Skip to content
Guides

Interactive deep-dives.

Hands-on guides that go beyond surface-level tips. Each one is built to change how you work.

Flagship guide

AI Orchestration: How We Run AI As An Operating System

The full playbook, nothing held back. How we run AI as an operating system, not a chatbot: models, frameworks, skills, memory, connectors, automations, and security. The method behind the AI Orchestrator OS, laid out end to end.

14 sections12 min read
  • What Is An Agentic Operating System?

    An agentic operating system is one shared AI stack under your business, not one more chatbot on the side. What it is in plain words, how it differs from a chatbot, a pile of AI tools and a single AI agent, the six layers, what running a business on one looks like day to day, six first steps for this week, eight mistakes that stall it, and when to build it yourself or get it built. One copy-paste starter prompt included.

    10 sections14 min read
  • A Free Claude Course for Business Owners

    A free written Claude course in eight short lessons. Set Claude up, teach it your business with instructions, projects and memory, brief it properly, give it your files, build a skill, connect your email and calendar, and check the work with 10/80/10. Each lesson has steps and one prompt to copy. No videos, no certificate.

    8 lessons14 min read
  • Claude for Business: A Working Guide

    How to put Claude to work in a real business, from an owner who runs on it. What it is good for and what it is not, the three things that make it useful (what it knows about you, how you do things, the tools it can reach), five jobs to hand over first with a copyable prompt for each, the 10/80/10 way to check the work, how to keep client data safe, and how to bring a team along.

    10 sections17 min read
  • Claude Mods: Do This First

    Most people install someone else's Claude mods, or build a game to play while Claude thinks. Do this first instead. One copy-paste prompt gets Claude to read your own chats and tell you which mods you need. What that audit found across 440 of my messages, seven of the Claude Code plugins I run, how to have Claude build yours with no code, and the traps to check before any mod goes on your machine.

    11 sections10 min read
  • Audit Your Claude Skills With One Prompt

    Skills pile up, get long and quietly stop working well. The five rules from Anthropic's best practices turned into tests you can run, why the check splits into a pass a script does and a pass that needs each skill read, eight traps we hit auditing 1,311 real skills, and how to fix a long skill without losing a rule. One copy-paste prompt at the end runs the whole audit in Claude or Codex.

    7 sections10 min read
  • Build An AI Voice Assistant That Answers To You

    Not the one that answers your customers' phone. The one you ring on the way to a meeting, that knows your queue and your numbers and still cannot approve a thing on its own. The full voice build: a worker that owns the audio and holds no keys, room tokens scoped to a microphone, turn detection and the pre-roll that stops quick commands arriving as mush, merging one thought into one turn, a measured latency budget with an honest filler line, spoken approval the model is given no tool to invoke, the four failures that make a live call go quiet, telephony against in-browser, and what a minute really costs. Four copy-paste prompts included.

    13 sections18 min read
  • One Operator. One Dashboard. A Fleet Of AI Agents Behind It.

    Once you run more than one agent, the bottleneck stops being what they can do and starts being where you say yes. The full build of an AI dashboard for agents, a local-first human-in-the-loop command centre: the boring stack and its deliberate absences, the one SQLite table every agent writes into, the approval board, an installable phone layer with a swipe deck, a voice call where the model gets no tool that can approve anything, a private mesh instead of a login page, and the order to build it in. Eight copy-paste prompts included.

    12 sections22 min read
  • Your AI Operating System, On A Machine That Never Sleeps

    Your AI team stops working the moment you shut the lid. The fix is self-hosted AI on a Mac Mini server in the corner of a room, running your scheduled routines and a standing agent whether you are there or not. The full build: the four settings that decide whether it survives a power cut, a private network with nothing exposed, a pinned toolchain, self-hosted AI agents with a local LLM fallback, backups, and the five failures that take it down silently. Three copy-paste prompts included.

    13 sections24 min read
  • Claude Plan Mode: The First Hour That Pays For The Whole Day

    Burning your AI limit before lunch and still getting work back you have to redo are the same problem, caught too late. Twelve sections on running plan mode properly: the dictated brief, the one line that makes it interrogate you, how to mark a plan up before you approve it, the four jobs that always need one, and the one-hour morning routine that moves four projects a day instead of one.

    12 sections12 min read
  • The Meeting Wiki Brain: Turn Calls Into Content

    For coaches, consultants and course creators. Your best teaching disappears after every call. Build a meeting wiki that captures it, turns it into content ideas, and trains an AI content team to sound like you. Wikis, routines, skills, and a build-it mega prompt.

