Skip to content
BLOGField notes from CompozyOS

Practical engineering for working with AI agents.

Experiments, guides, and reusable tools for reviewing agent work, recovering interrupted tasks, and building automation you can inspect.

All08Protocol00Runtime04Engineering04Network00RSS
LATESTNewest first
Engineering

Agent Frameworks: Find the Failing Boundary

Decide which part of your agent architecture needs to change. Work through approval, retries, and duplicate delivery with a failure drill and decision worksheet.

pedronauck7 min
Runtime

Coding Agent Task Briefs You Can Actually Review

Write a coding agent task brief with concrete acceptance criteria, staged permissions, and a review record. Work through a pagination bug from prompt to proof.

pedronauck7 min
Engineering

How to Compare Coding Agents on Your Repository

Compare Cursor, Claude Code, or other coding agents on your repository with a shared baseline, reusable task brief, and scorecard for correctness and review time.

pedronauck7 min
Runtime

Agent Session Recovery: What Survives a Stop?

Recover an interrupted AI agent session in CompozyOS. Inspect process state, preserve edits, and choose a continuation path with a diagnosis tree and handoff.

pedronauck7 min
Engineering

Build a Git Repository Briefing with Python

Build a useful Git repository briefing with a read-only Python collector, explicit commit boundaries, and a sourced review queue. Includes a runnable script.

pedronauck7 min
compozy-network/v0, three peers exchanging direct, receipt, and trace envelopes
Runtime

Introducing CompozyOS

CompozyOS v0.3 brings sessions, work, memory, permissions, automation, coordination, and an OS shell into one extensible local-first runtime.

pedronauck3 min