10Lesson

Staking, root, and why concentration decides outcomes

Earlier lessons showed how judgment turns into pay. This one asks who actually controls that judgment — and the honest answer is not "one participant, one vote."

Backing a validator with money doesn't require running one yourself. A nominator is exactly that: someone who commits stake behind another participant's judgment instead of judging anything themselves, and is paid a share of whatever that participant earns for judging well. [source] To attract that money in the first place, the validator publishes who it is and sets a take — the cut it keeps for itself before nominators are paid their share. [source] [source]

Meeting the chain's bare minimum to hold a permit is not the same as competing for a serious return. Real validating typically takes stake worth tens of thousands of TAO-equivalent, and because root network returns add up across every subnet a validator serves, the people with that kind of money tend not to pick just one — they register broadly and collect a slice of many contests at once, which is exactly how a large position keeps getting larger. [source]

Joining the root network itself works the same way as everything else here: placement is decided by stake, and a full one evicts whoever has the least. The body that used to let the largest root members vote on network changes — a formal senate — has been removed from the chain entirely; the small set of privileged, protocol-level changes still possible today are dispatched by a multisig of foundation-held keys, not by stake at all. [source]

This is why concentration, not headcount, decides outcomes. A validator's say is scaled by how much stake backs it, and the point that survives Yuma Consensus's clip on any given miner is the point a majority of that backing supports — not a majority of judges. Whoever holds the most money does not need to convince anyone else; they simply are the majority, more of the time, on more of the network. [source] [source]

Some of this concentration is not emergent at all — it is written into the rules. A subnet owner's own key is exempt from the entire permit contest: it always holds a validator permit, and its backing is never treated as too small to count. [source]

How it actually works

Validator permits are recalculated every epoch as the top-K neurons by stake weight, with K set by MaxAllowedValidators (default 128, owner/root-settable). A separate chain-wide minimum, StakeThreshold, zeroes out any stake weight below it — the runtime default is zero, but mainnet runs a non-zero floor (around 1,000 TAO-equivalent at time of writing), and the same floor doubles as the minimum stake needed to set weights at all. [source]

"Stake weight" is not raw TAO. It is the alpha a hotkey has staked on that specific subnet, plus its TAO stake discounted by a global TaoWeight factor — governance-set, currently 0.18 on mainnet against a runtime default near 0.053 — so alpha counts far more toward a permit than an equivalent amount of TAO does. [source]

There is also a clock attached to all of this: the activity cutoff, which scales with tempo as activity_cutoff_factor × tempo / 1000 blocks (factor defaults to 13,889, about 5,000 blocks at the default tempo). A validator that hasn't set weights within that window has its stake masked out of consensus entirely until it submits again — inactive stake, however large, stops counting. [source]

Stake itself can be repointed without moving it: set-children lets a parent hotkey delegate a fraction of its stake weight to other hotkeys on one subnet — commonly used to aim stake at a separate validating key — subject to a rate limit and a cooldown before it takes effect. [source]