Research
Irys University
Research essays from the team building Irys on how legal AI should work. Writing on auditability, matter context, source-backed answers, and reliable work product for legal teams.
ResearchReinforcement Learning Is the Wrong Bet: Four Structural Failures and What Works Instead
ResearchActive Inference & The Free Energy Principle: From First Principles to Working Systems
ResearchGLM 5.2: How Four Co-Designed Mechanisms Beat Raw Scale
ResearchWhat AlphaEvolve Teaches Legal AI About Evaluated Work
ResearchHow to Train Graph Neural Networks: From First Principles to Planet Scale
ResearchAuditability Is The Product In Legal AI
ResearchWhy Prompt-As-Memory Fails In Legal AI
ResearchMatter Memory Must Preserve Contradictions
ResearchPrompt Instructions Are Not Control In Legal AI
ResearchThe Exception Leak In Legal AI
ResearchA Citation Is Not An Audit Trail
Research