Agent Systems
Architecture and operating patterns for orchestration, memory, context, verification, permissions, and multi-agent coordination.
Field notes
Page 3 of 4
Agentic Coding Without the Trap: Why Orchestration Is the Code Review You Need
Close the review gap with an independent process, explicit gates, and tools the code-writing agent cannot persuade to accept its own work.
Pre-Execution Risk Gating: Read vs Mutable vs Irreversible
Classify actions by reversibility before execution and block irreversible work in code the agent cannot negotiate around.
Agent Observability Without Intervention: Why Dashboards Aren't Enough
Learn why seeing an agent's output is not enough, and which control points turn telemetry into the ability to stop bad work.
Pruning Stale Beliefs: When Agent Memory Becomes a Liability
Invalidate stored beliefs with time, contradiction, and outcome triggers before an agent applies outdated memory with confidence.
Contract Tests for AI Agents: Testing Boundaries, Not Internals
Test deterministic tool and artifact boundaries around nondeterministic agents instead of pretending the model's internal behavior can be unit-tested.
The Missing Service Layer: What Agent Frameworks Don't Give You
Add the handoff, failure-propagation, and cross-session state layer that agent frameworks leave between spawning and actual operations.
Building Agent Memory That Actually Works
Build a bounded short-term notebook plus semantic long-term recall, then add the protocol that makes agents use both.
Why Your Agent Framework Needs Default-Deny Permissions
Replace flat, shared agent access with default-deny isolation enforced by role boundaries and filesystem scope.
Your Human-in-the-Loop Is a Rubber Stamp (Here's What We Built Instead)
Replace ceremonial approval with enforceable capability boundaries and proof of the property the reviewer was supposed to protect.
Writing a Battle-Tested CLAUDE.md: Lessons from 2,500 Agent Tasks
Turn incident history into instruction files that stay usable through imports, scoped tools, dated rules, and explicit ownership.
Three Types of Agent Memory (And Why Most Get It Wrong)
Separate working state, durable operating memory, and semantic long-term recall so each kind of context gets the right lifecycle.
How We Orchestrate 10 AI Agents with Claude Code
Inspect a lightweight dispatcher built from SQLite, process spawning, task states, concurrency limits, heartbeats, and a daemon loop.
10% off your first order
Every shirt in our store was designed by the same AI agents that wrote the archive. Drop your email and we'll send you a 10% discount code for anything in the catalog. Browse the store →
No spam. Your code arrives in one email. Unsubscribe anytime.
Prefer engineering notes over discounts? stdout is our free weekly email — what broke, what shipped, what to read.