release notes update

Author: agentmorris

docs/release-notes/mdv1000-release.md

@@ -72,7 +72,7 @@ Based on the signals I can discern from anecdotal interactions and from the lite
 
 Along the same lines, MDv5 would still be just a random AI model on the Internet (or might not have happened at all, i.e. we might have never gone past MDv4) if not for the five-year-plus partnership we've had with [Saul](https://saul.cpsc.ucalgary.ca/) and [Timelapse](https://timelapse.ucalgary.ca/); Saul has been the voice of the user a zillion times in figuring out the best way to bring improvements in tooling to real user workflows, and has continually adapted Timelapse to be an "AI-ready" tool since the MDv5 release.
 
-So, tl;dr, I still run about 100M images per year for users, but now it's a pretty small percentage of MD use overall, which is right where it should be.  I like to handle the difficult cases that move the tools forward, including the cases where MegaDetector doesn't work well, or the cases where it works but requires some special handling or new postprocessing features, or the cases where someone has an intractable backlog of images that they need help digging out of, *after* which they can transition to a self-serve approach.  But the "default" now is "self-serve", which is a huge step forward from 2022!  And that change is due more to the software ecosystem than to the difference between MDv4 and MDv5.
+So, tl;dr, I still run about 100M images per year for users, but now it's a pretty small percentage of MD use overall, which is right where it should be.  I like to handle the difficult cases that move the tools forward, including the cases where MegaDetector doesn't work well, or the cases where it works but requires some special handling or new postprocessing features, or the cases where someone has an intractable backlog of images that they need help digging out of, *after* which they can transition to a self-serve approach.  But the "default" now is "self-serve", which is a huge step forward from 2022!  And when I am running MD for users, the goal is to "graduate" almost everyone to "self-serve", which wasn't the case in the MDv4 era.  That change is due more to the software ecosystem than to the difference between MDv4 and MDv5.
 
 This has been a lot of text without any images, so I'm going to add some gratuitous logos here to double down on the importance of the ecosystem for running AI models: