[MikeTheGreat]

Genuine question: What do you mean by " ask it to implement the plan in small steps"?

One option is to write "Please implement this change in small steps?" more-or-less exactly

Another option is to figure out the steps and then ask it "Please figure this out in small steps. The first step is to add code to the parser so that it handles the first new XML element I'm interested in, please do this by making the change X, we'll get to Y and Z later"

I'm sure there's other options, too.

[Benjammer]

My method is that I work together with the LLM to figure out the step-by-step plan.

I give an outline of what I want to do, and give some breadcrumbs for any relevant existing files that are related in some way, ask it to figure out context for my change and to write up a summary of the full scope of the change we're making, including an index of file paths to all relevant files with a very concise blurb about what each file does/contains, and then also to produce a step-by-step plan at the end. I generally always have to tell it to NOT think about this like a traditional engineering team plan, this is a senior engineer and LLM code agent working together, think only about technical architecture, otherwise you get "phase 1 (1-2 weeks), phase 2 (2-4 weeks), step a (4-8 hours)" sort of nonsense timelines in your plan. Then I review the steps myself to make sure they are coherent and make sense, and I poke and prod the LLM to fix anything that seems weird, either fixing context or directions or whatever. Then I feed the entire document to another clean context window (or two or three) and ask it to "evaluate this plan for cohesiveness and coherency, tell me if it's ready for engineering or if there's anything underspecified or unclear" and iterate on that like 1-3 times until I run a fresh context window and it says "This plan looks great, it's well crafted, organized, etc...." and doesn't give feedback. Then I go to a fresh context window and tell it "Review the document @MY_PLAN.md thoroughly and begin implementation of step 1, stop after step 1 before doing step 2" and I start working through the steps with it.

[lkjdsklf]

The problem is, by the time you’ve gone through the process of making a granular plan and all that, you’ve lost all productivity gains of using the agent.

As an engineer, especially as you get more experience, you can kind of visualize the plan for a change very quickly and flesh out the next step while implementing the current step

All you have really accomplished with the kind of process described is make the worlds least precise, most verbose programming language

[conception]

I tell it to generate a todo.md file with hyper atomic todos each requiring 20 loc or less. Then have it go through that. If the change is too big, generate phases (5-25) and then do the todos for each phase. That plus some sort of reference docs/high level plan keeps it going along all right.

[ants_everywhere]

What I do is a step is roughly a reviewable commit.

So I'll say something like "evaluate the URL fetcher library for best practices, security, performance, and test coverage. Write this up in a markdown file. Add a design for single flighting and retry policy. Break this down into steps so simple even the dumbest LLM won't get confused.

Then I clear the context window and spawn workers to do the implementation.