---

name: bitcoin-l2-bitvm description: | BitVM, BitVM2, BitVM3: off-chain computation framework using optimistic challenge games on Bitcoin script. Foundation for trust-minimized bridges to L2s and PoS chains. USE WHEN: building trust-minimized bridges, understanding ZK rollup peg-out mechanisms, evaluating bridge security models. allowed-tools: Read, Grep, Glob

BitVM

BitVM is a paradigm for off-chain computation with on-chain verification challenges on Bitcoin. Allows arbitrary programs to be verified via fraud proofs in Bitcoin script — without OP_CTV, OP_CAT, or other unactivated opcodes.

Original paper: Robin Linus, 2023. Refined versions: BitVM, BitVM2, BitVM3.

Core idea

1. Prover commits to a program execution (via cryptographic commitments
   to step states / hashes).
2. Verifier challenges: "what's the state after step k?".
3. Prover responds with state values + proofs.
4. Iterate via binary search over disagreement.
5. Final challenge runs on Bitcoin: a 1-step micro-computation
   verifiable in Bitcoin script.

If prover can't respond honestly, verifier wins funds (Prover's bond is slashed).

BitVM (original)

Single prover-verifier pair. Setup is heavy (lots of pre-committed hashes); single use per setup.

BitVM2

• Multi-prover: multiple challengers can dispute.
• Permissionless verification: anyone can challenge.
• Smaller setup: more efficient hashing schemes.

Used by:

• Citrea mainnet bridge.
• Fiamma, whose BitVM2 bridge lists Babylon as its Wasm-VM target chain (Fiamma's own docs, last updated January 2026). Babylon's trustless-vault docs do not credit Fiamma as of September 2026.

BitVM3

• Further efficiency improvements. The academic write-up — "BitVM3: Efficient Bitcoin Bridges via Garbled Circuits" (Robin Linus Woll, Alexopoulos, Aumayr, Avarikioti, Maffei, Tse; IACR eprint 2026/933, received May 2026, revised June 2026) — has the challenger evaluate a garbled circuit entirely off-chain, which the paper reports as a nearly 1000x saving: total on-chain cost ~$9, with the challenge transaction itself ~$0.20, against a ~$16,000 worst-case dispute cost under BitVM2.
• Babylon's Trustless Vaults were announced on BitVM3 (whitepaper, August 2025). As of September 2026 Babylon's own docs describe the vaults as public testnet — not mainnet.
• The RSA-based BitVM3 line was reported broken (by July 2025). Alpen Labs' post "Glock: A new standard for verification on Bitcoin" (15 July 2025) reports that Liam Eagen and Fairgate Labs found "separate core flaws in its RSA-based construction, rendering it unsafe to build upon", and that Alpen's "direction now fully orients towards Glock in the development of the Alpen ZK rollup and the Strata bridge" (Glock: Garbled Locks for Bitcoin, IACR eprint 2025/1485, August 2025). Strata's bridge moved off BitVM2 to Glock/Mosaic (as of September 2026) — see strata/SKILL.md.
• Reduces interactive challenge rounds.

Use cases

• Trustless bridges for L2 peg-out.
• Optimistic rollup verification on Bitcoin.
• Decentralized custody without federation trust.

Trade-offs

• Verification cost: large setup tx + many bytes for challenges.
• Challenge windows: 24+ hours typical → peg-out delay.
• Liveness: requires at least one honest challenger online.

Compared to other trust models

Model	Trust	Liveness req
Federation multisig	n-of-m honest	none for spending
Threshold (FROST)	k-of-n honest	k-of-n online to sign
BitVM2	1-of-n honest challenger	challenger online during window
ZK rollup direct	math	proof verifiers must be online

BitVM is the "1-of-n watchtower" of bridges.

Implementations

• bitvm-rs (research) — Rust BitVM implementation.
• Citrea bridge uses BitVM2 in production.
• Fiamma — BitVM2 bridge to Cosmos; Babylon is its Wasm-VM target chain (Fiamma docs, January 2026).
• Multiple research forks for variations.

Limitations

• Complex — hard to audit, may have bugs.
• Setup overhead — large pre-signed tx tree.
• Latency — challenge windows mean peg-out is not instant.
• Non-trivial economics — bond sizing matters for security.

Status (as of September 2026)

• BitVM2: production — Citrea's Clementine bridge on Citrea mainnet (clementine v0.6.6, June 2026; citrea node v2.8.0, September 2026).
• BitVM3: research / preprint, not deployed. Announced for Babylon's Trustless Vaults in August 2025; Babylon's vault-first roadmap (October 2025) targeted mainnet "early next year", and as of September 2026 the vaults are still public testnet. No primary source confirms BitVM3 live on mainnet anywhere.
• Active research on further efficiency.

See also

• citrea/SKILL.md
• babylon/SKILL.md
• strata/SKILL.md
• ../../protocol/proposals/SKILL.md