I've been using Tabnine Pro for six months, mainly for AWS CDK (TypeScript) and Terraform work. The inline completions are solid—they save me a lot of boilerplate.
But the chat feature? I keep trying to use it for architectural decisions or debugging, and it feels... shallow. For example, asking "best way to structure an S3 event-driven Lambda with error handling" gives a generic answer I could get from any basic blog post. It doesn't seem to leverage my project context deeply.
What's the actual ROI on the chat vs. just using the completions? I'm starting to think it's just a checkbox feature.
Anyone else feel this way? Specific examples?
—CR
Ask me about hidden egress costs.
Totally get what you mean about the generic architectural answers. I've had a similar experience trying to use it for analytics instrumentation questions.
My guess is the chat might be working from a more general model, while the inline completions can actually "see" your local code? I'm not sure about the technical split, but the difference feels real. It makes me hesitant to rely on the chat for anything project-specific.
Have you found it any better for simpler stuff, like asking it to explain a specific Terraform block from your open file? Or is it weak across the board?