A package to give Chris, Tony, Luke (or anyone curious) so they can try this themselves
Sam doesn't use AI as a chatbot. He uses it as a thinking partner plus delegated workforce. Over 36 hours, he built a sports betting analysis platform that backtests against 13,903 real NBA games, a financial planning hub, a customer quote system, 3D print management, and more.
The same approach scales to anything you're doing. This page shows how to start.
Part 1 — The 30-Second Pitch
If a friend asks "what's the deal with how you use AI?" — read them this:
"I treat AI as a partner I delegate work to, not a search engine I query. Give it the full situation, let it draft, push back when it's wrong. Over a weekend I built a betting analytics platform with real backtests against 13,000 games. The leverage is wild once you stop treating it like a fancy autocomplete."
Part 2 — The 3 Things You Need to Actually Start (10 min setup)
Claude.ai account — free works, paid ($20/mo) much better for serious work
Claude Code CLI (optional but recommended) — gives AI file-system access on your machine. docs.claude.com/claude-code
Persistent file storage (optional) — GitHub + Cloudflare Pages, or local. Lets work compound instead of vanishing per session.
Part 3 — The 7 Patterns Sam Uses
1Rich context dumps. Give AI the full situation, not just the question. Background, constraints, what you've tried, what you actually want.
2Push back when wrong. Don't accept output you doubt. "Actually no, here's what's happening: ..." Corrections compound — AI gets better in-session.
3Decompose hard problems. Sequence complexity. "First help me think through X. Now Y. Now combine."
4Verify AI output. Click through. Test. AI is wrong ~10–20% of the time and confidently so. Trust but verify.
5Strategy plus tactics together. Don't separate "thinking" from "doing." AI is good at both at once. Discuss approach AND have it implement, in same flow.
6Use AI for what AI is good at. Synthesis · drafting · research · scaffolding · code generation · pattern recognition. NOT replacing your judgment, your relationships, or your taste.
7Keep AI doing things you'd actually use. Building for building's sake is a trap. If you wouldn't use the thing, don't build it.
Part 4 — The 3 Starter Projects (try one this week)
Starter A — Daily Brief Generator
⏱️ 15 minutes · ⚡ Immediate value
Each morning, give AI your calendar + priorities. Get a 1-page brief you can act on.
"Each morning I'll give you my calendar + priorities. Generate a 1-page brief I can act on. Here's today's: [data]"
Starter B — Decision Analysis
⏱️ 30 minutes · 🧠 High-value thinking
Pick a decision you're stuck on. Get options + tradeoffs + what you're not considering.
"I'm trying to decide [X]. Here's the context: [details]. Walk me through options, tradeoffs, what I'm not considering."
Starter C — Domain Knowledge Base
⏱️ 1–2 hours · 📚 Compounds over time
Pick a topic you want to actually learn. Get a structured path.
"I want to learn [X] systematically. Build me a structured learning path with 5–10 lessons, each ~10 min."
Sam's personal financial hub, customer jobs, and household docs stay private. These pages are friend-shareable by design.
Part 6 — How to Know If It's Working
After 1 week:
✓ Saving time on at least one task? → Yes = working
✓ Doing things you couldn't before? → Yes = expanding capability
✓ More frustrated or less? → Less = good
If more frustrated: probably using it for wrong things. AI is good at synthesis + scaffolding + research. AI is bad at being your boss, therapist, or replacing human judgment.