Rise of the AI Agent

Holy moly, the past month (in my opinion) has possibly been the most rapid advancement period in AI history.

The biggest thing that stands out to us right now - AI can now reliably and capably act on your behalf. While most people are still using ChatGPT and Claude to search the internet, summarise topics, and get answers, others are using AI to run their lives.

The tools are finally capable of doing this without going off the rails, constantly making mistakes, or requiring supervision. AI Agents have been around since 2025, but they are finally useful in the real world.

Here’s a summary of AI’s progress so far:

The fundamental improvement here is the AI models. Claude Opus 4.6 and GPT 5.4 make the models that came before them look infantile. Tools like Claude Code and OpenClaw provide a harness through which AI agents powered by these models can be built and managed.

Big AI War

In broader news, the two big dogs of AI, Anthropic (maker of Claude) and OpenAI (maker of ChatGPT), have been at war.

Anthropic rejected a deal with the Pentagon & US Government over safety concerns in using AI for military operations. A few hours later, OpenAI swooped in and signed the deal, with Trump tweeting:

This caused massive backlash amongst ChatGPT users, with many now switching to Claude, whose revenue jumped to US $8B in one month!

A full write-up here.

Obviously, as well as building AI agents for other businesses, we’ve got a whole team of our own to run daily operations.

Glassy, my modified OpenClaw, manages this team of agents, which performs tasks like

analysing competitors, managing communications, and finding us new clients to work with.

The key is to connect a ‘brain’ of knowledge about your business/life so that it never forgets who you are or what you are working on, and learns over time:

After constantly struggling to get enough info on our clients to start building, we switched to doing ‘AI Audits’. This involves spending days in-person with clients, interviewing their team to discover where AI can help.

Check out this link for an explanation of how we do it.

That’s the newsletter! Thanks for reading :)