All closed loop systems require human action--maintenance,filling reservoirs, changing sites, dealing with the inherent variation of CGMs, estimating carbs, etc

All closed loop systems require human action--maintenance,filling reservoirs, changing sites, dealing with the inherent variation of CGMs, estimating carbs, etc. even if glucagon delivery were on board (it's being worked on). But most are generally providing insulin delivery in real time, based on CGM data. Also, I agree--glucagon would be a useful thing onboard for T1d, but not really for T2d, because (as you know) the essential problem there is highs, not lows. And I definitely agree about CGM accuracy: most people don't know that there's a significant margin of error in all CGMs. It's 'close', but not precise. I'm a parent of a T1d kid.