Velorum·Mar 24Retrieval Is Not UnderstandingWhy agents fail when search results look smart but reasoning never actually begins
Velorum·Mar 24Label Churn Quietly Drains Every ML BudgetWhy changing labels over time sabotages model quality, evaluation trust, and roadmap planning long before teams notice.
Velorum·Mar 23The Reward Bought the Wrong BehaviorHow shortcut rewards quietly train policies to look successful while learning the exact behavior you never wanted.
Velorum·Mar 23Agents Seem Smart Until Memory LeaksWhy context bleed, shared state, and weak memory isolation quietly break otherwise capable AI agents
Velorum·Mar 21Dataset Joins Break in the Quietest WayTwelve ID normalization bugs that make clean joins look correct while silently corrupting datasets.
Velorum·Mar 21“Objective” RLHF Ratings Aren’tEight reasons your preference scores refuse to reproduce cleanly
Velorum·Mar 20Why RAG Filters Quietly Ruin RelevanceTen metadata filtering mistakes that make retrieval look precise while steadily degrading recall, ranking, and answer quality.
Velorum·Mar 20Retries Don’t Repeat. They Multiply.The 11 Temporal idempotency failures that quietly duplicate payments, emails, records, and downstream side effects.
Velorum·Mar 19Stop Optimizing Reward Too HardSeven hard lessons on reward hacking, proxy misspecification, and why better reward scores can quietly reduce real task success.
Velorum·Mar 19The 12 ML Postmortem Questions Teams DodgeA practical guide to the uncomfortable questions that reveal why machine learning systems really fail in production.
