BENCH

BENCH is a legal-work simulation environment built to measure capability through work, not credentials. Users step into realistic legal matters, work through documents, evidence and changing information, make judgment calls, and produce the kind of work expected in practice. As they do, BENCH captures how they reason, apply law, handle evidence, adapt to new facts and exercise professional judgment. Its SCALE framework converts those behavioural signals into a structured picture of professional capability, forming the foundation for a lawyer’s Professional Capability Twin. BENCH was conceived and ideated before 21 August 2026, with the underlying product, research direction and simulation-based capability model under development. The Vibeathon submission represents a focused implementation of that broader vision with the protype for a student vertical, using vibe coding (Codex and laude Code) to build and test the legal-work simulation experience.

Matter-Based Legal Simulations — Users work through realistic legal matters with instructions, documents, evidence and changing facts rather than isolated legal questions.

SCALE Capability Framework — BENCH evaluates how a user reasons, applies law, handles evidence, exercises judgment and communicates, not just whether the final answer is correct.

Behavioural Capability Evidence — Actions taken during the simulation become evidence of professional strengths, gaps and working patterns.

Professional Capability Twin — Performance across simulations builds a continuously evolving picture of how the individual actually works as a lawyer.

Integrated Learning & BENCH Guard — Targeted reasoning resources and guardrails sit alongside the simulation so users can improve while maintaining responsible use of AI.

I am looking for three things: access to law firms willing to pilot BENCH with junior lawyers, feedback on what capability signals actually matter in hiring and development, and technical guidance on building the inference layer that turns behaviour across simulations into a reliable Professional Capability Twin.