Four parts.One playbook for leverage.
Thirteen weeks. Live sessions with replays. Downloadable workbooks, template libraries, and a capstone automation you'll run for years. Every cohort follows the same proven arc — understand the machine, master the craft, pick the right tools, and build automations that run without you.
Part 01The New OperatingSystem
Before you learn a single prompt trick, you need the right mental model. We start with how today's large language models actually reason, where they reliably break, and how to build trust between you and the machine. By the end of Part 01, you'll stop treating AI like a search engine and start treating it like a skilled junior who needs direction.
Week by week
How models actually think
- Tokens, context windows, and why length matters
- Training vs. reasoning — what the model can and can't know
- Hallucinations: the three failure modes and how to spot each
- Working memory: why the right context is 80% of the result
Prompt anatomy
- The four slots every prompt needs (goal, context, constraints, format)
- Showing vs. telling — when to use examples
- System prompts vs. user prompts — and when it matters
- Temperature, reasoning modes, and when to burn more compute
Context design
- Retrieval basics — feeding the model the right source material
- Building your personal 'context library' (Notes, Projects, Spaces)
- When to use a fresh chat vs. continue an existing thread
- Evaluating output quality — a 5-point rubric you can run in 30 seconds
The prompt
“Summarize this 40-page PDF contract for a non-lawyer in a 90-second read, flagging only the terms that shift risk.”
What you get back
Claude returns a 5-bullet risk summary with exact page references, separating 'standard' from 'unusual' language, and surfacing the 3 clauses most likely to matter in a dispute.
What you'll be able to do after Part 01
- Build a reliable mental model for where LLMs excel and where they break
- Write prompts that get usable output on the first try (not the fifth)
- Design reusable context templates for your most common tasks
- Evaluate AI output quickly and know when to push back
Part 02The CRAFTframework
CRAFT is the proprietary 5-letter framework that turns intuition-driven prompting into a repeatable system: Context · Role · Action · Format · Tone. It's what I use every day across Claude, ChatGPT, and Gemini — and by the end of Part 02, it'll be muscle memory. You'll leave with a CRAFT template library for the 12 tasks knowledge workers do most.
Week by week
C, R — Context and Role
- The Context ladder: from ambient context to provided context
- Role-setting that actually changes the output quality
- Persona pitfalls — when role prompts hurt more than help
- Live rewrites: weak prompts → CRAFT-formatted prompts
A, F — Action and Format
- Verb precision: 'Analyze' vs. 'Evaluate' vs. 'Compare'
- Chain-of-thought prompting when the task needs it
- Output formats: JSON, markdown tables, structured prose
- Format templates you can copy for memos, emails, and reports
T — Tone & the full framework
- Tone calibration for internal vs. external audiences
- Brand voice prompting (loading your style into the model)
- Putting it all together: 3 full CRAFT prompts built live
- Your CRAFT template library — 12 reusable prompts, yours forever
The prompt
“Context: I'm the fund manager of a $124M real estate portfolio. Role: act as my COO. Action: review the attached Q1 P&L and flag the three line items I should pressure-test in my weekly ops call. Format: a 3-row table with column headers 'Line item', 'Why it's a flag', 'Question to ask'. Tone: direct, no hedging.”
What you get back
Claude delivers a clean 3-row table with specific dollar variances, plausible root causes, and sharp follow-up questions — ready to paste into the ops agenda.
What you'll be able to do after Part 02
- Apply CRAFT to any AI task, in any model, without thinking about it
- Build a personal library of 12+ reusable CRAFT templates
- Cut your 'prompt-iterate-reprompt' cycle time by 60–80%
- Teach CRAFT to your team so outputs stay consistent across people
Part 03Tools & Platforms thatpay off
Everyone's subscribed to everything. Almost no one has a real system. Part 03 is the opposite of a tool-of-the-week newsletter — it's a ruthless operator's guide to which tools are actually earning their keep in 2026, and how to combine them into a stack that fits your work. Deep-dives into Claude, ChatGPT, Cursor, Perplexity, Notion AI, and n8n.
