When AI Can Solve But Can't Search: The MathNet Equation
MathNet shows why enterprise AI systems need structure-aware retrieval, not just stronger reasoning models with more context pasted on top.
Cognaptus Insights
Industry thoughts, trends, and commentary.
When RL Needs a Tour Guide: OGER and the Business of Smarter Exploration
A mechanism-first reading of OGER, showing why expert demonstrations become more valuable when they guide exploration instead of merely supplying imitation data.
WorldDB Memory Wars — Why Agent Memory Needs Structure, Not More Tokens
WorldDB argues that agent memory is not a bigger-context problem but a state-management problem: identity, time, provenance, and write-time rules need to be built into the memory layer.
CQ or Consequences: What This LLM Benchmark Reveals About AI Requirements Work
A comparison-based reading of CompCQ shows why LLM-generated requirements work needs model portfolios, not one-model faith.
CQ, AI & The Question of Questions
A controlled comparison of human, template, and LLM-generated competency questions shows why AI can accelerate requirements elicitation without replacing expert judgment.
Graph RAG, No Smoke: Why Explainable AI in Manufacturing Needs a Memory
A mechanism-first reading of how knowledge graphs and LLM-guided retrieval can make machine learning explanations in manufacturing more contextual, useful, and governable.
Lost in the Grid: Why AI Agents Still Can’t Spot the Impostor
SocialGrid shows why agent reliability depends less on model eloquence than on separating navigation, execution, and behavioral inference failures.
MARCH Orders: When AI Holds a CT Case Conference
A mechanism-first reading of MARCH, a multi-agent CT report-generation system, and what its hierarchy teaches enterprise AI about review, grounding, and controlled disagreement.
Silent Errors, Loud Consequences: ASMR-Bench and the Coming Era of AI Auditors
A research-sabotage benchmark shows why AI auditability is not a code-review feature, but an operating model for trustworthy AI work.
When AI Learns the Trick First: Why Insight Beats Brute Force in Theorem Proving
A mechanism-first reading of why explicit technique recognition may matter more than longer reasoning traces for informal theorem proving and enterprise AI workflows.