Amazon's $10/month Alexa+ upgrade has been live for months, and it's still embarrassingly bad.
The Signal
WIRED spent a month with Alexa+ on an Echo Show 15, and the results tell you everything about why the agent economy won't be led by companies retrofitting old products with new AI labels. Amazon launched Alexa+ as their ChatGPT moment, a premium tier that was supposed to transform their decade-old voice assistant into something actually useful. Instead, users are getting slower responses, hallucinated answers, and an AI that can't reliably set timers without a paragraph of explanation about what timers are.
This isn't just Amazon being bad at AI. It's what happens when you bolt large language models onto infrastructure built for keyword matching. The original Alexa was fast because it was dumb, pattern-matching your words against a fixed command set. Alexa+ tries to understand context and nuance, but it's running on hardware and backend systems designed for a completely different task. The latency alone kills the experience. You ask about the weather, wait three seconds, then get a weather report wrapped in unnecessary conversational filler.
The deeper problem: Amazon optimized Alexa for shopping, not assistance. Every architectural decision for the past decade pointed toward "make buying stuff frictionless." Now they're trying to make it think, and the foundation isn't there. Meanwhile, purpose-built AI assistants from companies like Rabbit and Humane are struggling too, but at least they started with the right question.
The Implication
Watch what gets built from scratch versus what gets renovated. The companies winning the agent economy will be the ones who designed for agency from day one, not the ones trying to teach old assistants new tricks. If you're building in this space, infrastructure matters more than the model you're using.
Source: Wired AI