Week by week
Reasoning: Claude & ChatGPT
- When to reach for Claude (Sonnet 4.6, Opus 4.6) vs. GPT-5
- Projects and Custom GPTs — knowledge bases that actually work
- Computer use, Skills, and the new agentic primitives
- Cost math: why a $200/mo plan pays back in two hours of work
Retrieval: Perplexity & NotebookLM
- Perplexity Spaces — a personal research desk in 20 minutes
- NotebookLM for synthesis of your own document corpus
- Citation hygiene and why AI research is only as good as its sources
- Live build: a market-intelligence workspace for your industry
Code: Cursor & the Claude-first workflow
- Cursor for non-engineers — yes, really
- Claude Code for drafting internal tools without a dev team
- When to use Replit vs. v0 vs. Lovable vs. Base44
- Shipping a working internal app in a single Saturday
Your personal stack
- The 5-tool rule: never pay for more than you'll actually open
- Stack audits for 3 archetypes (operator, analyst, creator)
- Integrating tools: when to use MCP, when to use Zapier
- Your stack plan — written, scored, and ready to subscribe
The prompt
“I'm a real estate operator paying $612/mo across 9 AI tools. Score my stack on signal-per-dollar and recommend what to keep, kill, or swap.”
What you get back
You walk out with a ranked list of your 9 tools, a 5-tool shortlist that covers 95% of your workload, and an estimated $340/mo saved — with zero loss of capability.
What you'll be able to do after Part 03
- Build a lean, 5-tool AI stack that fits your exact workflow
- Know when to use Claude vs. ChatGPT vs. Gemini for a given task
- Ship a working internal tool without hiring a developer
- Stop paying for AI tools you don't open and don't need
Part 04Automations thatcompound
The final part is where everything compounds. You take the prompts and stacks you've built and turn them into durable, multi-step automations that run without you — reviewing inbound deals, drafting weekly reports, triaging your inbox, monitoring markets. We build together, live, using the exact tools top operators use to 10× their output.
Week by week
From prompt to workflow
- What deserves to become an automation (and what doesn't)
- The 5-step anatomy of every durable workflow
- Zapier vs. Make vs. n8n — a clear-eyed comparison
- Your first automation: inbox triage in under an hour
Agent-grade workflows
- Long-running agents: when the model drives the loop
- Human-in-the-loop design — where to put the approval gate
- Scheduled tasks vs. reactive tasks — picking the right trigger
- Live build: an automated weekly report that writes itself
Ship & iterate
- Cost, latency, and reliability — the operator's triangle
- Instrumentation: logging, retries, and graceful failure
- Scaling what works: from one workflow to a system
- Capstone: present one automation you'll run for the next year
The prompt
“Every Monday at 6am, pull last week's Gmail and calendar activity, cross-reference against my CRM, and deliver a one-page weekly brief to my inbox with 'wins', 'in-flight', and 'at-risk' columns — plus the 3 follow-ups I most likely forgot.”
What you get back
A scheduled n8n + Claude workflow that consistently recovers 2–4 dropped threads per week and sets the week's priorities before your first coffee.
What you'll be able to do after Part 04
- Ship 3+ production automations that run on their own schedule
- Pick the right platform (Zapier, Make, n8n) for each job
- Build agents with the right human-in-the-loop gates
- Walk out with a capstone automation you'll run for the next year
Take the full syllabushome with you.
Every lesson title, every sample prompt, every outcome — in a single 12-page PDF. Drop your email and the download starts instantly. We'll ping you when the next cohort opens, and never before.
Seats closewhen the cohort starts.
No grandfathered pricing. No replays if you miss the live sessions. Lock your seat while it's still open.
Starter $29/mo · Pro $99/mo · Annual plans save 17%