It probably has its use-cases, but in platform engineering there is rarely a scenario where we are allowed/can trust using an LLM-based technology to improve efficiency. Usually it throws out well-formatted YAML manifests that look great, they work as expected, but are super insecure (CAP_SYS_ADMIN for no reason). You could keep prompting it until you get the result you want, but likely its just quicker if I do the job myself.
I think people often overlook the value of crystallizing what they actually want. Coming from a product management background, it is hard for many to clearly define or think through all the things they care about in a task.
AI helps eliminate the blank page problem — and it’s always easier to critique something than to think from scratch :) I often find myself going back and forth with the AI, debating, complaining… and then starting over — sometimes with it, sometimes without. But by that point, I know much more clearly what I want and what really matters for the task.
3 comments
[ 2.9 ms ] story [ 8.0 ms ] threadAI helps eliminate the blank page problem — and it’s always easier to critique something than to think from scratch :) I often find myself going back and forth with the AI, debating, complaining… and then starting over — sometimes with it, sometimes without. But by that point, I know much more clearly what I want and what really matters for the task.