I’ve always been a huge supporter of Alexa.

They did an amazing job making home automation voice accessible and they built a platform where developers could extend the device and make it their own. It was only natural that their language processing needed to grow up and adopt AI/LLMs to go from something that required a mapping of hard coded utterances to intents to be more natural. However, Alexa++ is not only one of the least capable agents out there but at the same time they have abandoned the developer and customizer in favor of custom integrations with only large partners. Favoring few quality integrations with services people sometimes use is not better than breadth when that was your only advantage before. Any decent developer can connect a modern LLM to a skills framework and get it to work far better than Alexa+. Hell, I’m sure Codex can generate one for me in a weekend. I’m sure it’s all cost and latency related but those trade-offs have really ended up with Alexa– not +.

To their credit they did finally enable legacy skills but even the basics don’t work, session continuation never times out so your skills are left as the context until you have to say “thank you” to get out, invocation works 1 in 3 times and the new API is still locked behind invite only.

What’s worse is for those of us who reverted back to Alexa, the service is degrading every day. The stuck in a session bug is now in old Alexa and even more requests than before have resulted in “I’m sorry I can’t help with that”. Even the age old termination phrases of “exit” or “stop listening” do not work.

I personally can’t wait until OpenAI comes out with their own device and wipes the floor with both Amazon and Google. In the mean time Amazon will be busy building another cheap Android tablet to save the world. Apple, well, we will see.

Blah.