agentlist

A field guide to working with agents.

Pick the right tool. Give it a clear job. Know when the work is done.

Ten illustrated guides for the wonderfully practical business of getting useful work out of AI. Read a little. Try something. Take it into your next task.

Start with your first agent Explore the fieldbook
A compass for the agent wildernessNyour next movea good place to start
A little orientation goes a long way.
Read. Experiment. Put it to work.

Every chapter includes a worked example, an interactive exercise, and a checklist. Start at the beginning or jump to the problem in front of you.

Find your next chapter

Begin anywhere. Make it your own.

10 guides found

A compass for the agent wildernessNyour next movea good place to start01
Start here12 min read

Choose your first AI agent

Start with a job you can judge. Find the right kind of agent, run a small trial, and choose with evidence you can actually use.

Find the agent for your next task
The anatomy of a useful briefa shared definition of donegoalinputsmake it clear02
Start here13 min read

Write a brief an agent can finish

Give an agent a destination, useful context, and a visible finish line. Build a brief that survives the first unexpected detail.

Build your next task brief
The carefully opened garden gatejust enough accessyou hold the key03
Start here14 min read

Give an agent the right permissions

Translate a task into specific access. Let useful work proceed while keeping consequential actions visible and reviewable.

Design a permission boundary
The small, meticulous bug workshopone small changelook closer04
Do the work15 min read

Ship a small change with a coding agent

Take one real issue from reproduction to a reviewed change. Preserve the project, test the behavior, and distinguish a passing build from a shipped result.

Would you ship this change?
A constellation of connected evidencetrace it back05
Do the work14 min read

Research with an agent and keep the receipts

Turn a broad question into a source-backed decision. Trace claims, resolve disagreements, and make uncertainty useful.

Practice source triage
Small steps through the browserlookactcheck06
Do the work14 min read

Put a browser agent to work, one checkpoint at a time

Turn a fragile sequence of clicks into a task you can supervise, recover, and verify. Start with a comparison job, then learn where to place checkpoints before actions change the outside world.

Where would you put the checkpoint?
An orchestra of useful connectionsa useful connectionthe right tool, on cue07
Do the work14 min read

Connect the right tools without creating a mystery machine

Give an agent a small, inspectable path from source to result. Understand what a connection exposes, test each handoff, and make failures recoverable before you depend on the full workflow.

Inspect the tool pipeline
The memory greenhousekeep what growsprune the rest08
Build a habit13 min read

Give your agent a useful memory, not a junk drawer

Keep the context that improves the next task. Separate durable preferences from current state, save decisions with evidence, and make old notes easy to correct or retire.

Sort the memory drawer
The evidence weighing machinesame taskreal evidencea fair comparison09
Build a habit15 min read

Find out whether an agent actually helps

Build a small, honest evaluation from your own work. Score observable outcomes, include your review time, and diagnose failures before changing the prompt or buying a different tool.

Calculate the value of a completed task
The little railway that gets things donereview, then repeata clear handoffarrive somewhere10
Build a habit15 min read

Turn one good result into a workflow you can trust

Capture the parts that worked, define the handoffs, and test recovery before adding a schedule. Build a small operating system for one recurring job, with an owner and a clear finish.

Build your repeatable workflow