Sign in Subscribe

Topic

framework

A synthetic, multi-part model for understanding a complex domain. Frameworks are the blueprints that connect multiple principles to explain why things matter.

The Agent Failure Bestiary: Sycophancy, Drift, and Recursion Loops

Most agent failures look novel when they happen. They are usually not. They belong to a small recurring bestiary of failure patterns that appear whenever stochastic systems are given too much authority and too little structure.

The Cognitive Firewall: PII Scrubbing Is Not a Feature. It's Part of the Runtime.

PII scrubbing is not a safety feature you bolt on later. If an LLM can see private data and untrusted instructions in the same token stream, your runtime boundary is already broken.

The Utilization Paradox: Why a 70B Model Can Be Cheaper Than an 8B Model

Most teams assume smaller models are cheaper by default. In production inference, that is often false. The real cost driver is utilization: how much useful work each GPU hour actually carries.

Why More Context Can Make an LLM Worse

The default response to agent failure is to stuff more context into the prompt. That often makes the system worse. A context window is working memory, not a hard drive.

The Event-Driven Operator: Agents That Wake on Events, Not Cron

Polling-based agents fail by construction — not occasionally, but structurally. The event-driven operator pattern eliminates that failure class by separating wake-up, execution, and recovery.

The Prompt Is a Hyperparameter

Manual prompt engineering is optimization without an objective function. DSPy reframes prompts as hyperparameters, programs as signatures, and optimization as a first-class operation — and in doing so, shifts the core skill from prompt craft to metric design.

Text-First Extraction Is an Epistemic Error

RAG pipelines degrade in a pattern consistent enough to be structural: a text extractor destroys semantic layout before a single token reaches the model. The failure is not in the retriever. It is at the extraction boundary — where two-dimensional structure gets linearized into noise.

Your Company's "Agentic Surface Area": The New Metric for Competitiveness

Your CEO asks: "How much of our operation is AI-powered?" The uncomfortable part is that the question sounds simple and usually has no clean answer. Teams can name pilots, copilots, and workflows, but not how much decision-making has actually moved to agents. Security teams have "attack surface

Intelligence Arbitrage: The Economics of Routing Thought

The default LLM architecture is economically incoherent: one expensive model for every task. Intelligence Arbitrage treats model selection as a routing problem and turns the spread between task difficulty and model capability into operating leverage.

The Iron Triangle of Inference: You Can't Have Smart, Fast, and Cheap

In distributed systems, the CAP Theorem dictates you can only have two of Consistency, Availability, or Partition Tolerance. In AI, we face a new, equally brutal constraint: The Iron Triangle of Inference. You must choose between Quality, Latency, and Cost. You cannot have them all.