init4
In 2025, the rollup landscape had delivered what it promised and also something nobody quite wanted. There were hundreds of L2s. You could deploy one in an afternoon. The infrastructure ecosystem had matured enough that spinning up a chain was a product decision, not an engineering heroic effort. And yet the conversations we kept having — at conferences, in group chats, on calls with teams who'd been through it — circled back to the same friction. Cross-chain operations were still slow, still probabilistic, still required a layer of trust in a bridge or a relayer that nobody had fully examined. Morpho had fragmented itself across half a dozen chains chasing liquidity that never quite cohered. The complexity hadn't been solved. It had been made more accessible, and more dispersed.
The diagnosis we kept arriving at was the same one from different directions: timeline separation. Every major rollup runs on its own clock. It produces blocks, accumulates state, and settles to Ethereum on its own schedule. The settlement is real, but it comes later, and 'later' is where all the complexity lives. The bridge trust assumptions, the withdrawal windows, the confirmation delays, the stuck funds — all of it is the cost of managing the gap between two chains that don't agree on time. The canonical bridge delays, the fraud proof windows, the relayer markets — these weren't the problem. They were symptoms of one upstream design choice: the rollup and Ethereum were allowed to diverge in time.
Signet started from a different premise: every rollup block corresponds to exactly one Ethereum block. Not synchronized to it after the fact — built to treat Ethereum's clock as canonical from the start. When we worked through the implications in those early sessions, the conclusions were almost too clean. If the two chains share the same block, assets sent to the deposit contract on Ethereum appear on the rollup in that same block. Not in the next relayer cycle. Not after a fraud proof window. The same block. Passage handles deposits. Transact enables L1→L2 execution — an Ethereum transaction that triggers L2 execution — within the same block. One clock eliminates an entire category of protocol surface area.
The block production layer was the other half of the diagnosis. Ethereum's PBS architecture had concentrated block building: a small number of specialized builders capturing the ordering rights for the chain. That concentration wasn't incidental — it was structural, a consequence of the auction dynamics where the entity willing to pay the most for the block wins it, and the entities willing to pay the most are the ones with the best MEV extraction pipelines. The same dynamic plays out downstream on rollups. On Signet, block building is round-robin: each builder gets an 8-second window in a 12-second cycle. No auctions. The blind sequencer signs only the block hash — it never sees contents. The precondition for ordering-based extraction is removed by design.
The incentive program data from 2024 reinforced something about how rollup growth was being manufactured. Arbitrum's STIP distributed 71.4 million ARB to drive activity on the chain. The on-chain data showed that 94% of that activity was mercenary capital — participants who came for the incentives and left when they ended. Optimism's airdrop data was equally direct: 48% of recipients sold as their first on-chain action. The chains were buying metrics, not building networks. The retention wasn't there because the underlying experience wasn't differentiated enough to justify staying. You can buy a number. You can't buy the reason to come back.
The filler economy was the mechanism that replaced extraction with service. Users post signed Orders — for example, 'one ETH on Ethereum for 3,800 USD on Signet.' Fillers compete to fill those orders at market rates. If ETH is trading at $3,750, a filler who can complete the swap captures the $50 spread. The competition between fillers drives those spreads toward the fair market rate. OrderDetector enforces the atomicity: fills execute first, crediting outputs to the recipient; orders execute second, debiting inputs from the sender. If any fill is missing, the entire order reverts. No partial state. The system's correctness guarantee isn't probabilistic — it's structural.
Block building without auctions changes the character of the chain. Predictability becomes the value proposition instead of extraction efficiency. Applications can reason about when they'll be included. Users see consistent fee structures. The blind sequencer design means the entity responsible for block validity never has access to the information required to front-run or sandwich. This isn't a policy choice or a commitment to good behavior — it's a consequence of how the cryptography is structured. The sequencer cannot extract what it cannot see.
USD-native gas is a small detail that reveals a set of assumptions. Signet's base currency is USD at 18 decimals. ETH deposits convert at market rate on arrival. Users see '$0.02 fee' — not gwei, not a number that requires a lookup to interpret. The denomination choice isn't cosmetic. It reflects a conviction about who the chain is for and what friction is acceptable to impose on them. A user who needs to understand gwei to reason about transaction costs is a user who has been asked to internalize infrastructure details that serve no one except the people who find gwei intuitive. That's a small number of people.
The application primitives that became possible once the timeline separation was removed were the parts of the work that made the underlying architecture feel like it had mattered. Flash.sol enables cross-chain flash loans without requiring a liquidity pool — the atomicity guarantee is the pool. GetOut.sol is instant exit: no withdrawal window, no waiting, capital is liquid when the user needs it to be liquid. PayMe.sol lets an L2 application require that an L1 payment arrive before L2 execution proceeds. PayYou.sol posts execution bounties. These aren't incremental improvements on bridge UX. They're a different model of what cross-chain operations can look like when the primitives are right.
In February 2026, Signet moved to the OP Stack as a rebased rollup. The framing at the time was direct: 'Rollups were supposed to be a bet on Ethereum. We're making that bet.' The rebased architecture means Signet validates correctness during execution, not after — fraud proof windows are eliminated not by shortening the delay but by removing the condition that makes them necessary. The phrase that captured the design philosophy was a precise one: extensions work better than replacements. Ethereum's security, Ethereum's finality, Ethereum's block timing — all of it extended into the rollup layer rather than replicated separately with a gap in between.
There is no Signet token. That sentence carries more strategic content than it appears to. The infrastructure value — the spread captured by fillers, the block rewards captured by builders, the economic activity that flows through the system — goes to the participants in the network, not to a protocol treasury managing a token schedule. It's a bet that infrastructure which captures economic activity sustainably doesn't require a token to create the initial conditions for that activity. The incentive programs and mercenary capital data from 2024 are part of the background for that bet. You can bootstrap numbers with tokens. Whether that produces a network is a different question.
By late 2025, agents were not a future scenario ›agents were fully present tense — they were deployed systems, running capital, making cross-chain decisions at speeds that made human fallbacks a design liability rather than a safety net. Every major agent framework still had those fallbacks built in, but not because the teams wanted them there. They were there because the infrastructure couldn't provide the guarantees required for fully autonomous execution across chains. A cross-chain operation that might fail, might get stuck, might require someone to check in is not a primitive an autonomous system can rely on. The work was never just about making cross-chain faster. It was about making it reliable enough that the human in the loop could be optional rather than mandatory.