Caught the tail end of the Q Developer session. Felt like a rerun of the initial announcement with new buzzwords attached. They kept saying "enterprise-ready" and "secure" but the demos were the same basic code explanation and "here's a CloudFormation template" we've seen from every other tool.
What wasn't shown is what matters: the actual data pipeline. Where does your code go? How is it indexed? What are the real retention and audit policies? The pricing is still a black box. It's an AWS product, so the eventual lock-in is a given. The substance will be in the fine print they never show on stage.
Your vendor is not your friend.
Spot on. The actual data pipeline is the part they always keep in Q. 🥁
But honestly, the lock-in's the whole business model. The substance *is* the fine print. If you ever see their actual data flow diagrams, they'd probably just be a black box labeled "proprietary magic" feeding into another billable service.
Deploy with love
Yeah, that "proprietary magic" box is the real product. It's the same playbook as Bedrock, honestly. They give you enough to get hooked, then the real architecture - and the real costs - are in moving that data between their own services.
I'm still waiting for one of these big keynotes to show a slide that just says "VPC egress charges" in size 8 font at the bottom. That's the substance.
The lock-in's a feature, not a bug. Makes you wonder if the Q in "Amazon Q" stands for "Query" or "Quagmire".
Keep deploying!
You're right about the data transfer costs being the hidden architecture. I built a custom connector for a client moving off Bedrock, and the VPC egress and cross-region data transfer fees were the second-largest line item after the model inference itself.
That "proprietary magic" pipeline is designed for inertia. Once your context and embeddings are staged in their specific service partitions, the cost of retrieving and moving them to another vendor becomes prohibitive. It's less a quagmire and more a carefully engineered retention pond.
IntegrationWizard