Tl;dr

S-two (pronounced: “Stew”), StarkWare’s fully open-source next-generation prover which is the fastest in the world, is now live on Starknet mainnet, replacing Stone as the network’s prover. Every Starknet block is now proven by S-two, securing every transaction with the most advanced circle STARKs technology. This milestone reinforces Starknet’s position as the leading high-performance and trustless blockchain stack.

S-two’s speed also enables client-side and general-purpose proving. From private DeFi, proof of identity and verifiable AI, S-two will make it possible to generate proofs locally and verify them onchain at minimal cost.

Finally, S-two sets the stage for decentralised proving. In the future, anyone will be able to sequence transactions and generate proofs, ensuring that Starknet’s state updates to Ethereum never stop. This architecture also strengthens network security by removing single points of failure and distributing trust across participants.

The S-two era has begun. The train has left the station.

What Exactly is S-two?

S-two is StarkWare’s next-generation prover and a major scientific breakthrough in zero-knowledge technology. Designed to be the most performant prover ever built, S-two delivers an unprecedented leap in proving speed and efficiency. Benchmarks show that it generates proofs up to an order of magnitude faster than its predecessor, Stone, achieving proof generation in seconds for computations that previously took minutes and outperforming every other prover in its class.

New to S-two? Here is everything you need to know.

 

Replacing Stone, which secured StarkEx and Starknet for the past six years, S-two introduces a new cryptographic foundation built on Circle STARKs and the M31 field – a small, highly efficient prime field optimized for performance. Implemented in Rust, it supports both the Cairo programming language and direct AIR (Algebraic Intermediate Representation) definitions, giving developers greater flexibility, precision, and speed when building or integrating proving systems.

Here is an illustration of S-two’s speed vs. competing provers, press “Start Race” and “Reset” to feel the speed:

Why This Matters for Starknet

1. Higher Starknet Efficiency and Reduced Costs of proof generation

This reduction in proof generation costs not only improves network efficiency but also expands what developers can build. Applications that require frequent or complex proof generation, such as privacy-preserving DeFi protocols, verifiable identity systems, or real-time data attestations, can now operate economically at scale. By lowering the cost of trust, S-two enables new classes of applications that were previously impractical on-chain.

 

2. Client-Side & General-Purpose Proving

Thanks to its efficient backend and hardware optimizations, S-two makes it feasible to generate proofs locally on any device, even on low-powered ones such as mobile phones, for any arbitrary logic (not just rollup-specific). This opens up possibilities such as:

    • Scalable private ERC20 transfers that hide both addresses and amounts

    • A large variety of DeFi applications such as private trading and dark pool.

    • Identity proofs such as “I am a unique human”, “My age is over 21”, or “I hold credential X”, without exposing personally identifiable information or data

    • Verifiable AI or ML inference, proving that a model ran on a given input and produced a specific output

With an efficient enough prover, these and many other use cases can now become reality, both on Starknet and both on SN Stack.

3. Path to Decentralised Proving

The reduction in the cost of proving paves the way for the next step in decentralization: as more sequencers will operate Starknet they will enjoy the benefit of proof generation via S-two which will be cost efficient and will enable them to run sustainably.

S-two in Motion: Real-World Client-Side Proving with Kakarot

Kakarot’s Fibrace proving game experiment, detailed in a white paper hosted on arXiv:2510.14693, demonstrates how S-two is already enabling large-scale client-side proving across the Starknet ecosystem. In this experiment, millions of zero-knowledge proofs were generated locally on thousands of different smartphone models, showcasing both the scalability and accessibility of S-two’s proving technology.

By leveraging S-two’s highly optimized proof system, Kakarot executed complex, high-throughput workloads directly from user devices, achieving performance levels that were previously possible only through centralized provers. This large-scale experiment validates the efficiency and robustness of S-two’s open proving framework and highlights that developer teams are already integrating it into real-world systems. The path toward user-driven proving and modular proving architectures is no longer theoretical; it is already underway.

What’s Next for Starknet & the Ecosystem

S-two has now moved from development and testing to actual use in production, starting to become more mature and battle-tested. So what comes next for S-two?

    • Lower proof-generation costs: As more blocks are proven by S-two, ongoing optimizations will continue to reduce the cost per proof and improve Starknet’s scalability and cost efficiency as well as every chain built with SN Stack.

    • New classes of applications: With S-two’s client-side proving capabilities, developers can now build privacy layers, identity systems, and other applications that were previously impractical due to performance or cost constraints.

    • Decentralised proving: S-two will enable more sequencers to participate in proving Starknet’s state, creating a more resilient and trustless network.

    • Developer tooling and libraries: APIs, templates, and abstractions will be introduced to help builders integrate S-two proving into their applications, supporting use cases such as identity, privacy, and zkML modules.

    • Ecosystem alignment and standardization: As S-two becomes the default prover for Starknet and a shared standard beyond it, builders across the ZK and blockchain ecosystem will align around its framework, accelerating innovation and interoperability.

Final Word

The launch of S-two on Starknet is more than an infrastructure upgrade. It’s a change in what’s possible. With every Starknet block now backed by high-performance STARK proofs and a proving framework that aims for decentralised participation and client-side proof generation, the stage is set for the next generation of applications: truly scalable, verifiable, privacy-centric and decentralised.

Whether you’re a developer building on Starknet, a user looking for trust, privacy, and performance, or just someone following the ZK and scaling space, now is the moment to take notice and participate.

 

Read the S-two book →