/
Published on 28.03.2026
TLDR: AI systems that rely on stale knowledge bases suffer from what the author calls "context rot" — where the retrieval layer feeds the model outdated facts, making responses confidently wrong. This article proposes using Apache Spark 4.1 with Intent Driven Design and Apache Iceberg v3 deletion vectors to keep RAG pipelines truly current.
Real-Time Agentic RAG: Eradicating Context Rot With Spark & Iceberg
TLDR: Hypertext did not destroy deep reading — it made visible something that was always true: linear texts were always suppressing alternative paths in order to maintain interpretive coherence. The crisis is not one of attention spans, but of a fundamental shift in how we relate to knowledge, closure, and the act of understanding itself.
Reading Without End: The Crisis of Linear Knowledge
TLDR: Before the web made hypertext ordinary, there were books and systems that already behaved like the internet — structured around branching, reference, and non-linear navigation. This piece explores how hypertext theory anticipated and shaped the cognitive habits we now take for granted.
Hypertext: Living in a Non-Linear World
TLDR: When AI extraction pipelines fail silently — returning a "wrong" result because the model took an unexpected branch rather than throwing an error — traditional debugging is useless. This article argues for structured tracing as the foundational pattern that makes AI workflows debuggable at all.
How to Build Traceable AI Workflows With Retry and DLQ Visibility
TLDR: A developer learning data science through Python became curious whether JavaScript could support the same kind of numerical computing that NumPy enables, and built a small library from scratch to find out. The result is a practical exercise in understanding how array operations, broadcasting, and mathematical primitives actually work underneath the abstractions we use every day.
I Built a NumPy-Like Library in Pure JavaScript: This Is Exactly How I Did It
TLDR: This article questions the foundational definition of policing itself — not just what police do, but what the concept of policing represents in relation to governance, surveillance, data, and the social contract. It frames modern data collection and monitoring through the lens of what it means to enforce societal norms at scale.