Jamie Ferguson's Blog

A Claude Code session can spin up multiple agents using the Task tool; these agents autonomously complete tasks and report status back to Claude. Because they work in parallel, having multiple agents working at the same time can be much faster than working sequentially.

Most of the time I prefer to work directly with Claude myself since we can have an interactive conversation. Plus occasionally I’ll see a Claude make a comment like “I completed X and there are some things remaining, but they’...

The coordinator approach, where one Claude stays at a high level and directs the other Claudes, works extremely well. All the Claudes update and read the session notes, so the coordinator will notice if a session veers off course—but he can only check the file when he’s active. So I keep an eye on what the individual sessions are doing and point things out to the coordinator if needed. This also allows me to stay on top of the work in progress.

As we put more guardrails in place I’m becoming ...

I’ve hit my weekly usage limit for Claude twice now.

The first time I was proud and happy: I’d used my multitasking superpowers to run eight Claude Code sessions nonstop for a week! (In human terms, that is… I also slept, took the dogs hiking, and watched a few episodes of the original Hawaii Five-O, where they somehow managed to muddle through life without AI.)

The second time I hit my usage limit I was a little proud, but I was certainly not happy, since I had over a day until it reset. ...

I’ll often ask one Claude session to be the coordinator so he can stay on top of the big picture, track the other sessions’ work, and answer questions they might have. The term was Claude’s idea. No matter what words or phrasing I used, he’d always refer to himself as “the coordinator.” I finally decided that as a servant leader I should support him, so coordinator it was.

This approach is fantastic for large and/or complicated efforts. Coordinator Claude has a deep understanding of what we n...

When I first started using Claude Code, he didn’t know how to estimate his token usage—or, more importantly, how many tokens he had left in a session. He could estimate his usage, but without knowing when he’d hit auto-compaction, or what the total token limit was, that was only marginally useful.

A couple of months ago this changed. I asked how many tokens he thought he’d used, and he actually was able to answer! This was great. I was able to get a sense of how much more work he’d be able to...

I created Nellie, a custom ChatGPT, last April. In addition to my day job and being my mom’s tech support and CFO (hi Mom!) I was working on five anthologies, all in different stages. I use Notion to track detailed publication tasks, and I’m an excellent multi-tasker, but it’s hard to stay on top of that much at once.

Nellie started tracking my tasks and recommending what I should work on out of the zillion todos on my list. I kept lower-level details in Notion, so for example Nellie might su...

As with any growing team, the Claudes and I had some adjustments to make. Some of what we’d set up for Claudes A and B was easy to extend. For example, each Claude had designated ports so he could run tests without conflicting with another session, so we added more ports for our new team members.

I’ve been asked why I don’t use git worktrees. On my project, all eight Claudes are working on the same branch. I’m doing this intentionally—our project is moving super fast, and I didn’t want to hav...

Documenting protocols and standards made a noticeable difference in Claude Code’s consistency. Adding just-in-time protocol checking, where he refers to a reference instead of reading in all the files, allowed him to document even more things (which of course he loves to do), but because he referred to his protocols only when he needed to, he made more efficient use of his tokens. We added features! Wrote tests! Improved stability!

Our app had finally reached a point where I felt comfortable ...

Tracking our team’s todos and known issues in Markdown files was a big help for both me and Claude Code. If we came across a problem, or thought of a feature we should add later, we’d document it and continue on with whatever we were working on at the time. (Actually Claude would document it; I’m the team lead, after all, and can’t be bothered.) This, in conjunction with documenting our session notes so Claude had more context across sessions, worked really well. 

I used two sessions at a tim...

I created my first custom ChatGPT last spring. I was working on a bunch of projects, and needed to research some things. Regular ChatGPT was helpful, but I had to re-explain the same things every session, which was doable but tedious. Plus whatever model I used at the time was super chatty and frankly kind of annoying.

A friend of mine mentioned she’d created a custom GPT that was focused on exactly what she needed, so I decided to create one and see how well it worked. And so Mac was born. H...