Most chains get slower as they scale. Aptos keeps getting faster. Introducing Consensus Observer: Cutting End-to-End Latency by Half at 10k+ TPS. What you need to know about the latest latency breakthrough to arrive on Mainnet.
TLDR; Sub-second isn’t a claim — it’s a commitment.
Traditionally, fullnodes wait until validators execute and commit a block before syncing it. That delay can add hundreds of milliseconds — or more at high TPS.
Consensus Observer (CO) changes that by allowing blocks to be shared immediately after ordering, before execution or commitment. CO is the answer to a key question: What’s the fastest way to get fullnodes access to block data—so they can process in parallel with validators?
With CO, fullnodes now:
• Receive blocks in real-time
• Execute alongside validators
• Commit once consensus finalizes
The result? Massive latency gains.
Since its deployment, CO has cut end-to-end latency by ~50% at higher loads (8.5k+ TPS). And unlike most systems, CO improves both average (P50) and tail (P90) latencies — making performance more predictable, too.
Why it matters: Consensus Observer is live on Mainnet, unlocking use cases that require ultra-low latency and low-cost transactions: • Micropayments (e.g., per-second content billing) • Streaming payments • Autonomous payments triggered by real-world events These demand speed, predictability, and scale. Aptos delivers all three.