💌 Part one of the course is live / Don't forget to update Claude Code

[IN SHORT] This week

• [WHAT I BUILT] Part one of the course is on YouTube, and the full material is free to a limited number of people who reply.
• [CASE STUDY] The founder of MORTAL turned years of research on grief into a content engine.
• [WORTH YOUR TIME]
  • Update Claude Code today, because Opus 5.5 is now the default.
  • Ignore the Jev hype for now, unless you're building production systems.

[WHAT I BUILT] The agent ecosystem course

Most people use AI by sending a message, it does something and sends something back, so you're still the agent.

I think it should be the other way round, an ecosystem that knows how your business runs and what's going on, so it takes work off you without being asked.

So I built a course, ten stages that take you from the basics of Claude Code to an ecosystem that will support your business to scale.

(...and all the content is constantly being refined and improved through the 1-2-1 coaching sessions I do with business owners).

The first video in the series is live now on YouTube!

It covers:

• Why you'd build an ecosystem, and what mine already does.
• VS Code with the Claude Code extension, and why it beats a chat in the Claude app.
• A dictation tool, because talking gets far more out of your head than typing.
• CRIT, the prompting framework I use to get what's in my head into the ecosystem.

Watch it →

While I'm recording the rest of the videos, I'm making the material available free to a limited number of individuals, so if you want full access, reply to this email and I'll get you set up!

[CASE STUDY] The content flywheel

This week's one is from work with James Sowden, the founder of MORTAL.

The mission, in his words:

"MORTAL is building a new way to help people grieve better: bringing the world’s wisdom about loss into everyday life.

We’re currently in research and prototyping, and looking for grief and end-of-life experts, researchers, technologists, advisers, collaborators and people interested in testing what we’re building."

What we covered on our first call wasn't an introduction to the tools.

He was already quite well versed in using them, instead it was how we could set up a content engine.

Something that would allow him to quickly produce a high volume of authentic content to make people aware of MORTAL.

We decided this for two reasons:

1. They're going to need beta users and quick feedback cycles to figure out how to improve the product.
2. They want to work on the demand indicators investors like to see, like long waiting lists and cult followings.

What he already had was the material, a founder story, the brand documents, and years of reading and research on the subject the app is built around.

So the first job was getting all of that into the folders Claude could read before coming back with suggestions.

This wasn't easy, and in his words:

"Getting the right information into the folders meant I rewrote the foundational documents. There were gaps and there were conflicts."

Once this was done, next was the skills, each one a routine written once and run as often as he likes:

• One reads the research, finds an angle, and interviews him until it has his point of view rather than a generic one, then writes the beats and, if he asks, the script.
• One takes a single piece of inspiration, a poem, a scene from a series, and turns it into script beats the same way.
• Next in the queue is one that takes a finished video's transcript and turns it into a carousel.

And any feedback he gives automatically updates the skills, so the next time it'll be produced quicker, starting the flywheel.

And in his words:

"The last week has been tough in lots of other parts of my life, but actually, this has been like a bit of a highlight."

If you want to learn more about MORTAL, check the link for the waitlist below.

Join the waitlist →

His first job, getting everything Claude needs into one place, is where every build starts, so I've written up how to map yours before you set anything up.

Read it →

[WORTH YOUR TIME] Update Claude Code, and ignore the Jev hype

Update Claude Code

Opus 5.5 is out, but what matters more is what's changed in Claude Code, so if you haven't yet, go to Extensions in VS Code and click Update.

Opus 5.5 is now the default model on Pro as well as Max, and Anthropic say:

• it works at Fable 5.1 level on most work
• over 30% faster
• around 40% cheaper than Opus 5.

Once you've updated, three things are worth knowing:

• The agent map: Click the agent count to see every subagent it's spun up, how many tokens each one has used, and what each one was asked and which tools it called.
• The clock: That's your prompt cache, which on a Claude plan keeps the whole conversation in Claude's head for an hour, like your team still being in the meeting room. It counts down the minutes left, and once it turns red, the next message rereads everything and uses more of your limit.
• Remote control: It's been flaky for a long time, and in my experience it just works now, so with the desktop app open you can pick up any chat from your phone, and it pings you when it's finished or needs a decision.

Ignore the Jev hype for now

Jev is the new model all over your feed, and it's built to make quick decisions inside software rather than to chat.

Instead of writing a reply, you give it some text and a set list of answers to choose from, a category, a score or a yes or no, and it hands back the one it thinks is right, with how sure it is.

Because it only ever picks from your list, it can't answer with anything that isn't on it, and it comes back in under half a second for a fraction of the cost of a chat model.

It is an amazing breakthrough, and people will find great applications for it, for now the main one seems to be sorting and routing inside other software, like which pile an email goes in or which step an agent takes next, because it tells you how sure it is every time.

They've paused new signups while they cope with demand, and unless you're building production systems, you're likely not going to use it, but no doubt it'll be coming to a product near you soon.