Not really a bug, but weird change since last upgrade + upgrade to Golden Gate 27
Closed#66 · by User #d8a6 · Oct 7, 2026 · Released v1.0.0-rc.40
Not really a bug, but weird change since last upgrade + upgrade to Golden Gate 27.0.1 - guessing (with no context - just assumptions) that Local AI memory/cores are competing with Siri behind the scenes?
since the last update local processing is VERY slow - what would normally take about 5 seconds to process a local transcript is now taking upwards of a couple minutes after the recording stops.
it's happening on both my M4 with 24gb and my M2 Pro with 64gb - it feels like a low level setting/config change might be needed, or Siri is just a hog behind the scenes?
Comments (2)
Ohhh dangit that's a bad one. Geez ok getting something out now must be releated to the mdm policy work
GLM actually got to the root of this one:
"why since the upgrade". Apple's model has to be available on the system before our summary runs at all. If it wasn't ready on his previous macOS, his Apple choice silently did nothing, and stopping was fast because there was no summary. macOS 27.0.1 made the model live, so the same setting suddenly costs minutes.
So the fix: You'll want to change all those selectors from "Apple (on-device)" to "off". That label is confusing too, it should say Apple Intelligence... which I haven't found to be usable. It's more like a parked experiment at this point.
Having supporting features show Gemini when you have the meetings set to offline is also confusing and looks like it could have been a privacy issue but the search would have asked before sending with offline enabled.... confusing. I need to make a bunch of UI fixes.