Legalcomplex designs, builds and verifies AI products for legal, tax and regulated work, and runs its own in public. The same design and engineering is available to your team on a monthly engagement: architecture, retrieval, AI integration, and the evals that prove it works.
Start a conversationEvals gated on committed runs, and checks built so an absent input can never read as a pass. The S3 benchmark sweeps 2,716 citation statements past four models, and every one lands near chance without sources.
See the benchmark results →The changelog is generated from the commits, monthly. The health endpoint reports the commit actually running. Nothing here is a claim you cannot check.
Read the changelog →More than 31,900 profiles with funding history, kept current by a daily sync from the Spark dataset. Check yours, and what your segment raised last quarter.
Open your profile →Tell us what you are building and where it is stuck. A free 15 minute intro chat follows, and if it is not a fit we say so quickly.
Everything here is required. We read every application ourselves.