How Large Language Models actually work
Tokens, context windows, temperature, and why LLMs predict text — not retrieve facts
Learning paths
Organized tracks for
real AI fluency
Four structured paths from AI basics to professional workflows. Complete lessons in any order, at your own pace.
New to AI? Start here
Your first 5 days with Claude
30–45 minutes a day. One focus per day. By Friday you'll be more effective with AI than most people on your team.
1
Create a free Claude account and set Custom Instructions
Go to claude.ai, sign up, then go to Settings → Custom Instructions. Paste your job title, company, and what you use AI for. Claude applies this to every conversation.
2
Send your first structured prompt
Use this format: You are [role]. I need [task]. Context: [details]. Respond in [format]. — adding a role and format alone dramatically improves output quality.
3
Paste your real work — not test prompts
Give Claude an actual email to rewrite, a document to summarize, or a decision to think through. Generic tests produce generic impressions. Real work shows you what AI actually does for you.
4
Create a Project for your main work context
In Claude, click Projects → New Project. Name it for your role. Paste your team background, recurring tasks, and tone preferences. Every conversation in this project starts already informed.
5
Ask Claude to critique its own answer
After any response, send: "What's the weakest part of that answer and how would you improve it?" This single habit eliminates most disappointing AI outputs and trains you to see what better prompts look like.
Pro tip
Don't ask Claude to "help" with things. Tell it exactly what to produce. "Help me write an email" is weak. "Write a 3-paragraph follow-up to a prospect who went silent. Tone: warm but direct. End with one specific question." is strong.
Track 1 · 9 lessons
AI Foundations
Everything you need to understand how modern AI works — before you build with it. Knowing why AI behaves the way it does makes you dramatically better at using it.
Prompt engineering fundamentals
Role-setting, constraints, examples, output formatting, and the 5-part prompt structure
Understanding AI agents
Tools, memory, planning loops, and when agents go wrong
RAG — Retrieval-Augmented Generation explained
How AI uses your documents without retraining on them
What is MCP and why it matters
Model Context Protocol: connecting AI to real tools and live data
Fine-tuning vs prompting vs RAG
When to use each approach, what it costs, and what problems each one solves
Multimodal AI: images, audio, and video
What today's models can see, hear, generate, and where they still fall short
AI safety, hallucinations, and how to verify outputs
Why AI makes things up, how to catch it, and the 4-question pressure test
Choosing the right model for the job
GPT-4o vs Claude vs Gemini vs open-source: a practical comparison for real tasks
Track 2 · 12 lessons
ChatGPT Mastery
Most ChatGPT users only use basic chat. This track covers memory, Projects, Custom GPTs, Deep Research, Code Interpreter, voice mode, and automation — the features that actually save hours.
Setting up ChatGPT Memory and Custom Instructions
What to store, what to skip, and how persistent memory changes the way you work
ChatGPT Projects — organize your AI workspace
Create project contexts, upload documents, and control scope per project
Deep Research: the complete playbook
Briefing strategies, source evaluation, output cleanup, and when to trust the results
Canvas for editing and long-form writing
Collaborative editing, tracked suggestions, and version control without leaving chat
Custom GPTs: build your own AI assistant
Instructions, knowledge files, actions, and publishing for team use
Data analysis with Code Interpreter
Upload CSVs, run Python, create charts, and export results without writing code
Image generation with DALL-E 3
Prompt patterns for consistency, style control, and editing workflows
Voice mode for hands-free AI work
Real-time voice, best use cases, and when to switch back to text
Automating with ChatGPT and Zapier
No-code workflows that run while you're doing something else
Research to presentation pipeline
Deep Research → Canvas → DALL-E → Slides in one connected session
ChatGPT for email and communication
Drafting, rewriting tone, summarizing threads, and building reusable templates
Advanced prompt patterns for professionals
Chain-of-thought, self-critique, LLM Council, and structured output techniques
Track 3 · 8 lessons
AI Coding
Ship faster with AI-assisted development — the tools, habits, and workflow patterns that actually work in production.
Cursor vs Windsurf vs GitHub Copilot: pick your tool
A practical comparison for professional developers who need to decide this week
Getting the most out of Cursor
Composer, rules files, codebase indexing, and writing a .cursorrules file that works
Claude Code for terminal-first development
Agentic coding, file editing, CLAUDE.md setup, and long-context code review
AI-assisted test writing
Generate unit tests, edge cases, and testing strategies with AI doing the heavy lifting
Code review with AI
Security, performance, readability — what to ask and when to trust the result
Prototyping with Replit and AI
Zero-to-deployed in an afternoon, without a local environment or build tooling
Debugging with AI assistance
Error interpretation, fix proposals, and avoiding hallucinated solutions in code
Shipping habits for AI-assisted teams
PR reviews, commit messages, documentation, and quality gates with AI in the loop
Track 4 · 10 lessons
Workplace Copilots
The AI tools already built into your work apps. Most companies are paying for Copilot or Gemini and most employees have never touched it. This track fixes that.
Microsoft 365 Copilot: the practical setup
Enabling features, permissions, and what it can actually do vs what's marketed
AI in Word: drafting, rewriting, and summarizing
Copilot in Word for reports, memos, and long documents — with honest limitations
Excel AI: formulas, analysis, and Copilot
From formula suggestions to natural-language data analysis without knowing Python
Outlook AI: triage, drafting, and scheduling
Inbox summarization, smart replies, and meeting prep that works in under 5 minutes
PowerPoint Copilot: from outline to deck
Auto-generating slides, designer AI, and speaker notes that don't sound like AI
Teams AI: meeting summaries and follow-ups
Transcripts, action items, and live AI assistance that saves the post-meeting hour
Gemini in Google Workspace
Docs, Sheets, Slides, Gmail, and Meet — what each Gemini feature actually does
Claude for professional writing
Long documents, research memos, editing, tone adjustment, and the 200K context window
Building a personal AI workflow stack
Combine tools across platforms without paying for everything or creating chaos
AI for managers: delegation, reporting, and strategy
Use AI for 1:1s, status reports, decision documents, OKRs, and performance reviews