Engineering notes forAI agent governance.

Long-form technical writing from the open-source MARGINAL project. We focus on the mechanics that can be inspected: no-progress detection, runtime evidence, privacy boundaries, false stops and the conditions under which a governor should—or should not—interfere with an agent.

Evidence first
Open source
Falsifiable claims
No marketing math

Latest

Start with the failure mode, then inspect the mechanism.

Each article separates what the runtime can observe from what it can infer, and links product claims back to public code, documentation or evidence.

Diagram-style cover for detecting AI agent no-progress without reading prompts

AI agent observabilityAug 19, 2026

Detecting No-Progress in AI Agents Without Reading Prompts

Why repeated tool calls are not automatically waste—and how action identity, outcomes, state and evidence can expose a stronger no-progress signal without making raw prompts required telemetry.

Read the technical article →

Diagram-style cover for evidence-bound authority in AI agent governance

Runtime governanceAug 19, 2026

Why an AI Agent Governor Should Start Powerless

Installation is not evidence. A governor that can block useful work should observe first, earn narrow authority from local evidence, and lose that authority when its assumptions drift.

Read the architecture argument →

Editorial scope. This publication describes engineering work behind MARGINAL. Product behavior is checked against the public implementation and documentation at publication time. Negative, ambiguous and unsupported outcomes are kept explicit rather than rewritten into performance claims.

Explore the system

Engineering notes forAI agent governance.

Start with the failure mode, then inspect the mechanism.

Detecting No-Progress in AI Agents Without Reading Prompts

Why an AI Agent Governor Should Start Powerless

From search intent to inspectable evidence.