How (A)I Messed Up My Data Gathering

Okay, buckle up...

Earlier today I dropped my post sharing my thoughts about the PSF Grant changes. Let's talk about how things went wrong.

I used AI to get my data points

After scrubbing through the PSF board resolutions and trying to list all of them. I quickly gave up and asked Claude by Anthropic to do it for me. I was reminded that "Claude can make mistakes. Please double-check responses." and I did that. I did a search and find for each of the points in the resolutions and indeed I did find them. Claude didn't add any fake data or amounts, but it didn't give me ALL the data points.

It only gave me half the data. Later on my worry of basing my results on AI led me to run through the data points again. I caught my mistake, issued an apology, cried for a second because I was embarrased and then started writing this.

This isn't me saying don't trust AI (although, don't trust it without some validation), although my trust in it parsing text that I give it has been eroded a little. It still serves as my proofreader and tone-checker for my blog.

Personal brand and being the angry Black man

In some ways I have taken it upon myself to be the loud person in the room wanting to examine the actions of the community that I love. Some of this is because of concern. I would be dishonest if I didn't also attribute some of it as frustration in the feeling that few people really don't care about the Python community if it doesn't benefit them.

There is a fear and stereotype that I cannot stand and that is the one of the angry black man. Any time I veer into believing that I'm doing this, I reach out to people I trust to be honest with me and they let me know, "Jay you're being ridiculous". That didn't happen here but mostly because the majority of the people that I talked to looked at the data that I gave them (with no reason to believe the data was missing information) and had the same concerns I did.

I really tried to approach that post with empathy for the PSF and the Grants Working Group. I don't know all the details but I know that this stuff is incredibly hard and a blog that has been proofread several times and fact checked. Will talk a little more about how hard this stuff really is.

Why even write the other post (or this one)

That is a simple answer. I'm trying to write more.

By writing more and having more conversations in the last year I have learned so much. I've also got this unknown (because I don't track analytics) popularity thing going on with my blog. To put is simple...

I want conversation around things to happen so badly

My co-host of Python Community News, Jon, asked me why I keep bringing up PCN from time to time and that is because I truly miss it. I think we had this great dynamic and was covering a part of the Python community like no one else was. We had so many inspiring ideas and deep dives that we wanted to investigate and report on. This was something that we talked about all day¹. We tried to look at pieces and understand the situation and talk about what we could do. It felt so great and I was happy.

Here's what I told him.

"It's the thing I want to bring back most but have no bandwidth for I even thought about it as a blog/newsletter but was like that thing would be posted to like 1-2 times a quarter. And all the (reasons) I want it back for are the things that take the most time. Like finding conferences and things and talking about this kind of stuff. " ... "I feel like the biggest thing is that when I talk to you about stuff it always feels like there is so much good information that needs to be shared with more than me. You do a wonderful job of sharing it with folks in Conf-Organizers Discord, but I'm horrible at getting the infromation out past me.

"trying to fix it by getting things off my plate (which is why bringing back PCN is not a good idea at the moment) A part of me also thinks it's a muscle thing. The more I write about the things I see and the more I share, the easier it will be but it's still hard."

Sadly, I'm Python Community popular which is great at PyCon but I cannot pay my bills off of PCN downloads and Black Python Devs volunteer work. This means that I'm burning the candle at all ends again balancing trying to excel (or at least not get fired) at work and running a successful non-profit, all while continuing to want to write more code, learn more Python and about the community and package management, and inspire a generation of Python developers that will have to compete with AI for Python jobs (although I'm taking a point away from AI after this debacle, even if it was my fault).

I wrote this because again if I hold myself accountable then I can do and get better at all of this. I removed the following section from the last post because while spewing my thoughts onto the screen I couldn't find a place for this:

We've learned that funding needs to go into building expertise. There are a few events and initiatives that BPD has chosen to support knowing that it won't be as successful as we hoped it would, but it's important for them to have the ability to learn by doing and that over time things will improve. This willingness to let them learn and develop does affect how much we support, but it never prevents support.

I hope my fumbles and foibles will never prevent your support.