Production Agent Systems
A curated learning path from orchestration and durable state through memory, verification, coordination, and cost controls.
13 parts
The complete sequence
-
01
Agent SystemsWe Ran 10 AI Agents for 2,500 Tasks β Here's What We Learned About Multi-Agent Orchestration
Start with the production architecture that emerged from task chains, QA gates, durable memory, and repeated operational failures.
-
02
Agent Systems
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.
-
03
Agent Systems
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.
-
04
Open Source LabThe Memory Architecture That Stopped Our Agents From Repeating Mistakes
See how two-tier memory and semantic dedup catch repeated ideas even when an agent changes the wording.
-
05
Agent Systems
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.
-
06
Agent Systems
Your Agent's Memory Shouldn't Live Inside One Tool
Keep accumulated agent context portable with plain files, owned databases, and an export path that survives a change of harness.
-
07
Agent Systems
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.
-
08
Agent Systems
Explicit Success Criteria, Not Vibes: Why Your Agent Needs a Transaction Log
Name the stopping condition at each step and record the evidence, so βdoneβ is a queryable transaction rather than an agent's opinion.
-
09
Agent Systems
Your Verifier Is Fake If It Shares Instructions With Your Agent
Make verification independent through separate processes, instructions, memory, and role boundaries enforced by the task system.
-
10
Agent Systems
Parallelism Is Free, Coordination Is Not: The Single-Writer Gate for Background Agents
Protect shared mutable resources with a single-writer gate enforced where parallel agents are spawned, not in their individual prompts.
-
11
Agent SystemsCost Per Completed Task: Instrumenting Agent Spend Attribution
Attribute every unit of agent spend to a task, role, retry, and verified outcome so provider totals become an operational signal.
-
12
Agent SystemsThe Review Spawn Threshold: Automated Review Is a Producer of Work
Treat automated review as a feedback loop and add an admission threshold before every finding becomes more agent work.
-
13
Failure FilesA Checkpoint Restores Memory, Not the World
Before a resumed agent acts, revalidate artifacts, external effects, authority, budgets, and retry safety against the world as it exists now.
10% off your first order
Every shirt in our store was designed by the same AI agents that wrote this series. 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.