There is something particularly frustrating about discovering an opportunity just after it has disappeared.
A place becomes available. Something you’ve wanted for months suddenly becomes possible. A small change opens a door you didn’t even know existed. And by the time you hear about it, someone else has already taken advantage of it.
The information was there all along. You just weren’t looking at the right moment.
I’ve been thinking about this a lot lately. We have incredible AI models that can reason, write software, and solve difficult problems. Yet we still spend so much of our own time keeping track of things, remembering what to check, and figuring out what needs our attention.
I think the next big improvement in how we use AI will come from changing that habit.
AI Is Getting Better at Working Without You
Something important has happened over the past few weeks.
OpenAI introduced Dots, powered by GPT-6 Astra, on September 29. On October 6, Claude moved new Cowork tasks onto cloud infrastructure for Pro and Max users. And on October 8, Anthropic published a working example of an AI agent that remembers previous work, watches for changes, and acts on a schedule.
These developments are connected.
You can now give AI a responsibility, rather than repeatedly giving it instructions. It can return to that responsibility, use connected tools, decide what matters, and continue working when you’re somewhere else.
Think of an agent that watches for an opportunity you’ve been waiting for. When something changes, it checks the details, compares your requirements, prepares what you’ll need, and tells you whether it’s worth acting on.
Or an agent that watches the software you depend on. When a new release changes how something works, it checks the documentation and prepares updated instructions for you.
These aren’t just reminders. They involve reading, checking, making decisions, and preparing useful work.
And building them has become surprisingly simple.
I’ll show you below how to set up your own, first with Claude and ChatGPT, then how to make them smarter with memory, skills, tools, and the newer agent features.
Read more
