Working with agents
- Ask for proof of work instead of a summary. Ask a worker for evidence you can inspect.
- Find implicit state machines with Pstack. Model each subsystem and chart its key flows.
- Review untested branches. Decide which checks need tests and which states the types can rule out.
Reconsider the task
- Could a human working for free do it? Test what it would take to hand off the work.
- Restate the goal. Check that the agent understood your goal before it starts.
- Find a better method. Reconsider the method you assumed the work required.
- Design a system to do the work. Reframe a task as a system to build.
- What is worth building now? Look for work that became practical as the cost changed.
- Deeper without bigger. Ask what better practice would add to the same task.
- The best next question. Get the question that would move the work forward.
- The same with half the code. Ask for a smaller implementation of the same thing.
- Stay at the frontier. Resolve the decisions you can see now.
- Brittle and durable fixes. Compare a fragile fix with one that will hold up.
- Audit the bottlenecks. Use the last day of work to identify what slowed you down.
- Abundant bounded labour. Look for work you can split into simple tasks.
Reconsider what feels necessary
- The trusted clone. Imagine handing off today’s work to see what you would choose next.
- Ten times the agency. Surface requests you are holding back.
- Aversion as a pointer. Identify what you would have to let go of to solve the problem differently.
Reconsider how to meet the world
- The smallest real thing. Find the smallest thing another person can actually try.
- Prevent the error type. Change the system so the same class of mistake is hard to make.
- Make stronger testing concrete through history. Use past bugs to find better tests.
- What would have to be true? Name the conditions a proposal depends on.
- What would make this wrong? Name the evidence that would overturn your conclusion.
- Similar cases, different decisions. Find the distinction behind two different decisions.
- Make the miss visible. Ask what evidence would change the decision on its own.