I always enjoy chatting about how folks use AI and what tools they use. Partly because everything is changing all the time and you never know if you are missing out on the best tool ever, because let’s be honest, it’s easy to miss, given the torrent of AI-related information coming out every day.

Opus 5.5

The GOAT, enough said!

Chief of Staff / Firstmate

Most of my work now happens through a firstmate-type session that manages other ‘worker’ sessions. The firstmate’s job is to talk to me and manage the other agents for me. These are not subagents but full Claude Code/Codex sessions. This really allows me to scale and do more. While I do lose some visibility into the actual details in the process, I can always jump into an individual session and chat directly if I want to. Which I sometimes still do, but not that often.

I was introduced to the idea via Kun’s firstmate project, but I haven’t actually been using the project itself. I’ve built my own mini version of it and it’s been working pretty well so far.

Firstmate: talk to one agent, orchestrate many crewmates across your repos The idea in one picture (banner from the firstmate repo).

If you want to see it in action, Kun has a long video building a full app this way. It starts at the Herdr + firstmate part:

This is a viable way of working with agents now that they are really good at communicating and carrying out long-running tasks.

Voice

The voice models have vastly improved compared to 6 months ago or so. They are much more intelligent than they were before and can delegate tasks to subagents. At least for me, I default to just talking to my agents on the phone for most daily tasks.

I also really enjoy voice mode when I’m learning something or wanting to understand something complex. Voice makes it super easy to progressively directionally explore complex topics. When paired with visualisation skills like diagrams in a harness with a rich UI, like the Codex app, this is even better.

I think the next step would be to have live duplex video as well. This might be a human character that you speak to, or it could be live visualisations/video/images that update as you chat about something, in the same way live duplex audio works today. It’s going to unlock so many things. While we’ve had a preview of live video input with Gemini Live, it would be amazing to explore how things would look if agents could do live duplex video.

Dictation

I dictate whenever and wherever I can. At least for me, this is clearly a superior way to communicate with agents. And to be honest, even outside of agents, more generally. Whenever I’m in the office or a public space where I can’t speak out, I feel slow and constrained now.

I’ve been using the fully local and open-source Handy app for a while now. Although I need to go evaluate the landscape again, as there have been a bunch of interesting releases recently.

Orca

Orca has been my “IDE/terminal emulator/harness” (honestly not sure what any of these words mean anymore. It’s all in flux ATM) or “Agent Multiplexer” for about 3 months now. And that’s a very long time in the age of AI. It’s basically a terminal emulator (e.g. iTerm2, Ghostty, etc.) with almost everything you need to do AI-native development (or non-development agent work 😄). I use only about 5% of all the features it has, but these have been sticky enough to keep me coming back. The value add here is that they are all included in one integrated experience. I used to use 10 different apps/services for the same features before.

Orca with Claude Code and Codex running side by side, plus the Orca mobile app Claude Code and Codex side by side in Orca, with the mobile companion app (screenshot from the Orca repo).

  • Vertical & horizontal tabs, because you can never do too much context switching! 🙃
  • Claude/Codex usage monitoring & subscription switching (this is still a bit buggy though)
  • Remote control via mobile and desktop: especially useful if you are switching between subs and harnesses all the time. And if you want to control sessions running on your MBP/Mac mini from your phone while on the train (assuming you have a decent connection 😂)

Looking at Orca as a product, it has really, really good taste. The authors are like my personal curators for the latest AI tools and patterns; they implement them inside Orca and ship them over to me. This is great! And they do an awesome job keeping Orca reasonably performant and stable, especially given how fast they ship. And they ship really fast, at AI-native speed.

Here’s the official launch video from the Orca team:

Herdr

I think of Herdr as Orca on a diet. It’s minimalist and is more of a TUI. I run it on my VPS boxes as it allows me to SSH into them and control them in a nice, persistent way. While I haven’t really played around much with their new app, I really love the full customisability of it (kinda like the new DeepSeek harness). A more minimalist, more performant take on Orca would be super duper interesting.

Herdr’s app showing spaces, agents and a Claude Code session

Granola

Capturing transcripts and notes from all your meetings is super useful for your agents. I love how minimalist and useful Granola is as an app. From a product and design PoV, Granola is what I want my products to look and feel like! How did we ever live without it?! 😄

How do you AI? Let me know in the comments or hit me up for a chat! I promise not to get too excited 😂!