The Recording
The Panelists
- Kevin Feasel
- Mala Mahadevan
Notes: Questions and Topics
ABORT_QUERY_EXECUTION
Our first topic covered a new Query Store hint coming to SQL Server 2025. Joey D’Antoni has a great write-up of the feature and we used that to launch into a discussion.
There are some specific uses for this, but it’s a pretty big hammer, as Joey indicates. I see it as more of a last resort for recalcitrant queries versus something people would regularly apply.
Goodbye, AOL
Our next topic was on America On-Line ending its dial-up service after 34 years. We reminisced a bit, discussed alternatives (generally in the sense that broadband is near-ubiquitous and satellite is way different from the days of yore), and I griped a bit about why FiOS left North Carolina.
AI is Chef Mike
Colin Cornaby provided the next topic with an outstanding satirical post around AI as developer of everything. Frequent Shop Talk listeners know that I’m pretty negative on generative AI as the basis for long-term success doing most tasks. I think this post summarizes much of my complaint, while implying that there is good utility in the concept and people will probably continue to (over)use it. 10/10 great analogy.
Measuring the Impact of Generative AI on Open-Source Developer Productivity
This topic comes from Mala, who got it originally from Mark Wilkinson. There’s an interesting paper on arXiv covering a study of 16 senior-level open-source developers and how generative AI tools impacted their ability to do specific tasks. The sample size is small and these are very experienced people, but it was interesting to see that generative AI tools made them less productive in their tasks, whereas they believed they were more productive.
Developer Trust in AI Tools Down in 2025
The last story we were able to cram in comes from Ars Technica and covers the Stack Overflow 2025 Developer Survey. The Ars post focuses on one interesting finding: that as more people are using generative AI tools more often, the trust in those tools has dropped considerably. The main problem is that code generation tools will often do something that is close to what you expect but not quite, meaning you have to spend considerable amounts of time carefully reviewing the code, so at that point, are you actually more productive? Well, see the above story for one take…