rustyrustling

After 2-3 weeks of going hard at Claude Max - yeah it's got major limits. At first I was trying to make full ready-to-go-to-prod entirely with vibe coding. Then I progressed to reviewing all the code it made. However, it will spend at least a third of it's tokens and my constantly ( 5 hr ) replenishing opus on spinning in a circle - checking random dirs, random files, and unrelated things for the answer to a problem that it THINKS is solution ( and that's not even considering the price increase ( by token reduction and 7 day waiting period lol ) that they will be doing in August.

Then I just stopped letting it run rampant and reviewed every step it took ( this is when I got the best results. ) So realistically - all it would be good at in a complex enterprise software stack ( My primary experience is with Guidewire ) it would be great for quickly scaffolding new parts of a micro service, adding onto pieces of one very specifically, and just removing the grunt work of manually hitting tab, cmd ->, ctrl ->, opt -> through big files and letting me just read it and confirm rather than getting carpal tunnel.

Honestly the best use I've gotten out of it has been updating and adding onto my emacs config primarily used for org-roam.

As far as replacing engineers? After at least 300 hours using max - I can say no it will not. I realized this after I spent more time configuring the rules and prompting the ai than actually just writing the code myself.

Location: Birmingham, AL Remote: yes Willing to relocate: yes Technologies: guidewire, Gosu, Java, C#, Js, Blazor Résumé/CV: https://drive.google.com/file/d/1V2_Nm1N18L9jHhKMTa2CY4eViEb... Email: [email protected]