How Argumatix thinks about data, causality, and trust. If we haven't answered your question here, ask at hello@argumatix.com — we read everything.
Exclusively from public federal sources. Opinions and dockets from CourtListener and the RECAP archive. Metered PACER pulls for the gaps. Regulatory context from SEC EDGAR, FDA enforcement, and CFPB actions. No proprietary legal-research feeds, no scraped third-party databases, no black-box licensing. Because everything is public, every claim we surface is independently auditable.
No. Argumatix is a causal-inference and structured-ML platform, not a writing assistant. Large language models are used only at narrow, schema-enforced extraction boundaries — never for free-form generation that reaches a user-facing claim. Every output that names a statistic, a judge's tendency, or a causal effect is backed by a structured record in our database and gated by an independent truth-checker before it ships.
No — we complement them. Legacy research platforms are built around finding the law; Argumatix is built around a different question: given your judge, your opposing counsel, your case posture, and your brief, what will actually happen — and what should you do differently? It's a predictive and causal layer on top of the research layer, not a substitute for primary-law access.
Never. Your briefs, drafts, and case work stay in your workspace and are never used to train any model that another tenant can benefit from. Public-record data powers our predictive models; your work product powers only your analyses. Data is encrypted in transit and at rest, tenants are isolated by design, and nothing you upload becomes training data for anyone else.
Three layers. First, LLMs are scoped to narrow extraction tasks with schema validation and byte-exact span grounding — no free-form generation touches a user-facing claim. Second, every claim must resolve to a source row in our database before it can be rendered — an independent truth-checker runs on every output and fails closed. Third, nightly negative-control telemetry runs synthetic outcomes through the full pipeline to catch drift before any real user sees it. Confidence below threshold? The claim is suppressed, not dressed up.
Initial steps and guidance to help you begin using our services seamlessly.
Information on managing your account and maintaining security.
Details on payment options, billing cycles, and handling transactions.
