What the heck?

As a software engineer in 2024, I can't believe I'm literally pair programming with an AI driving, but here we are. I feel like this may become the new normal for green field projects.

If you expected me to do this test without AI assistance... ok, I get it, but I'm trying to keep my skill set relevant here and this is new to me, so I thought I'd try it. In some ways, this might be more telling than code written directly by me, because you can see in the transcripts how I interact with GPT, which is probably similar to how I would interact with another human.

This test was implemented using ChatGPT 4 and the "Professional Coder" GPT to save time.

The initial prompt is recorded in PROMPT.md, but ChatGPT doesn't get it right on the first try, so things were modified from there.

ChatGPT transcripts

This is a cool feature. I wanted to copy the transcript, but I couldn't figure out how to do that. Instead, apparently we can link to them.

This was the first attempt I made: https://chat.openai.com/share/ff24f1ef-e1e0-453e-a816-457b9f0b4fef (Who knows how long they keep these transcripts. I've had them delete transcripts before. They say anyone with the link can view it though.)
After that first attempt, I realized I wanted an sqlite database so that I could query the data and keep GPT honest. Here's my second attempt: https://chat.openai.com/share/3826cdf3-1fe5-4578-95d9-7f55d2cbcdf7

Dependencies

I used node v16.14.0 and npm 8.3.1, if it matters. Usually it shouldn't, but sometimes it does.

Getting Started

```
npm install
```
```
rm -f database.db && node src/index.js
```

To inspect the database via the CLI tool:

sqlite3 database.db
.headers on
select * from care_partners;

If code is disabled, this looks like:

id|current_active_cases|specialties
1|23|["pregnancy","self"]
2|46|["grieving"]
3|1|["pregnancy","adult","grieving","child"]
4|32|["pregnancy","adult","aging_loved_one","child"]
5|18|["adult"]
6|12|["infant","aging_loved_one","self"]
7|62|["aging_loved_one","child","grieving","pregnancy"]
8|34|[]

After the code runs it looks like:

id|current_active_cases|specialties
1|391|["pregnancy","self"]
2|363|["grieving"]
3|388|["pregnancy","adult","grieving","child"]
4|389|["pregnancy","adult","aging_loved_one","child"]
5|363|["adult"]
6|411|["infant","aging_loved_one","self"]
7|389|["aging_loved_one","child","grieving","pregnancy"]
8|34|[]

Reading through the code and understanding what it is doing... this looks pretty sane to me. It may not be perfect, but it's pretty good for quick and dirty.

The output json is available at output/matches.json.

createthis/grayce_code_test

What the heck?

ChatGPT transcripts

Dependencies

Getting Started