The official docs make clear that =nil; should be tracked as more than a generic high-throughput L2 claim. The architecture is explicitly split into execution shards, one consensus shard, local committee consensus, global consensus, and a proof-aggregation path back to Ethereum.
The overview page frames =nil; around three explicit design goals: horizontal scalability, modularity through Ethereum DA/consensus, and a single proof attesting to the correctness of all shards.
The principles and shard pages expose an unusually crisp control surface for cross-shard composition: outgoing messages that target accounts on another shard become necessary messages, and blocks are invalid unless those messages are included by the destination shard.
The async-execution tutorial shows that =nil; wants developers to reason in explicit message-passing terms, not only synchronous call terms. asyncCall, awaitCall, and callback-style request flows are part of the application model, not merely validator internals.
The docs also surface two analytically useful deviations from the usual Ethereum mental model: all accounts are smart contracts rather than EOAs plus contracts, and internal messages can pay for their own top-level processing while external messages are paid by the receiving contract.
Contract co-location is worth keeping in view because it makes locality an explicit policy surface. =nil; is not only about sharding state; it also lets builders decide when tightly coupled contracts should share a shard and absorb the tradeoffs.
No single canonical whitepaper was surfaced during this pass. The docs and public research/org links were the strongest current primary-source packet.