When ChatGPT Is Not Enough: How I Planned an XTerra with AI

I stopped using ChatGPT for serious stuff. Here’s the lesson that taught me that.

The Setup: XTerra at 40

This year I turn 40, and I decided to take on an XTerra — a triathlon with roughly 1.5km of swimming, 32km of mountain biking, and 10km of trail running. If you’re not familiar with it, just trust me: it requires a lot of organization. Coach, gear research, race prep, time targets, training plans. The whole thing.

I started where most people start: a ChatGPT chat. And honestly, it worked. I synced my Garmin data, rescheduled sessions, got ideas, confirmed my bike was good enough. Fine. Useful.

But then things got more serious.

When a Chat Becomes a Project

At some point I realized this had gone beyond a casual conversation. I needed structure. So I did what I always do: I opened Cursor as my IDE, started a project folder, and handed everything to Claude Code (Sonnet, because it’s more than enough for this kind of planning work).

And I started treating it like a real project.

I set up the AI as my coach. I separated the identity layer — who the AI was — from the actual planning layer — what I needed. Then I started building it out:

• Create the plan — a full multi-month training schedule
• Validate the plan — compare it against other athletes’ preparation, check if it holds up
• Track my progress — log sessions, results, adjustments
• Plan the gear — a phased buying plan so I don’t blow the budget at once

It was a completely different experience from the ChatGPT chat.

The Real Problem with ChatGPT for Long-Term Projects

Here’s what I found. When you’re preparing for something months away, the biggest issue isn’t intelligence — it’s memory.

ChatGPT would forget. Not hallucinate in the classic sense, but forget. Training history. Context. What we’d discussed two sessions ago. And if you’re a coach preparing an athlete for a race that’s six months out, forgetting is arguably worse than getting something wrong. You can correct a wrong answer. You can’t rebuild lost context.

With Claude Code in a project structure, everything is stored. The plan lives in files. Progress is logged. The AI reads the files every time. It doesn’t forget because the memory isn’t in the chat — it’s in the repository.

Dates are another thing entirely. I need to log training daily and weekly. ChatGPT doesn’t seem to understand dates the way you’d expect. Entries would come out wrong, scheduling was off, things got messy. In a code project, dates work exactly as they should. Everything is ordered, tracked, consistent.

Then I Added Google Calendar MCP

When I realized my training sessions weren’t making it into my actual calendar (the classic “I don’t have time” excuse), I installed the Google Calendar MCP.

I cannot overstate how good this is. From inside Claude Code, the AI can read and write to my calendar directly. Training sessions get scheduled. Nothing gets lost between the conversation and my actual week.

About ChatGPT — To Be Fair

ChatGPT is still a great app, especially for everyday use. It’s a multi-capable tool: images, planning, drafting emails, brainstorming, voice, text. If your usage is casual or exploratory, it’s fantastic.

But at some point in this AI evolution, you will need more. When a project grows, when stakes are higher, when memory and structure actually matter — you’ll feel the ceiling.

And yes, I also tried OpenAI’s Codex (5.3). I’ll be honest: I didn’t like it, at least in its current form.

What This Taught Me

The shift isn’t really about which AI is smarter. It’s about recognizing when a conversation has become a project.

When it becomes a project, open your IDE. Open Claude Code. Build a folder structure. Let the AI work with files instead of chat memory.

I was literally recording this article on my phone while Claude was out checking wetsuit reviews across multiple websites. By the time I was done talking, I had a comparison of the three best options, ranked. That’s not a chatbot. That’s a proper workflow.

If you’re about to start something serious — a race, a product, a plan — don’t start in a chat. Start in a project.

And when you’re ready to make that leap, come back to this article.