, latency, and infrastructure and geographic diversity. The whitepaper says selection "is systematic and published".

Stake: the 1.2M FLOP bond

Each validator posts a self-stake bond of 1,200,000 FLOP. Per the whitepaper, there is "no delegated staking into the set", and unbonding takes about 21 days.

The intro page mentions that holders can delegate stake to a validator or miner, while the whitepaper rules out delegation into the validator set. Treat delegation as unresolved until the Yellow Paper is final.

On testnet, the network sponsors the stake. For genesis, the airdrop makes the bond possible. Each validator admitted to the genesis set gets a fixed 1,200,000 FLOP seat allocation, equal to the required stake, for up to 1,000 seats. That allocation is frozen until the first halving, about two years after genesis. It's then released one validator per day in a pre-set order. An active validator still has to keep the required stake bonded.

Rewards

• 10% of every block reward, divided across the set.
• 15% of every inference fee. This leg exists only when an agent paid for work that was actually delivered.
• Transaction base fees go to the operators who produce and secure blocks.

Slashing

Penalties are graduated. Brief downtime costs a small fraction of stake. Extended downtime costs more and leads to removal. Equivocation costs a substantial share, rising to the full bond if a third or more of the set acts together. Fraud brings the full penalty and a permanent ban. Slashed stake goes to the Flop Foundation.

Governance role

Until the first halving, only the Flop Foundation can submit protocol FIPs, but validators decide whether they pass. The whitepaper says Hayes has described this interim openly as "a constitutional monarchy" for the first two years. Upgrades take effect after a 14-day enactment delay. A multisig superuser key exists at genesis until a specified handover is carried out.

How to prepare

1. Provision hardware that meets or exceeds the provisional spec.
2. Register interest as a validator through the official testnet page.
3. Plan for high uptime during the testnet. Ranking decides admission to the genesis set.
4. Ignore anyone selling "validator slots" or "whitelist spots". Admission is by on-chain testnet ranking. See scam warnings.

Related: testnet, airdrop, mining, FAQ.