earlyink
104 watching first flagged14 Oct 2026

Contract Diff: behavioural tests for AI-written Solidity before deploy

CI that differential-tests AI-generated smart contracts against a reference before deployment. Stablecoin and tokenisation teams pay.

Competitive · SolBehave shows compiling isn't correctness

Written for: Small startup team

Next step

Call the CTO of a seed-stage stablecoin wallet startup building on embedded-wallet infrastructure. Ask what share of their contract changes a coding agent writes now and what they run between merge and audit.

Why now

Embedded stablecoin wallets (Rain) and x402 payment rails (Privy on XRPL, Laravel middleware) mean more small teams are shipping money-moving contract code, often written with coding agents. Audits are slow and expensive, and the agent writes faster than auditors can read. A CI-level behavioural check fits between commit and audit. Demand is unproven, though, and audit firms own the relationship.

What became possible

SolBehave evaluates AI-generated Solidity on 250 real Ethereum contracts with differential and behavioural tests, not just compilation. That gives a reusable method for catching faults that can never be patched after deployment.

Two ways in

  1. Diff CI

    Runs agent-edited contracts side by side with the previous or reference version on fuzzed transactions and flags any change in behaviour.

    Customer
    Engineering leads at stablecoin wallet and tokenisation startups
    Model
    $500 to $2,000 per month per repo
  2. Pre-audit Report

    A behavioural test report that audit firms accept as input, cutting audit time.

    Customer
    Smart-contract audit firms, sold as a partner channel
    Model
    Per-report fee

Competition

Competitive

The audit-firm partner channel looks weaker than it sounds. Several of the closest players are audit firms themselves or are tied to one: Trail of Bits, Cyfrin, Nethermind, and Recon, which also sells to auditors. They would rather use their own tooling than accept a third-party report that cuts their billable hours. I found no sign that any audit firm accepts outside behavioural reports as standard audit input.

Full competition notes (8 competitors)

The broader category of AI security and testing in CI for Solidity is crowded and well funded. Olympix and Octane already sell into GitHub CI and are both listed as Circle partners, so they are already in front of stablecoin teams. Certora has an open-source product for checking AI-written contracts, and Recon offers cloud fuzzing in CI with AI-generated test suites. The narrower idea still looks open: automatic behavioural-equivalence checks between an agent-edited contract and a reference version. The only purpose-built tool I found, Trail of Bits' Diffusc, was archived in February 2025. None of the commercial products I found market old-versus-new differential fuzzing as their main feature. They look for vulnerabilities or check properties that users write themselves. A new entrant could win by making differential testing work with no setup. That means automatically building the harness that runs the old and new versions side by side, forking the deployed version on-chain as the reference, and treating intended changes as an explicit approved list so that every other change in behaviour fails the build. It could also target what stablecoin and tokenisation teams specifically worry about: mint, burn, freeze, blacklist, role and upgrade-storage behaviour, and how compliance hooks behave across upgrades. The trigger for each check would be an AI agent's pull request, not a manual upgrade. The risk is that this becomes a feature. Olympix, Octane, Recon or Certora could add a diff mode quickly, and Foundry users can write differential harnesses by hand. The audit-firm partner channel looks weaker than it sounds. Several of the closest players are audit firms themselves or are tied to one: Trail of Bits, Cyfrin, Nethermind, and Recon, which also sells to auditors. They would rather use their own tooling than accept a third-party report that cuts their billable hours. I found no sign that any audit firm accepts outside behavioural reports as standard audit input.

  • Diffusc (Trail of Bits / crytic)

    The closest match to Diff CI. It is open source, uses Slither and Echidna to differential-fuzz two upgradeable contract implementations before an upgrade, and offers standard, fork and hybrid modes. The repo was archived in February 2025 and is now read-only, so nobody is maintaining it.

  • Olympix

    A funded proactive security suite with static analysis, unit test generation, mutation testing, fuzz test generation and PoC generation, running in VS Code and GitHub CI. It is listed as a Circle partner, which puts it directly in front of stablecoin teams. It does not market old-versus-new behaviour diffing.

  • Octane

    AI security platform that raised $6.7M, led by Winklevoss Capital and Archetype. A GitHub App analyses every pull request for more than 50 vulnerability classes. It is also a Circle partner and is integrated with Avalanche. It looks for vulnerabilities rather than diffing behaviour.

  • Certora AI Composer

    Open-source alpha launched in November 2025 that puts the Certora Prover inside the AI code generation loop, so AI-written Solidity is checked against formal properties before it ships. It is aimed squarely at safe AI-written contracts, but it depends on properties the user writes rather than a reference implementation.

  • Recon (Recon Pro / Chimera)

    Cloud invariant testing and fuzzing for EVM contracts, with CI/CD campaigns and AI-generated fuzzing suites (Recon Magic), built on the open-source Chimera framework. Its own figures list clients including Liquity, Centrifuge and Balancer. A team could write differential harnesses on it, and it is also an auditor-side player.

  • Nethermind AuditAgent

    AI audit scanner from Nethermind, which is also an audit firm. A GitHub Actions workflow and REST API can fail builds on new critical findings. It is a scanner, not a differential tester.

  • Aderyn CI Assistant (Cyfrin)

    A free GitHub Action for static analysis of Solidity in CI, from the audit firm Cyfrin. It is the baseline that buyers already get for free.

  • Matterhorn / ASI Alliance

    Announced tools for auditing AI-generated (vibe-coded) blockchain code before deployment, combining automated analysis, human review and testing. I could not confirm that it has shipped.

Checked 2026-10-14

Risks

  • Foundry and existing fuzzers already cover much of this for skilled teams
  • Audit firms may see it as a competitor rather than a channel
  • Crypto developer budgets swing with market cycles

Best on buildability; market pull is the weak point (4).

tech 6 · pull 4 · space 5 · fit 6 · build 7

Exposed

Companies that could benefit or be threatened. News and context, not investment advice.

Momentum of its main topic

54 developments this week +22 vs last week
21 Sep: 0 developments22 Sep: 0 developments23 Sep: 0 developments24 Sep: 0 developments25 Sep: 0 developments26 Sep: 0 developments27 Sep: 0 developments28 Sep: 0 developments29 Sep: 0 developments30 Sep: 0 developments1 Oct: 0 developments2 Oct: 0 developments3 Oct: 0 developments4 Oct: 0 developments5 Oct: 0 developments6 Oct: 0 developments7 Oct: 0 developments8 Oct: 13 developments9 Oct: 4 developments10 Oct: 7 developments11 Oct: 0 developments12 Oct: 4 developments13 Oct: 4 developments14 Oct: 7 developments15 Oct: 11 developments16 Oct: 10 developments17 Oct: 8 developments18 Oct: 5 developments19 Oct: 6 developments20 Oct: 7 developments
Stage: Coverage

Sources

  1. SolBehave: Evaluating the Behavioural Correctness of LLM-Generated Smart ContractsarXiv
  2. Rain launches stablecoin walletFinextra
  3. Ripple, Stripe-owned Privy partner to bring embedded wallets to XRP LedgerThe Block
  4. requestway/laravel-x402GitHub

Status history

  • 2026-10-14 Added to the watchlist