Build your AI operating system.

How I write agent goals that do not wander

My seven-step workflow for turning a messy idea into a bounded agent goal, a context rider, and a finish line the agent can prove.

How to build your second brain with Fathom

A practical workflow for turning captured meetings into project requirements, reusable skills, and a second brain you can actually operate.

A 90-minute startup validation sprint

The seven-part sprint I use to pressure-test a business idea before spending weeks building it.

Companion video

How to build an AI operator for your business

Four architecture decisions that turn expert content into a tested operating system instead of another generic chatbot.

Companion video

Claude Schedules vs Routines (what's the difference)

Routines, Desktop scheduled tasks, and /loop each run Claude on a timer — but they differ on where they run, what they can touch, and how fast. Here's the decision matrix.

Claude Code Ultraplan, clearly explained

When a browser-reviewed cloud plan is worth the handoff, when local planning wins, and how to keep the plan accountable to the repository.

Hermes Agent, clearly explained

What Hermes Agent is, where its learning loop helps, and the safeguards I would put around messaging, browser, memory, and scheduled work.

Companion video

Claude Managed Agents: deploy your first governed agent

The four Managed Agents concepts, a safe deployment sequence, and the permission boundary to understand before production use.

Companion video

Your second-brain system, done with you

The installation chain, four maintenance operations, and domain-vault pattern behind a second brain that stays trustworthy.

Five workflows for Perplexity Computer and Claude Cowork

A task-by-task comparison of two agent work surfaces, with the permissions, checkpoints, and current-product caveats that matter.

Companion video

The non-developer's guide to Claude Cowork

Claude Cowork puts an AI agent inside your files, browser, and apps without writing code. A practical walkthrough for operators who want results, not a dev environment.

A Claude Channels setup for remote, local work

How I think about a Mac Mini, a laptop, selective sync, and a message channel without turning chat into an unrestricted remote shell.

Companion video

How I built a voice agent for our customer discovery calls

The six-node conversation structure I used for Annie, plus the disclosure, consent, privacy, and human-escalation controls a production voice agent needs.

Ten OpenClaw workflows worth building carefully

Ten useful agent workflows, rewritten around read-only defaults, minimum access, human approval, and measurable outcomes.

Ten practical business experiments inspired by autoresearch

How to adapt the keep-or-discard experiment loop to business problems without pretending every number is safe to optimize automatically.

Twelve OpenClaw power-user patterns that actually help

A dated operator checklist for context, routing, schedules, subagents, logs, voice, and defense in depth.

Companion video

Installed Paperclip—now what?

A governed three-layer workflow for coordinating agents, applying reusable operating skills, and running constrained experiments.

The Claude Dispatch starter pack

How Dispatch, Cowork, plugins, connectors, local tasks, and cloud schedules fit together without conflating their execution boundaries.

Companion video

How to coordinate a company of AI agents with Paperclip

Paperclip's organizational model—roles, goals, tasks, budgets, adapters, approvals, and heartbeats—reframed as a human-governed control plane.