TL;DR
- FAQ, key points, and content brief runs in Moose chat now keep working if you leave the screen, and the finished result is waiting in the chat when you come back.
- Ask Moose to change a FAQ or key points document it generated and it edits that document in place instead of writing a new one.
- Moose chat can read your project's entity graph and answer questions about what your site is built around.
- Quieter improvements: crawl page snapshots are stored compressed, the local model only warms up when your machine has memory to spare, and manually added monitored URLs stay on your project's own domain.
- Our own marketing site took our own advice this week: FAQ sections on key pages and a visible last-updated date on every page.
Stay on the job
Some jobs in Moose chat take a while. Generating a FAQ page means research first, and research takes as long as it takes. As of this week, those long runs keep working when you leave the chat screen. Go check your monitoring results, poke around settings, come back whenever. The chat picks the thread right up, shows you where the run got to, and the finished work is sitting there when it is done.
The run stays. You get to move. As a dog, I appreciate the reversal.
Edits, not do-overs
Generating a FAQ page takes research. Changing one answer in it should not. When you ask Moose to tweak something it generated, it now applies your edit to that document in place and leaves everything else exactly where it was. One answer reworded means one answer reworded, not a fresh document that reshuffles the nine answers you already liked.
Ask about the graph
The entity graph maps the entities your site talks about and how they relate. As of this week, Moose chat reads it. Ask what your site is built around, or how two topics connect, and Moose answers from the graph it built out of your own crawl. You need a completed crawl and a built graph first. Moose is good, but he cannot read a graph that does not exist yet.
Housekeeping
A few quieter improvements. Crawl page snapshots are now stored compressed, so monitoring a large site takes less disk. The app cleans up its own spent installer files at launch, which is the software equivalent of putting your toys away. The local model only warms up at startup when your machine has memory to spare, so the app stays a polite neighbor to whatever else you are running. And manually added monitored URLs now have to belong to the project's own domain, subdomains included, which keeps a project about your site being about your site.
Taking our own advice
We tell people that pages should answer real questions on the page. This week our own marketing site got the treatment: proper FAQ sections went onto the AEO audit, local AI, AI visibility, and workflows pages, and every page now shows a visible last-updated date.
One more thing. There is a word Sam has banned from everything we publish, a filler word that sneaks into AI writing and his own. A sweep found it still loitering on a few marketing pages. It has been escorted out. No, I will not write it here.
FAQ
What happens if I leave the chat screen while Moose is generating something?
The run keeps working in the background. When you come back to the chat, it picks the thread up, shows you where the run got to, and the finished result is waiting when it is done.
How does Moose change a document it already generated?
Ask for the change in chat. Moose applies the edit to that document in place and leaves the rest as it was, instead of generating a fresh version from scratch.
What does chat need before it can answer questions about my entity graph?
A project with a completed site crawl and a built entity graph. Once the graph exists, you can ask about it in plain words in Moose chat.
Why do manually added monitored URLs have to be on my project's domain?
It keeps a project about your site being about your site. Subdomains of your domain are allowed on purpose, so a docs or blog subdomain can still be monitored alongside the main site.
The one who's under the desk for all of it.
