hyperlane

When we got to 2023, everyone in the interoperability space was still processing 2022. Ronin. Wormhole. Nomad. Harmony. A billion dollars gone across a handful of exploits, and the forensics on all of them told the same story: concentrated trust. Someone had decided that a small set of validators, or a guardian committee, or a single multisig, could be trusted to secure hundreds of millions in bridged assets. The attackers agreed that it was a small set and concentrated their attention accordingly. The conversations at conferences that year had a particular texture to them — teams that had been comfortable with permissioned security models were now looking at the design choices they'd inherited and asking harder questions than they'd asked before.

That was the atmosphere when I came on at Hyperlane. The team was small and technically serious and building something that felt, from the inside, like a genuine answer to the question the industry had been asking. The Interchain Security Module architecture gave each chain that deployed Hyperlane the ability to define its own verification logic. Not inherit someone else's guardian set. Not trust a committee chosen by the protocol team. Choose your validators, compose your security checks, own the outcome. The sovereignty claim wasn't marketing — it was structural, and you could feel the difference when you went into the design.

The competitive landscape in 2023 was specific. Wormhole had 19 guardians — major institutions, heavily backed after the exploit — and required a threshold of signatures to verify messages. LayerZero ran on an oracle-and-relayer model where collusion between those two parties was the attack surface. Axelar was pursuing enterprise partnerships with a proof-of-stake validator network. All three had one thing in common: to add a new chain, you made a business development call, waited for prioritization, and accepted whatever trust model you'd been given. The long tail of rollups launching in 2023 — the OP Stack chains, the Cosmos appchains, the experimental VMs — wasn't anyone's priority. There was no sales cycle fast enough to keep up with the deployment pace.

We spent a lot of time on the road that year. Chain teams, RaaS providers, validator operators, asset issuers. The conversations were usually about the same thing: what does cross-chain connectivity actually need to look like for this specific context? Not in the abstract, but for this chain's user base, this asset's security requirements, this team's operational capacity. The protocol was genuinely capable. ›The protocol had the right architecture. Whether it was ready was a harder question. The gap was consistently product — what the user actually experienced when they tried to move something from one chain to another. A chain running Hyperlane but not shipping a coherent transfer interface might as well not have deployed it. That gap was most of what we worked on.

Chainlink's CCIP launched late in 2023 and clarified things. Enterprise-audited and service-oriented on one end; trust-minimized, permissionless, and fast-to-deploy on the other. The teams that wanted to ship without getting on anyone's BD calendar were ours. The teams that needed a compliance-reviewed integration with guaranteed SLAs were CCIP's. That division wasn't sharp at the edges, but it was real enough to organize around. By the time the rollup explosion of 2024 was underway — Eclipse, Taiko, Manta, a long tail of OP Stack chains — the permissionless deployment model started showing what it could actually do at scale.

What made the work feel urgent was that the interoperability problem was being actively contested. Not just technically — competitively. We were a small team with a real architectural thesis competing against protocols with more funding, bigger BD teams, and longer chain relationships. The travel, the chain conversations, the late-night calls across time zones — all of it had that quality that comes with working on something where the outcome genuinely isn't certain. The bridge hack era had opened a window. We were trying to build through it.