Staking means two quite different things and conflating them is how protocols end up with economics that do not work.
Securing something with capital at risk is one thing. Paying people to lock tokens is another. Both get called staking.
Start by naming what it secures
If the answer is a chain, an oracle, a bridge or a service guarantee, staking is doing real work and slashing gives that work teeth.
If the answer is nothing — if stakers lock tokens, receive newly minted tokens, and no participant is taking on an obligation — then the mechanism is a distribution schedule. That can be a legitimate design choice. It should just be described accurately, because sophisticated holders work it out immediately and the credibility cost of the euphemism is higher than the cost of the honesty.
Reward funding decides everything
Two sources, and the difference is fundamental.
Protocol revenue. Fees generated by real activity flow to stakers. Sustainable indefinitely, and the yield reflects genuine demand for the protocol.
Inflation. New tokens are minted and paid out. The nominal yield is whatever you set it to, and the economic reality is that non-staking holders are being diluted to pay staking holders.
Inflation-funded rewards are not automatically wrong — bootstrapping a validator set or a supply side often requires them. They should be finite, declining on a published schedule, and explicitly justified.
What fails is perpetual high inflation presented as yield. It is a transfer, not a return, and the market eventually prices it as one.
Unbonding is a security requirement
The most common product complaint about staking is the withdrawal delay, and the most common design mistake is shortening it to reduce that complaint.
The unbonding period exists so capital remains at risk long enough for misbehaviour to be detected and penalised. Remove it and a participant can act badly and exit before consequences arrive, which means the stake was never securing anything.
The length should follow from how long detection and challenge actually take in your system, not from what feels competitive.
Liquid staking derivatives
The genuine solution to the liquidity complaint: a token representing the staked position, tradable while the underlying stake stays bonded.
This works and it introduces real complexity. The derivative can depeg from the underlying, particularly under stress. It creates a second asset with its own liquidity requirements. And it becomes collateral elsewhere in DeFi, which means your staking design is now load-bearing for protocols you do not control.
Worth doing, worth doing carefully, and worth being explicit with users about what the derivative is and is not.
Slashing, defined narrowly
If staking secures something, there must be a penalty for failing to secure it.
Good slashing design is narrow and provable: specific, on-chain-verifiable conditions, proportionate penalties, and a clear distinction between malicious behaviour and operational failure.
Aggressive slashing for honest faults — a node going offline briefly, a missed attestation during an infrastructure incident — drives away exactly the careful, professional operators you most want participating. The careless ones stay because they were not managing risk anyway.
Reward accounting at scale
The unglamorous engineering problem, and the one that causes the most production bugs.
Rewards accrue continuously across a changing set of participants with changing balances. Computing each participant’s share on every block is not viable, so protocols use accumulator patterns that compute lazily on interaction.
These are well-understood and easy to get subtly wrong, especially around rounding, and rounding errors compound over millions of interactions into a real discrepancy nobody can explain.
This is precisely the kind of logic that invariant testing catches and unit testing does not: assert that the sum of all claimable rewards never exceeds what the protocol has actually allocated, then try hard to violate it.
Common questions
What does staking actually do?
In a proof-of-stake network it secures the chain by putting capital at risk behind honest behaviour, with slashing as the penalty for misbehaving. In application-level protocols it usually secures something narrower - an oracle, a bridge, a service guarantee. Staking that secures nothing and simply pays holders for locking tokens is a distribution mechanism with a security-sounding name, and it should be described accurately.
How should staking rewards be funded?
From protocol revenue wherever possible, and from inflation only with a declining, finite schedule. Rewards paid purely from new issuance are dilution with a friendly interface - the nominal yield is real and the economic return to a holder who does not stake is negative. Markets price this correctly eventually, and protocols that were honest about it from the start fare better.
Why do staking protocols have unbonding periods?
Because without one, a participant can misbehave and withdraw before the misbehaviour is detected and penalised. The unbonding period keeps capital at risk long enough for slashing to be enforceable. It is a security requirement rather than a liquidity preference, which is why liquid staking derivatives exist to give holders tradability while the underlying stake stays bonded.
What is slashing and when should a protocol use it?
Slashing destroys or redistributes part of a staked balance when a participant provably misbehaves. It should exist wherever staking secures something real, because without a penalty the stake is not securing anything. It should be narrowly defined, provable on-chain, and proportionate - overly aggressive slashing for honest operational faults drives away exactly the careful operators you want.
Does Corum8 build staking protocols?
Yes. We build staking and delegation contracts, reward accounting and distribution, unbonding and slashing logic, liquid staking derivatives where appropriate, and the interfaces around them. We coordinate independent third-party audits before mainnet.
