How To Set Up Your First Agent In Claude
A practical first step into agent workflows: create one focused Claude Project, give it clear instructions, and make it useful before adding more agents.
I write and teach how to build AI systems that expand human intelligence, creativity, and agency: memory, skills, workflows, and creative strategy.
A practical first step into agent workflows: create one focused Claude Project, give it clear instructions, and make it useful before adding more agents.
AI becomes more useful when it stops being one vague chatbot and starts becoming a set of clear roles, tools, and boundaries around your judgment.
Scattered information only becomes useful when it can be inspected, routed, and turned into action. This week's recap connects newsletter archives, paywalled links, and the inbox layer of an AI second brain.
Most second brains become searchable clutter. A useful AI second brain needs an inbox that turns raw inputs into routed action.
An AI Second Brain has to notice when something changes, connect that change to the right part of your work, and turn it into the next reasonable action.
A paywalled link can still become a research agenda. The useful trick is to use AI to work around the paywall legally, by reconstructing the story from open sources.