Ironically enough - ai is really good at making small tools reliably. Feel like some form of AI automation that builds the automation/code live - then slots it in and tweaks it as it goes - is more the future. Self healing systems basically. Like giving AI access to AOP in a Java system.
It will bootstrap a socket, and with its accompanying claude hook I call the plugin with the latest adjusted file and lines change in nvim. You can follow along as Claude works on your files and see nice in line change notifications. Works great with folke's sidekick.nvim github.com/folke/sideki...
So I've been looking for the right workflow for claude. I'm a big fan of the cli, and I really dislike vscode. So with the help of Claude I wrote an neovim plugin called claude-follow-hook github.com/zach-source/...
One of the best parts of being a dad is randomly giving your random collected shit from your many years of life and the child acting like it’s your the best thing they’ve ever received. Not sure if my wife appreciates the nooks of random stuff in their room now but I love it.
This is pretty good. I’d just add like baked in support for SPIFFEE/SPIRE and/or multi cluster service account definitions. Federated auth basically out of the box.
I’d also love ABAC policies or the way to split cluster roles a bit more granularly. Like support filtering by namespaces with labels.
Well sort of - you want your otel data to go to a storage backend you have to convert it from otel types to something the backend will accept. So sometimes you want more middleware like Kafka or rabbitmq. Sometimes open source like Prometheus or zipkin. Sometimes a paid for platform like datadog.
Eks to 1.26 crashes kube proxy and totally disrupts traffic for about an hour. All versions of plugins the same before and after. Control plane upgrade shouldn’t have caused kube proxy to crap the bed.
If it’s networking on the kube service address space - check/restart kube proxy. If you can access the nodes with the trouble pods - try checking the ip tables.
