Avalanche

In May 2018, a pseudonymous group called Team Rocket circulated “Snowflake to Avalanche.” The paper proposed leaderless consensus based on repeated subsampling and metastable preference. The reviewed official history leaves the group's identities undisclosed.

Emin Gün Sirer, Kevin Sekniqi and Maofan “Ted” Yin were pursuing related work at Cornell. Cornell's account and Kevin Sekniqi's later history say that this group carried the paper's ideas forward with Team Rocket and formed the team behind Ava Labs. That makes them the builders and company founders in the documented story, not named authors retroactively attached to the anonymous paper.

Avalanche mainnet launched on September 21, 2020. The Primary Network divided work among the X-Chain for native assets, the P-Chain for validators and platform coordination, and the EVM-compatible C-Chain for smart contracts. The separation gave Ethereum developers a familiar execution path without forcing asset and validator records into the same chain.

AVAX connected the three jobs economically. Validators bonded AVAX for stake-weighted participation, users paid Primary Network fees in AVAX, and those fees were burned. Staking rewards were newly minted. The token therefore combined a hard documented ceiling with an inflationary path below that ceiling.

The Snow family uses small repeated peer samples rather than asking every validator to exchange votes with every other validator. Snowman turns that confidence process into a linear order suited to blockchains. At launch, P- and C-Chains used the linear engine while X-Chain used the Avalanche DAG.

Cortina changed the public architecture in April 2023. It stopped new X-Chain vertices and used the final DAG state as the starting point for a Snowman++ chain. The move simplified state sync, exchange integration and cross-chain messaging. It also made old descriptions of a currently operating Avalanche DAG misleading.

Horizontal scaling creates another problem: isolated chains need to authenticate one another. Durango activated on March 6, 2024 and made Avalanche Warp Messaging available to EVM chains. A source validator set can produce an aggregate signature for a message, while relayers deliver it and receiving contracts decide what to do.

That division preserves important boundaries. A valid source-chain signature cannot prove that the receiving application is bug-free, that a relayer will deliver promptly or that two L1s have equal validator quality. Interoperability joins security domains; it does not erase their differences.

Under the original Subnet model, a validator of an application chain also had to validate the Primary Network and stake at least 2,000 AVAX. The rule tied additional chains to Primary Network membership but made every validator carry the same capital and hardware obligation.

Etna activated on December 16, 2024. New or converted L1s could register L1-only validators governed by a validator-manager contract. Those validators pay a dynamic continuous P-Chain fee and do not need to validate X- or C-Chain or stake 2,000 AVAX. They still synchronize current P-Chain state for validator-set and messaging information. Primary Network validators kept their own staking rules.

Etna did not silently migrate every existing network. An owner must convert control to a validator-manager contract and register the new validator set; an unconverted network can continue with pre-Etna validators. An L1 may use proof of stake, proof of authority or custom logic, including permissioned membership.

The lower common requirement makes application chains cheaper to operate, while weakening any shortcut that treats the Avalanche name as one shared security guarantee. Validator admission, rewards, administrator keys, gas tokens, upgrades and emergency controls can differ from one L1 to the next.

A Primary Network validator currently self-stakes at least 2,000 AVAX. Each delegation must be at least 25 AVAX, and total validator weight is capped at the lower of 3,000,000 AVAX or five times the validator's own stake. Terms run from two weeks to one year, the minimum delegation fee is 2%, and more than 80% uptime is required to earn rewards. Sampling weight follows stake, but the protocol's documented penalty is reward loss rather than confiscation of principal.

The supply schedule also needs both sides of the ledger. Official documentation says 360 million AVAX were minted at genesis and supply can rise toward 720 million through validator rewards. Primary Network fees are burned, yet the documentation says issuance normally outweighs burns while supply remains well below the cap. Burn totals do not prove a falling supply.

A 2020 Avalanche allocation explainer assigned 10% of the 720-million maximum to the Ava Labs team and 9.26% to the Foundation, alongside sale, community, partner, airdrop and testnet programs. Those launch allocations describe the original plan, not present ownership or liquidity.

A July 2025 audit found that stateful precompiles handled delegatecall semantics in a way that could enable caller impersonation and forged privileged actions. The project rolled Foundation validators back, issued a soft-fork containment release and reported no evidence of exploitation on Mainnet or a public L1.

Granite activated on November 19, 2025 and made the permanent fix: affected precompiles reject delegated calls and the EVM library distinguishes raw from EVM-semantic caller contexts. The episode matters because custom precompiles are part of Avalanche's flexibility, and the same customization surface can widen the consequences of an undocumented assumption.