The Observability Stack

A curated technical publication surface for essays, frameworks, and labs on observability for complex compute systems — from silicon to autonomous agents.

A structured body of work, not a newsletter feed.

Each artifact is meant to be read, cited, or inspected. Published work comes first; research arcs are labeled honestly by maturity.

Read the flagship essay Frameworks Labs

Core pillars

Four pillars, separated by maturity.

Core practice

Silicon observability

State capture, scan access, and debug platforms for AI/HPC silicon.

Published framework

Compute economics

The Cost of Usable Intelligence — useful output under real constraints.

Emerging arc

Agent observability

Tracing, replay, escalation, and failure taxonomy for autonomous systems.

Emerging arc

Quantum-classical

Accelerated classical control planes for fragile quantum systems.

Featured artifacts

Public work you can read, cite, or inspect.

Flagship essay

Scan, fault-injection, and LCI demos

Frameworks

Reusable models for complex-system observability.

Agent Debuggability Stack

Six inspection surfaces every agent platform should expose around consequential action.

Observability maturity model

Levels 0–5, from final-output-only to failure-taxonomy-and-eval-integration.

Silicon → agent mapping

How bring-up debug principles translate to agent observability.

Publication roadmap

Published artifacts first. Future directions second.

Published

Shipped

—Observability safety primitive (essay)
—Debuggability for autonomous agents (essay)
—The Cost of Usable Intelligence (paper)
—Scan / fault / LCI labs

Planned

—Agent trace schema
—Escalation eval suite
—Agents as infrastructure risk
—Failure-mode taxonomy