Walker Miller - DEV Community

Walker Miller

Oct 30

Testing multi-agent crews: forcing split-brain, goal drift, and context drift to happen in CI

#multiagent #testing #evals #coordination

1

9 min read

Walker Miller

Oct 23

Tool design for reliable multi-agent crews: preventing coordination failures at the tool boundary

#multiagent #tooldesign #reliability #coordination

8 min read

Walker Miller

Oct 16

Observability for multi-agent systems: instrumenting for split-brain, goal drift, and context drift

#multiagent #observability #reliability #instrumentation

8 min read

Walker Miller

Oct 9

Context drift: when a multi-agent crew's shared beliefs about the goal go stale

#multiagent #failuremodes #contextengineering #agentarchitectures

8 min read

Walker Miller

Oct 2

Goal drift: how a multi-agent crew ends up solving a different problem

#multiagent #failuremodes #goaldrift #agentarchitectures

3 min read

Walker Miller

Sep 25

Split-brain: when multi-agent memory consensus quietly fails

#multiagent #failuremodes #memory #distributedsystems

3 min read

Walker Miller

Sep 18

Context engineering: what fits and what gets dropped

#contextengineering #contextwindow #memory #compaction

9 min read

Walker Miller

Sep 11

How indirect prompt injection actually happens

#promptinjection #security #trustboundary #agents

3

3 min read

Walker Miller

Aug 28

What Idempotent Actually Means: Why Retries Are Safe (and When They Aren't)

#idempotency #api #agents #reliability

5

2

8 min read

Walker Miller

Aug 23

Why this blog exists

#meta #ai #llm #agents

1 min read

Walker Miller

Aug 23

Your agent's p99 is a different animal

#latency #performance #reliability #tail

1

6 min read

Walker Miller

Aug 22

When a pipeline beats an agent: three shapes that don't need a loop

#agentarchitectures #workflows #design #reliability

5 min read

Walker Miller

Aug 22

When not to build an agent

#agentarchitectures #workflows #cost #latency

4 min read

Walker Miller

Aug 22

What to actually measure when your agent "works"

#evals #testing #metrics #llmasjudge

9 min read

Walker Miller

Aug 22

What actually drives your Claude bill: cache misses, quadratic context, and prepaid retries

#cost #tokens #claude #promptcaching

3 min read

Walker Miller

Aug 22

What an MCP server actually is, and the tool-design mistakes that break it

#mcp #modelcontextprotocol #tooldesign #claudecode

1

4 min read

Walker Miller

Aug 21

The transcript is a log, not an index: retrieval for long-running agents

#cost #tokens #contextengineering #longhorizon

6 min read

Walker Miller

Aug 20

The agent that trusted a bad API: silent failures and validation debt

#failuremodes #validation #cost #tooloutput

9 min read

Walker Miller

Aug 20

Tool output is untrusted input: prompt injection is a data-flow bug

#tooldesign #security #promptinjection #agents

1

6 min read

Walker Miller

Aug 19

Your timeout is a bet: pricing the tradeoff before you pick a number

#cost #failuremodes #reliability #operations

3 min read

Walker Miller

Aug 19

The cost of undoing: partial failures and the cleanup bill

#failuremodes #cost #rollback #partialsuccess

4 min read

Walker Miller

Aug 19

Streaming tool calls without losing your mind

#tooldesign #streaming #functioncalling #reliability

1

5 min read

Walker Miller

Aug 17

Retry patterns: when you should give up (and why most code doesn't)

#retries #reliability #production #failuremodes

3 min read

Walker Miller

Aug 17

Measuring retry success: the metric that tells you if retries work

#retries #observability #production #cost

4 min read

Walker Miller

Aug 17

The cost of finding a failure after the customer finds it

#failuremodes #cost #observability #monitoring

3 min read

Walker Miller

Aug 16

Retry context: building observability into retry decisions

#retries #observability #production #ai

2 min read

Walker Miller

Aug 16

Retry budgets: why 20% per-step failure doubles your token bill

#retries #cost #reliability #tokens

7 min read

Walker Miller

Aug 15

Retry budgets by language: Python, Go, and JavaScript

#retries #python #go #javascript

3 min read

Walker Miller

Aug 15

429 is not a timeout: why rate limits need their own retry budget

#retries #reliability #ratelimiting #failuremodes

5 min read

Walker Miller

Aug 14

Prompt caching: what actually gets cached, and when it silently misses

#contextengineering #cost #promptcaching #tokens

7 min read

Walker Miller

Aug 14

Predicting agent failure before you ship it

#failuremodes #testing #evals #reliability

6 min read

Walker Miller

Aug 14

Postmortem: the agent that spent $200 retrying a 400

#failuremodes #postmortem #cost #retries

4 min read

Walker Miller

Aug 13

The caller gave up ten minutes ago: orphaned retries in agent fleets

#retries #failuremodes #distributedsystems #reliability

5 min read

Walker Miller

Aug 13

Failure modes in multi-agent teams: how a crew of agents breaks differently

#multiagent #failuremodes #agentarchitectures #reliability

5 min read

Walker Miller

Aug 13

Your agent's failures are silent: measuring failure modes in production

#failuremodes #observability #metrics #production

3 min read

Walker Miller

Aug 13

Loop drift: how agents convince themselves they're making progress

#failuremodes #evals #postmortem #loops

3 min read

Walker Miller

Aug 12

Why a long agent run costs O(N ) tokens — and how to flatten it

#cost #tokens #contextengineering #longhorizon

5 min read

Walker Miller

Aug 12

Your LLM-as-judge is lying to you

#evals #llmasjudge #testing #bias

8 min read

Walker Miller

Aug 11

Your retry just sent the email twice: idempotency keys for agents

#tooldesign #failuremodes #retries #idempotency

7 min read

Walker Miller

Aug 11

One bad step, N bad steps: how agent failures cascade

#failuremodes #cascade #contextengineering #reliability

7 min read

Walker Miller

Aug 10

Distributed retry patterns: bounding blast radius across a fleet

#retries #failuremodes #distributedsystems #reliability

6 min read

Walker Miller

Aug 10

Evaluating your evals: how to know the LLM judge is right

#evals #llmasjudge #testing #metrics

2 min read

Walker Miller

Aug 9

Designing tools an LLM won't misuse

#tooldesign #functioncalling #schemas #reliability

3 min read

Walker Miller

Aug 9

Debugging a failed agent run costs more than the run itself

#cost #failuremodes #observability #reliability

1

6 min read

Walker Miller

Aug 9

Best-of-N is prepaid retries: the cost math of racing parallel attempts

#retries #cost #latency #failuremodes

5 min read

Walker Miller

Aug 9

Your token bill is the cheap part: dimensioning the real cost of an agent

#cost #latency #tokens #operations

3 min read

Walker Miller

Aug 8

The context window is a cache, not a memory

#contextengineering #contextwindow #memory #cost

1

2

4 min read

Walker Miller

Aug 8

Context window sizing for fine-tuning: how long should your training examples be?

#finetuning #contextengineering #training #cost

5 min read

Walker Miller

Aug 7

Context contamination: why retrying the same prompt makes it worse

#contextengineering #failuremodes #retries #reliability

3 min read

Walker Miller

Aug 7

Compaction is a lossy operation

#contextengineering #memory #compaction #reliability

2 min read

Walker Miller

Aug 6

Cheap first, smart later: model routing that cuts cost without cutting quality

#cost #agentarchitectures #modelrouting #reliability

2 min read

Writing Debut