Our first topic of the night was a recap of this year’s Raleigh Day of Data. We talked a bit about how the event went and our lessons learned. The biggest lesson learned was, don’t schedule a conference on Memorial Day weekend.
Language Models and Malicious Helpfulness
Our next topic came from Ars Technica, which reported on Meta’s support chatbot being so friendly that it would allow random people to change the e-mail address associated with Instagram accounts. Along the way, I spent a good bit of time covering different services for rooting out likely LLM-based writing. My favorite at this point is GPTZero.me, which is designed more for educators but it does a good job of determining if long-form text comes from a human or a bot.
Day of Data Badge Generator
The final topic of the evening was my discussion of the Day of Data Badge Generator, a simple application I threw together (with the significant help of Claude Code). Its purpose is to make it quick and easy to check in a person at a Day of Data-style event and print out a thermal label that you can affix to a placard or badge holder. My favorite such placard, by the way, is a set of dry erase index cards. They’re about 20 times cheaper than custom badges, and much more durable than cardstock. For a free conference, I think they do a really good job, even if they required people to think hard in the morning about which color to choose.
Our first topic of the night was a sales pitch for Raleigh Day of Data 2026. We talked a bit about the speaking roster, event details, and a bit of behind-the-scenes operations.
Our first topic of the night came from Mala, who couldn’t make it this evening. The people behind ggplot2 (implementing the Grammar of Graphics for R) have come out with a SQL variant called ggsql. It’s still pretty early on in development, but it does look promising.
DROP USER and EXTERNAL MODEL Permissions in SQL Server
Our next topic comes from Andreas Wolter and deals with a permissions bug in SQL Server. To me, it looks like certain permissions get cached (or maybe not cleaned up correctly) and can lead to people having a bit more access to specific external models than you’d think.
The final topic also came from Mala and covers SQL Server’s MCP offering. Jerry Nixon describes using the Data API Builder (DAB) to serve as a database intermediary, so that you don’t offer up direct database access to a language model.
Our first topic of the night was an update on Raleigh Day of Data, coming up on May 23rd of 2026. You can still sign up for the event at theĀ Day of Data website. Attendance is free and lunch is a nominal fee.
I also spent a bit of time talking about a pair of posts from Anthony Nocentino on the SQL Server Kubernetes operator, as well as planned failover. Since the time of recording, Anthony also put together a post on unplanned failover.
Jeffrey Snover has a great rant on how Microsoft has had multiple, overlapping but incompatible strategies for implementing graphical interfaces on Windows.
Along the way, I got sidetracked a bit on Trusted Platform Modules and my annoyance at Windows 11 for not working without a module.