Pre-launch · building · devnet target Q2 2026read the notes
LigateLabs
Back to notes
April 15, 2026·6 min read·Ligate Labs

Why we're building a sovereign rollup for AI provenance

Verifiable AI does not live on a general-purpose chain. It lives on a chain where the unit of work is a prompt.

The instinct to build AI provenance as a smart contract on an existing L1 is strong. It is also wrong. The cost structure, the latency envelope, and the opcode surface of an EVM or Solana chain were designed for financial transactions, not for the thing that actually happens a billion times a day: a human types a prompt and an AI answers.

Ligate is a sovereign rollup built on the Sovereign SDK, using Celestia for data availability. That choice makes three things cheaper and three things possible that otherwise are not.

Three things that get cheaper

Gas. Posting a Proof of Prompt receipt to Ligate costs roughly $0.0009 of $LGT. The same receipt posted as calldata to an L1 would be dollars, which is fine for a wire transfer but not for a ChatGPT message.

Latency. Our sub-200ms attestation target only works because the chain is tuned for one transaction shape. On a shared L1 you are queued behind every NFT mint and failed swap.

Infrastructure. Running a sovereign rollup on Celestia DA costs us somewhere between $110 and $250 per month at launch. That is not a typo. Modular chain architecture has collapsed the cost of running chain-as-a-service to single-digit cloud bills.

Three things that get possible

First-class AI transaction types. We can add opcodes for SubmitPrompt, ClaimProof, CreateBattle instead of shoehorning everything into generic CALL semantics. The gas cost of a prompt submission is a line item, not a function of how clever a Solidity developer was five years ago.

Redactable receipts. With SP1 / RISC Zero in Phase 2, we can prove a prompt was submitted at time T to model M by wallet W, without revealing the prompt or the output. Enterprises using AI on trade secrets actually need this. You cannot bolt it onto a chain that already committed to its transaction schema.

One token, shared security. Because Ligate hosts two native products (Themisra for AI provenance, Kleidon for Web3 infrastructure), they share validator set, liquidity, and governance. No fragmented L2 nightmare.

The honest caveat

Sovereign rollups are young. Sovereign SDK is not yet battle-tested at Ethereum or Solana scale. We are betting that the architecture is right and the maturity will follow — the same bet Celestia made on modular DA in 2022, and Cosmos made on app-chains in 2018. It is a bet. If you want a production-proven chain today you use Ethereum. If you want the right substrate for AI provenance five years from now, that is us.

The devnet targets Q2 2026. If this sounds interesting to build on, run a node.