Layer 2 Boundaries & Trust Models
vProgs Status: ACTIVE DEVELOPMENT / PROTOTYPE | Igra / Kasplex L2: EXTERNAL ARCHITECTURES
As Kaspa expands its capabilities post-Toccata, the term "Layer 2" (L2) is often overloaded. It is critical to distinguish between the different architectures leveraging Kaspa.
Sovereign Verifiable Programs (vProgs)
- Execution: Off-chain sovereign runtime.
- VM Compatibility: Custom verifiable programs (e.g., RISC Zero guests). Not EVM.
- Settlement: Kaspa L1 covenants using
OpZkPrecompile. - State: Application state is off-chain; L1 only tracks state roots and sequencing commitments.
EVM-Compatible Based Rollups (Kasplex L2 / Igra)
Kaspa L1 does not execute EVM bytecode, and Toccata is not an EVM hard fork.
External projects (like the Igra implementation or Kasplex L2 architecture) may build EVM-compatible "based rollups".
- Execution: Off-chain EVM node.
- VM Compatibility: EVM.
- Sequencing: Derives ordering from Kaspa L1 (Based).
- Bridge: May rely on trusted MPC (Multi-Party Computation) bridges initially, or validity-proof (ZK) bridges if/when fully implemented.
Note: Kasplex L1 (Asset Hub) is distinct from Kasplex L2 Architecture.
The "Trustless" Terminology Trap
Avoid using the word "trustless" casually when describing L2s or bridges.
- Validity-Proof Enforced Exit: A ZK bridge uses Kaspa's
OpZkPrecompileto verify that an L2 withdrawal is valid according to the L2's state rules. This removes the need to trust an MPC committee for correctness. - Remaining Trust Assumptions: Even with a ZK bridge, you must still trust:
- Data Availability: Can you get the data needed to construct the withdrawal proof?
- Operator Liveness: Will the sequencer/prover actually process your exit?
- Censorship Resistance: Can you force an L1 exit if the L2 sequencer ignores you?
A bridge is only as secure as its weakest assumption. Toccata's consensus primitives do not automatically make an external bridge "trustless" simply by existing.