Leios monthly spotlight: October highlights

Summary:

• An Ouroboros Leios prototype demonstrates how votes can be compressed into a single-block certificate, supporting fast and efficient endorsement.
• A test setup measured how Leios overlay traffic interacts with base-layer block diffusion, providing a baseline for future optimizations.
• Simulations revealed that system capacity levels off near 300 transaction kilobytes per second (TxkB/s), with a stable delay once endorsement blocks carry the majority of the transaction volume.
• Work continues on node-to-client interfaces and integration paths to limit downstream impact.
• The team progressed the technical design, testing strategy, and roadmap toward real-world deployment.

The October Leios monthly review and demo call focused on measurable progress across design, testing, and simulation. The team shared a working path for certification, explored performance under high load, and refined tools to support future implementations.

Voting and certification primitives

Hamza Jeljeli, a cryptographic engineer at Input | Output Engineering (IOE), presented a visualization of the cryptographic operations of the voting process, from committee selection to on-chain verification, ensuring that endorsement block (EB) certification remains both compact and verifiable. The committee is a combination of persistent seats (fixed for the epoch) and rotating seats (selected for each EB), which helps to balance efficiency and fairness.

A block producer aggregates all votes using the Boneh-Lynn-Shacham signature (BLS) scheme, checks for a 75% quorum, and verifies on-chain. On a standard Apple M1 system, raw votes were compressed from roughly 12 kilobytes down to a certificate of about five kilobytes in milliseconds. As a result, it can be included in a block immediately after creation, without delay. This predictability in inclusion timing allows block producers to plan more effectively and contributes to reduced bandwidth and storage overhead.\ Important: performance may vary due to local hardware resources and network conditions.

Measuring Leios traffic alongside block diffusion

Nick Frisby, a senior software engineer at Tweag, presented a three-node setup where a patched node managed Leios mini-protocols while relaying base chain blocks. Without any Leios load, block latency hovered at 0.6s, increasing to around 3s after a burst of EB traffic. 

The objective was to demonstrate and measure the interaction between these components, allowing for iterative improvements. While the prototype is still incomplete and uses experimental data stores, the focus lies on establishing a test harness and benchmark as a baseline to improve upon. The key takeaway is that Leios traffic, when processed natively, disrupts block diffusion, and that developing fetching logic and prioritization mechanisms to minimize the potential impact on diffusion latency under attack is needed.

Updated simulations and impact

William Wolff, a solutions architect at Input | Output Group (IOG), demonstrated that throughput scales with demand and then flattens near 300 TxkB/s (about 200 average-sized TPS or 1,000 small-sized TPS). Below capacity, inclusion times remain stable; above capacity, clients experience back pressure and longer waits, rather than a collapse, which aligns with the behavior observed in the current mainnet deployment of Cardano using Ouroboros Praos.

When demand is low, ranking blocks (RBs) handle many transactions. As load increases, more transactions flow through EBs, incurring a roughly constant additional delay. The CPU and network costs per transaction decrease and then level off as demand rises. 

Recent work has introduced bounded mempools, allowing inclusion times to reflect back pressure in subsequent runs more accurately. For more details, refer to the simulation updates in the slides.

Ecosystem prototypes and interfaces

TxPipe’s Pedro Sánchez Terraf outlined an initial Leios path inside the Palace toolkit, with an EB certificate and optional EB announcement in the header, plus a lightweight server-client demo to exercise message flows. 

Blink Labs’ Chris Gianelloni focused on the node-to-client interface, proposing a merged-block approach to minimize changes for chain-sync clients and moving that discussion into the Cardano Improvement Proposal (CIP) process, which would help reduce downstream integration cost as the technical design matures.

Roadmap and impact analysis

IOG’s Sebastian Nagel, lead technical architect, outlined how the team is turning proposal-level ideas into concrete node scope, integration points, and measurable milestones. The near-term focus includes stabilizing cryptographic application programming interfaces, publishing microbenchmarks for voting, aggregation, and verification, adapting a trace verifier for conformance across implementations, and iterating on the network measurement harness to lower the EB-traffic impact on Praos diffusion. For updated roadmap pointers, please refer to the October deck.

Governance and scope questions

October's discussion centered on crucial topics, including layering choices, system behavior under heavy load, and potential front-running vectors. During the review, the team successfully identified and fixed a simulator issue, which led to improved reported results. Future runs will utilize bounded mempools, and a public thread will be initiated to gather community input on effective mitigations. Interface work is proceeding in the open to minimize downstream churn as the specification continues to mature.

Open questions for the community

• Do you agree with the rationale for the proposed Linear Leios in CIP-164? We would like to hear from you on this thread.
• Would the node-to-client interface change proposed in this pull request work for your application?
• Are you concerned about front-running on Cardano right now? Would this change if the number of transactions sequenced by block producers increases tenfold?

📣 Join the discussion on GitHub or during the next community call.

Get involved

Watch the October Leios call recording on YouTube and don’t miss the next November edition:

• Date: Wednesday, November 26, 2025
• Cadence: Last Wednesday of every month
• Who: Open to the entire community.

Bring your questions, feedback, and demos. We look forward to seeing you there!