Damn! Beat me to it!

I've had Claude Code for a while now.

Figured it out.

You do all the work, then as climax is starting, but you're exhausted, you bro tag it in and cuck. It's got this.

Then one of two things happens:

1. Climax. No great. Standard, really. As expected. But you got a drink of water in that time, the job's done. It's hard to fuck up at that point. You just let it know what it should've done instead. But it's okay, first time for everyone. Just ignore the fact you're making yourself redundant. All good.
2. Spend more time halting things, explaining how to do the deed, as you and the project make eye contact expressing how deeply unsatisfied everyone is.Just ignore the fact you're making yourself redundant. All good.

So. If you want to use an LLM as a primary coder, do 80% of it yourself and expect these two outcomes.

Otherwise, sit in the corner and explain to your project lead that many screams happened but the safeword was between Claude and the project, so how could you know when it went too far? You burned tokens as requested.

Claude will remain employed but you will not, obviously. Unless you ignore this advice entirely and old school it, in which case Claude will remain employed but you will not BUT you'll be reemployed to go through everything above again.

Edit: Honestly, if you want a code assistant—I get it, the annoying things of coding are no longer annoying—go local.

With my tongue 😈

Tsoding calls it "software prostitution"

No, no, that's prostitute programming.