@thorstenball It's pretty rare, but I have hooks set up for plan mode that allow me to review & edit the plan prior to implementation.
Only use for complicated, multi-phase efforts that will span several sessions and touch many parts of app
@mfranz_on Raspberry pi case challenges: RF and Thermal.
Would be interesting to throw in some specs for these and see what it comes up with, e.g. how to counter heat accumulation when running local models 24/7
@engineers_feed ? 6.99 is 6.99 with some margin of error, which depends on the measurement device and conditions. It's also probably too low precision to be useful for anything but house construction
@mfranz_on We were mostly counting people (that's where demand was pre-COVID), but the algorithms could detect most objects, and obviously have improved greatly at that since then.
Counting anything suffers from drift (make one mistake and the count is off), so we had to adjust for that.
@liontarakos@mfranz_on@dee_hw It's not agentic, just machine vision. And the questions you ask are the questions I got as a PM doing this. Test, iterate, improve, and do better than the competitors is the answer