Capstone: Build Your AI Toolkit

🔄 You’ve compared models, pricing, writing, coding, research, and business. You have a three-question decision framework and a routing table. Now it’s time to make it personal.

In this final lesson, you’ll build a toolkit you can actually use tomorrow — a set of go-to prompts optimized for each platform, a personal routing cheat sheet, and a plan for keeping it all current.

Exercise 1: Map Your Tasks

Grab a piece of paper (or a note app) and list your 3-5 most common AI tasks. Be specific — “write stuff” is too vague. “Draft weekly client update emails” is perfect.

Here’s an example:

Now do yours. The three-question framework from Lesson 7 handles the “Tool” column: Does it need images/voice? Big context? Code execution?

Exercise 2: Build Platform-Specific Prompts

Here’s something most people miss: the same task benefits from different prompting strategies on each platform.

Claude responds better to:

• Detailed context upfront (it uses its context window well)
• Explicit role assignments (“You are a senior code reviewer who…”)
• Asking it to think step by step (its reasoning chains are strong)
• Giving it room to push back (“Tell me if my approach has problems”)

ChatGPT responds better to:

• Structured output requests (“Format as a table with columns…”)
• Specific formatting instructions (markdown, bullets, headers)
• Chained tool use (“Search for X, then analyze the results”)
• Clear constraints (“Keep it under 200 words”)

Here’s the same task, optimized for each platform:

Task: Write a project update email

Claude prompt: “I’m a project manager updating my VP on a website redesign. The project is 2 weeks behind because the API vendor delayed their SDK. I need to be honest about the delay but confident about the recovery plan. Write a 150-word email that acknowledges the issue without sounding panicky. Push back if you think my framing is wrong.”

ChatGPT prompt: “Write a 150-word project update email from a project manager to their VP. Context: website redesign is 2 weeks behind due to API vendor SDK delay. Tone: professional, confident. Format: subject line + body. Include: (1) status, (2) root cause, (3) recovery plan with dates.”

Both produce good results. But Claude’s version leverages its strength at nuanced tone. ChatGPT’s version leverages its strength at structured output.

✅ Quick Check: Look at your task list from Exercise 1. Pick one task and write two prompt versions — one optimized for Claude’s strengths, one for ChatGPT’s.

Exercise 3: Your Personal Routing Cheat Sheet

Create a one-page reference you can pin to your monitor or save in your notes. Here’s a template:

My AI Routing Rules:

1. Needs images, voice, or code execution? → ChatGPT
2. Long document (50+ pages)? → Claude
3. Quick question or lookup? → ChatGPT (save Claude’s messages)
4. Writing that needs my voice? → Claude
5. Data analysis with computation? → ChatGPT
6. Complex reasoning or architecture? → Claude
7. Everything else? → [Your default tool]

Customize rules 4-7 based on your actual work. The first three are universal.

Course Review: What You Learned

Let’s recap the major findings across all eight lessons:

Models (Lesson 2): Claude Opus 4.6 leads Chatbot Arena (#1 Elo). GPT-5.2 leads math/science benchmarks. Both released their latest flagships on the same day (Feb 5, 2026).

Pricing (Lesson 3): Both cost $20/month. ChatGPT gives 4-8x more messages. Claude gives longer outputs and a bigger context window. Annual Claude plan saves $36/year.

Writing (Lesson 4): Claude writes more naturally and handles long-form better. ChatGPT adds DALL-E images and voice mode. For pure text quality, Claude edges ahead.

Coding (Lesson 5): Claude leads SWE-bench (80.9%). Developers prefer Claude 78% in blind tests. ChatGPT leads Terminal-Bench for DevOps. Most devs benefit from using both.

Research & Business (Lesson 6): ChatGPT’s Deep Research is the best autonomous research tool. Claude synthesizes provided documents better. 70% of Fortune 100 use Claude.

Decision Framework (Lesson 7): Three questions handle 90% of routing decisions. Power users ping-pong between tools within projects.

Keeping Your Toolkit Current

AI models update every few months. Here’s how to keep your toolkit relevant:

1. Follow the benchmarks. Chatbot Arena and SWE-bench are the most trustworthy leaderboards. When rankings shift significantly, reassess your routing.
2. Test new features yourself. When either platform announces a major update, try it on your specific tasks before changing your workflow.
3. Reassess quarterly. Every three months, check: Am I still routing tasks to the right tool? Has a new feature changed the equation?
4. Don’t chase every update. New models drop constantly. Only change your workflow when the improvement is meaningful for YOUR tasks.

The three-question framework is designed to survive model updates. Even when GPT-6 or Claude 5 arrives, you’ll still be asking: images/voice? Big context? Code execution? The answers might shift, but the questions stay the same.

What to Do Next

1. Finish your task map from Exercise 1 — list your top 5 tasks and map them to tools
2. Write platform-specific prompts for your 2-3 most common tasks
3. Save your routing cheat sheet somewhere visible
4. Try the ping-pong strategy on your next big project — start in one tool, finish in the other
5. Explore related skills — the AI Skills Directory has prompt templates optimized for both ChatGPT and Claude

Key Takeaways

• Start with YOUR tasks, not the tools — map your 3-5 most common tasks to the right platform
• Same task, different prompts — optimize for each platform’s strengths
• Create a personal routing cheat sheet for instant decisions
• Reassess your toolkit quarterly or after major model updates
• The three-question framework survives model updates — the questions stay the same even as answers evolve

Congratulations — you now know more about ChatGPT vs Claude than 99% of people who use either tool. The difference is you won’t just know which is better. You’ll use the right one every time.