🔧 Must-Read for Anyone Who Wants to Avoid Simple Mistakes!
How to Create “Leak-Proof” Specifications Using ChatGPT
■ A Sudden Realization of a Critical Oversight
Lately, I’ve been watching tons of YouTube videos related to robot design.
One day, I stumbled across a video where someone mounted a generative AI on a Raspberry Pi 5—and boldly attached a fan directly for cooling.
That’s when it hit me:
> —— I completely forgot to account for a heatsink in my design!
Even though I’d been in a bit of a creative slump, this was a classic oversight.
I had completely neglected the cooling system.
■ Why Did This Mistake Even Happen?
When I consulted ChatGPT, it told me that existing heatsink products could be used, so luckily, things didn’t turn into a disaster.
But then the question arose:
> Why couldn’t I prevent this mistake, even though I was using ChatGPT?
In the end, the lack of cooling components in my parts list was just the surface-level issue.
The real problem was that I hadn’t included any consideration of thermal management in the spec sheet itself.
■ So, What’s the Solution?
Reflecting on this, I tested three different approaches with ChatGPT to create a “mistake-proof” process:
✅ Attempt 1: Create a checklist
Too serious—doesn’t fit well with hobby projects
Feels less like AI, more like a manual task
✅ Attempt 2: Build a dedicated AI for spec writing
Tried something similar with Gemini
Managing multiple AIs seems like too much effort
✅ Attempt 3: Let AI ask questions, and specs are generated based on the answers
This was fun!
Felt very "AI-like" and matched my workflow
I actually wanted to try this!
By the third attempt, I finally found an approach that excited me.
■ Trying the Q&A-Based Spec Creation
Here’s how ChatGPT started the conversation:
> 🤖 Step 1: Basic Design Questions
[Q1] Where will this robot be used?
[Q2] What is the robot’s main purpose or function?
[Q3] Are there any size or weight restrictions?
[Q4] What kind of movement system will it use?
[Q5] If you’ve already decided on any key components, please list them.
Just by answering these five questions, ChatGPT automatically generated a basic design specification (v1)!
You can see my answers and the generated specs here:
👉 [Link to the Spec Document]
“AI Questions and Spec Sheet”
Excerpt:
Q1 Where will this robot be used?
→ Indoors, with no steps or level changes.
Q2 What is the robot’s purpose?
→ For studying ROS2, and testing odometry using LiDAR...
(etc.)
■ In the End, It’s Still Your Responsibility
Honestly, checklists are probably the most reliable way to prevent careless mistakes.
But since this is a hobby project, I didn’t want it to feel like work. That’s why the Q&A format worked perfectly for me.
Regarding the heatsink issue—thanks to this approach, ChatGPT remembered my past mistake and included cooling measures in the new spec sheet.
I’ll examine how effective this method is in more detail another time.
✨ Summary
> “The best way to prevent mistakes is to bake them into the specs from the beginning.”
Letting AI ask the questions and build the specs might be a surprisingly good fit for a lot of people.
——Before you find yourself saying,
> “Wait, I forgot the heatsink!”
why not try reviewing your design specs with AI？