Alexa is super buggy now anyway. I switched my Echo Dot to Alexa+, and it fails turning on and off my Samsung TV all the time now. You usually have to do it twice.
This has been my impetus to do Home Assistant things and I already can tell you that I'm going to spend far more time setting it up and tweaking it than I actually save, but false economy is a tinkerer's best friend. It's pretty impressive what a local LLM setup can do though, and I'm learning that all my existing smart devices are trivially available if anyone gets physical access to my network I guess!
This is the kind of thing Claude Code (bypassing permissions) shines at. I‘m about to setup HA myself and intend to not write a single line of config myself.
Something I love about HA is that all thr gui can always be directly edited using yaml. So you can ask claude for a v1 then tweak it a bit then finish with the gui. And all of this directly from the gui.
Ugh. Reminds me that some time ago Siri stopped responding to “turn off my TV.” Now I have to remember to say “turn off my Apple TV.” (Which with the magic of HDMI CEC turns off my entire system.) Given how groggy I am when I want to turn off the TV, I often forget.
How can this be? I had great luck with GPT3 way back when… and I didn’t have function calling or chat… had to parse the JSON myself, extraction “action” and “response-text” fields… How has this been so hard for AMZN? Is it a matter of token cost and trying to use small models?
that's a reasonable theory. they've likely delayed the launch this long due to the inference cost compared to the more basic Alexa engine.
I would also guess the testing is incomplete. Alexa+ is a slow roll out so they can improve precision/recall on the intents with actual customers. Alexa+ is less deterministic than the previous model was wrt intents