    10 sections10 min read
  • Build Your Own AI Stack: Track What You Run

    Once you run AI across more than one project, you lose track of what you have. Build a small, self-maintaining system that records every skill, MCP, plugin, and agent, maps how it connects and why, and shows it in a dashboard. A manifest engine, a Karpathy-style wiki, a Next.js dashboard, and a mega prompt that builds it for you.

    11 sections12 min read
  • Agent Skills: One File That Runs a Whole Method

    A skill is a folder with one file inside it. Say the trigger, it fires, and the model runs a tested process instead of winging it. What skills are, the tools that read them, how to install them in the desktop app and Claude Code, the right structure, how to vet and find them, and how to track your whole stack.

    9 sections10 min read
  • AI Agent Orchestration: How To Make Agents Work As A Team

    One agent does a task. A team of agents, run right, does the work. The patterns that hold up, the hand-off rule that keeps it honest, where multi-agent setups break, and how to orchestrate your own agents from scratch.

    7 sections5 min read
  • Agentic AI For Business: What It Is And Where It Pays

    Agentic AI explained without the hype. What AI agents actually are, the use cases that move the needle for a $1M+ business, where the ROI really comes from, and how to start without betting the business.

    7 sections5 min read
  • Are Autonomous Agents Right for My Business? (2026 Guide)

    OpenClaw, Hermes, or build your own? An honest, no-hype guide to one decision: do you actually need an autonomous AI agent, or does what you already run do the job? Eight sections, a three-question test, and the security reality for client data.

    8 sections8 min read
  • Claude Fable: Before It Got Banned

    Everything we learned about Anthropic's most capable public model in the 72 hours it was live, before a US export-control directive took it offline. What it was genuinely good at, how we ran it at orchestration level, and the business lesson behind the shutdown.

    7 sections9 min read
  • How to Turn Your Quote History into an AI Estimator

    Turn years of old quotes into an AI tool that prices a new job in minutes. A practical, no-code build for any service business in Claude Cowork or Code, with six steps and eight copy-and-paste prompts. Worked through with a software agency and an electrical business.

    6 steps13 min read
  • Claude Mythos: The Honest Intelligence Briefing

    Everything we know about Anthropic's next frontier model. Twelve sections covering the two leaks, the verified capability, the inflated claims, the agent infrastructure exposed in the Claude Code source leak, Project Glasswing, the safety profile, and four realistic release scenarios. Built from system-card data, partner reports, and benchmark debate.

    12 sections14 min read
  • Claude Output Styles: Make Claude Answer Like You Think

    You have already told Claude to be concise and it drifted back within three exchanges. One markdown file fixes the shape of every answer, on every session, on every local surface. What an output style is, which layer the rule belongs in, six mistakes that make a style useless, and a ten-minute build with the measurement step everyone skips.

    9 sections9 min read
  • Claude Code Routines: The Field Guide

    What people are shipping in the wild with Routines. Eight sections covering the four-part architecture, three trigger types, daily run caps, ten power-user hacks, six anonymised business use cases, and a 30-minute first-routine walkthrough.

    8 sections11 min read
  • GPT Micro-Magnets: The 2026 Lead Magnet Playbook

    Ten sections on building a custom GPT that actually grows your list. Pick the job, design the four-step in-GPT flow, write the instructions, solve the email capture problem, distribute via ManyChat or GHL, and track the five numbers that matter.

    10 sections12 min read
  • Claude Code vs Claude Cowork: The April 2026 Playbook

    Two agents, same Claude, very different jobs. An interactive guide to choosing between Claude Code and Claude Cowork, with a feature matrix, a five-question decision framework, eight scenarios, and the operator playbook for running both together.

    12 sections15 min read
  • The Opus 4.7 Leak: Designing Prompts That Actually Work

    Claude Opus 4.7's system prompt leaked in April 2026. Ten insights and three deeper patterns from how Anthropic writes their own prompts, turned into a practical reference for your agents.

    10 insights16 min read
  • Claude's 2026 Stack: Five Levels of Unlock

    An interactive walkthrough of Anthropic's 2026 ecosystem, framed as five levels of unlock. Cowork, Scheduled Tasks, Dispatch, Routines, and Managed Agents, with a decision framework, persona playbook, and a full launch timeline at the end.

    5 levels14 min read
  • Claude Chat vs Co-work vs Code: Which Mode Does What

    Three modes of Claude, three different jobs. Interactive breakdown with a live scenario switcher, side-by-side comparison, and a decision helper so you always pick the right mode for the work.

    9 insights8 min read
  • Claude Code: The Power User Guide

    Nine insights from the Claude Code architecture that separate the top 1% of users from everyone else. Interactive guide covering configuration, sub-agents, hooks, permissions, and more.

    9 insights10 min read