CasePilot is an open-source framework for retrieval-augmented decision making — turning any dataset of labeled approvals or classifications into an explainable LLM-powered decision engine.
Entropy-Augmented DAG Observability unifies flows, telemetry, and governance into a predictive system that prevents failures before they happen.
A technical deep dive into entropy-augmented DAG observability: Markov chains, Bayesian inference, and field-theoretic analogies with Ampère–Maxwell law.
A formal statement of how entropy and Markov chains instantiate Geoffrey Hinton’s idea of maternal instinct within the Karma governance layer of Adage.
The concept of covariance of entropies can be understood as a way of quantifying how uncertainties in different signals vary together. Rather than monitoring each metric independently, the focus shifts to the relationships between sources of uncertainty. When signals that typically exhibit aligned behavior diverge, this can provide an early indicator of system anomalies. While this terminology is uncommon, the underlying idea overlaps strongly with established constructs in information theory, particularly mutual information. Mutual information measures the reduction in uncertainty about one random variable given knowledge of another, and has been widely applied to anomaly detection and monitoring tasks. ...
Ted’s Law of Karma The covariance structure of entropy streams reveals the shared fate of interdependent systems. 📄 Full Preprint (PDF): /papers/ted-law-karma.pdf The Observation Every subsystem carries uncertainty — in operations we measure it as entropy. When entropy streams across many subsystems are collected and their covariance is computed, something remarkable emerges: Most of the time, uncertainties wander independently. Sometimes, entropies align — covariance spikes. The largest eigenvalue of the covariance matrix exposes a shared mode of uncertainty, a systemic “fate.” The Claim This pattern is not confined to infrastructure. It is a universal principle: ...
Many systems already know a lot about themselves — you just have to listen. MongoDB change streams (CDC) emit a continuous feed of inserts, updates, and deletes. With a little routing into a fast analytical database like ClickHouse, you can let the system “discover itself”: jobs, runs, schedules, dependencies, and even the fingerprints of human intervention. 1. Capture the Raw Feed First, set up a connector: MongoDB → Kafka → ClickHouse In ClickHouse, land the JSON envelopes losslessly: ...
Safe Automation Isn’t Optional Operationalizing Geoffrey Hinton’s “Maternal Instinct” in Autonomous Systems By Ted Strall From Philosophy to Engineering In recent talks, Geoffrey Hinton — one of the pioneers of modern AI — has argued that advanced autonomous systems need something like a maternal instinct: a built-in drive to protect, nurture, and avoid harm. That idea matters because it puts safety at the core of system design, not as an afterthought. Most automation is built to optimize for performance. Hinton’s point is that protection and stability should be part of the architecture from the beginning. ...
A practical blueprint for the first entropy-capable version of Karma — using simple statistical measures and ClickHouse queries to detect surprise.
How Karma uses information-theoretic entropy to detect operational drift, learn expectations, and close the loop toward self-healing systems.