Sonic Labs (@SonicLabs) Shit poster 100% escape. Technical posting due to 100% shame. ☠️ Short review .. It's very computer science-oriented, so it might not be easy to understand.. (I'll try my best) > Sonic one-line definition "EVM 100% compatible L1 + aBFT(DAG) consensus aiming for ultra-low latency and high throughput chain" If we say it as per the official document, it is a performance-oriented L1 that prominently claims 400k TPS. > So what is aBFT + DAG consensus? First, let's understand DAG. DAG can be likened to "instead of standing in a single line, multiple lines proceed simultaneously." DAG = Directed Acyclic Graph Directed: There is a direction of arrows (who references whom). Acyclic: There is no cycle that returns to the starting point → the flow of time only goes in one direction. Graph: A collection of points (nodes/events) and lines (references/edges). The concept of that graph (Node, Edge) is something learned in traditional computer science classes on data structures/algorithms. (A circle is a Node and a line is an Edge) Consider a task as a Node, and sequentially solve it with directional Edges, while simultaneously increasing performance through parallel processing where possible. At this point, the order of tasks to be executed simultaneously must be determinable, so the condition is that the graph must not have cycles. A graph like below can cycle as B -> C -> E -> D -> B... and we aim to prevent that from happening. Ugh.. At this point, I feel like there will be dropouts, and I think I should have just asked ChatGPT to do it and copy-paste instead. Anyway, how is this different from blockchain? Traditional blockchains (like Bitcoin) are linear chains: … → Block N → Block N+1 → Block N+2 … (only one can expand at a time) In a DAG system, events (units like small blocks) arise simultaneously, and each event references multiple previous events as it spreads. aBFT: "Let's reach consensus safely even if not fully synchronized" First, what is Byzantine Fault Tolerance (BFT)? It is a consensus principle that ensures the system does not reach a "wrong conclusion" even if some participants act arbitrarily (lying, hacking, going down) or the network is temporarily chaotic. Fault: Not just simple crashes but also includes malicious actions like lying (Byzantine) Tolerance: Designed so that even if there are a few such participants, the whole can safely reach the same conclusion. BFT vs aBFT Classical BFT (e.g., PBFT, Tendermint) Usually assumes partial synchrony: "Messages arrive within a certain time" and believes that such an upper bound (Δ) will eventually hold. aBFT Does not assume any upper bound on message delivery delays (just that messages will arrive eventually). It is designed to maintain safety even in this environment and recover liveness when the network stabilizes. → Result: Even if the network is temporarily chaotic, the focus is more on preventing contradictory confirmations, even if the announcement (confirmation) time may be delayed. In summary, Sonic adopts asynchronous Byzantine Fault Tolerance (aBFT) consensus and exchanges blocks asynchronously in a DAG. Each node verifies transactions and accumulates them in a local DAG, and once sufficient observation/propagation occurs, the main chain order is assigned and confirmed. Effect: Designed to reduce slot waiting/synchronization bottlenecks and build consensus trust quickly as the network runs simultaneously. Prerequisite: The network/validator configuration must be solid, and the final perceived performance depends on operational conditions (peering, hardware, network quality). > Let's implement a super-fast Layer 1 through distributed processing ^^ This is the goal of the chain.... It seems that cryptography + computer science struggles are well integrated. ... It is difficult to understand just by reading through. Thank you very much if you read it carefully... GG