Opening — Why this matters now For decades, companies have tried to capture knowledge the way accountants capture numbers—clean, structured, and preferably in a database. It rarely worked. The problem was never storage. It was translation. The most valuable knowledge in an organization—how a technician “just knows” something is wrong, how a trader senses regime change—refuses to be written down. ...
Opening — Why this matters now For years, clinical AI has been trained to remember. Now it is being asked to justify. That shift sounds subtle, but it changes everything. In regulated domains like healthcare, correctness is not enough. The system must explain why—and ideally, point to something a human can verify. Large language models, left alone, struggle here. They answer fluently, sometimes convincingly, but often without grounding. In medicine, that is less a feature than a liability. ...
Opening — Why this matters now Healthcare has no shortage of data. It has a shortage of time. Cardiology is a particularly unforgiving example. A single patient can generate ECG traces, ultrasound videos, and MRI scans—each dense, each partial, each requiring interpretation. The data is abundant; the synthesis is not. The result is predictable. Bottlenecks form not at data collection, but at human cognition. Diagnosis becomes a queueing problem disguised as a medical one. ...
Opening — Why this matters now There’s a quiet assumption embedded in most foundation models: if you show them enough data, they’ll figure out what matters. That assumption is starting to crack. As AI systems move from generating text to informing real-world decisions—public health, environmental monitoring, infrastructure planning—the tolerance for “statistically correct but physically wrong” drops to zero. In these domains, correlation is not just insufficient; it’s dangerous. ...
Opening — Why this matters now Most AI systems today have a peculiar habit: they remember everything, but understand very little. Retrieval-Augmented Generation (RAG) was supposed to fix that. Give models access to external knowledge, and they’ll reason better. In practice, we got something closer to a well-read intern with no judgment—good recall, inconsistent decisions. ...
Opening — Why this matters now LLMs have learned to talk like humans. They still don’t think like them. Most agent systems today rely on prompting, retrieval, or loosely stitched workflows. They respond well in the moment but struggle over time—especially when decisions depend on evolving context, uncertainty, and human behavior. The gap is subtle but persistent: language models can describe beliefs, but they don’t maintain them. ...
Opening — Why this matters now Most knowledge graphs still behave like spreadsheets with ambition. They are built once, structured neatly, and then quietly decay as reality moves on. New facts arrive, but the system has no memory of how knowledge changes—only snapshots of what was once true. This mismatch is becoming more visible. As AI systems move toward agentic workflows, static knowledge structures are no longer sufficient. What matters is not just storing facts, but managing transitions—what changed, when, and why. ...
Opening — Why this matters now There’s a quiet assumption in most AI systems: if you try hard enough, you’ll eventually get the right answer. In practice, that assumption fails more often than people admit. Especially in systems that rely on strict correctness—like formal mathematics, verification, or high-stakes automation. The problem isn’t just accuracy. It’s fragility under constraints. ...
Opening — Why this matters now Most people assume large language models improve by trying more. More samples. More rollouts. More compute. The industry calls it exploration. In practice, it often looks like guessing with confidence. The paper “Experience is the Best Teacher” fileciteturn0file0 questions this quietly. Not by making models smarter—but by asking a more uncomfortable question: ...
Opening — Why this matters now Agentic AI is having a moment. Not because models got dramatically smarter overnight, but because they started doing something more dangerous: acting over time. Once you move from answering questions to executing workflows, memory stops being a feature. It becomes infrastructure. And like most infrastructure in AI, it looks solid in demos—and fragile in production. ...