HotStuff-2

• Name: HotStuff-2
• URL: https://eprint.iacr.org/2023/397
• Category: two-phase responsive BFT consensus design / chained-commit consensus primitive / low-latency sequencing comparison baseline
• Summary: HotStuff-2 is worth cataloging not as a generic consensus-paper footnote or just a point of comparison inside later Espresso/HotShot discussions, but as a distinct chained-BFT design that claims an unusually attractive bundle of properties at once: two-phase commit within a view, optimistic responsiveness, $O(n^2)$ worst-case communication, and optimistically linear communication. The reviewed primary materials frame HotStuff-2 as a simplification of the original HotStuff rather than a wholly different family, with the key move being that a block can commit once a quorum certificate itself receives a second round of certification. That makes HotStuff-2 a useful comparison point for HotStuff, Tendermint, HotShot, and later shared-sequencing systems because the real control surfaces are commit-rule depth, lock / justify discipline, view-change mechanics, pacemaker behavior, and whether lower latency is bought by extra protocol complexity or by a cleaner safety argument.
• What it does:
  • Defines a partially synchronous leader-based BFT protocol that aims to preserve HotStuff’s responsiveness and communication properties while reducing the happy-path commit regime from three phases to two
  • Replaces the original three-chain intuition with a two-step certificate structure where a quorum certificate on a block is itself certified again to form the commit trigger
  • Keeps leader-driven proposal flow and lock-based safety checks, with replicas comparing a proposal’s justify certificate against their currently locked certificate before voting
  • Treats view change and timeout handling as first-class liveness machinery rather than an afterthought, with implementations like Hotmint surfacing status messages, timeout certificates, and pacemaker backoff explicitly
  • Serves as a lower-layer consensus baseline for later systems that reuse leader-based BFT ideas but add other concerns such as data-availability committees, rollup confirmation, or modular sequencing
  • Provides a cleaner analytic baseline for distinguishing consensus-core questions from the extra DA, proof-publication, and application-routing layers that later systems often bundle into one shared sequencer story
• Key claims:
  • HotStuff-2 clears the corpus bar because it exposes a reusable consensus sublayer already referenced throughout the corpus queue, but not yet preserved as its own mechanism page.
  • The ePrint abstract is explicit about the protocol’s claim: two phases are enough to achieve partially synchronous BFT with optimal quadratic worst-case communication, optimistic linear communication, and optimistic responsiveness. That combination is the main reason the design matters.
  • The practical mechanism worth preserving is the shift from HotStuff’s deeper chained-commit intuition to a two-step certificate path. In implementation notes such as Hotmint’s protocol writeup, a quorum certificate followed by a double certificate becomes the concrete commit trigger.
  • HotStuff-2 is analytically useful because it isolates where later systems differ. If HotShot, Espresso, or other shared-confirmation stacks add DA committees, VID, L1 proof publication, or namespace routing, those layers should be compared against a clear lower-bound consensus baseline rather than against a vague HotStuff-like label.
  • The strongest caveat from this pass is source type: the canonical primary artifact is still an extended abstract and author commentary, while Hotmint is an implementation-oriented interpretation rather than the canonical protocol spec. That is enough for corpus intake, but the entry should still be treated as a mechanism page first, not as a production-network profile.
• Whitepaper: The canonical primary artifact in this pass is the ePrint paper Extended Abstract: HotStuff-2: Optimal Two-Phase Responsive BFT, supported by author commentary and a modern implementation-oriented protocol writeup; see ../../whitepapers/hotstuff-2-primary-sources-2026-05-14.md.
• Sources:
  • https://eprint.iacr.org/2023/397
  • https://malkhi.com/posts/2023/03/hs2/
  • https://github.com/NBnet/hotmint/blob/main/docs/protocol.md

Internal linkages

• Best direct descendants or cousins in the corpus: hotshot and alpenglow.

• Sequencing stacks that layer more machinery above a BFT core: espresso and astria.

• Reusable lens: many fast finality claims reduce to commit depth, certificate structure, and pacemaker behavior, not a new control plane.

• Last reviewed: 2026-05-14 UTC