@MrTacticalX I'm not gonna lie, even though I know that most things only need opus at low/medium effort I still have a hard time not going max on everything
It's probably a perfectionist problem
I stopped writing my own prompts.
My projects have never felt more powerful. It really started with Grok Bot.
I gave it my Claude sub and decided to try letting it steer my Claude sessions instead of me. I gave it all the project context, yapped a little to iron out the idea, and then told it to go for it.
I saw the Claude session spin up on my Mac under the guidance of grok bot. I then read its prompt and it was ridiculously clear. Way clearer than I’ve ever been in any prompt that I’ve given an LLM before.
That’s when I decided I’m done prompting myself.
Since then grok bot has done my prompting and session steering for me. I’ve had it start verifying and re-prompting too when the output isn’t what I asked for.
It’s like having a software engineer on my team (and this is great cause I’m only a rookie)
Even if you don’t have grok bot, use your ai to create a prompt for another ai to do the task. Or ask your ai session to spawn a subagent to complete the task you had in mind and then verify the results! The ai’s prompt will be much better than any prompt you write.
@mattpocockuk Ohhh yeah this is crucial for me too. I have a HUGE notion page for all of my company info (ideas, tasks, procedures, PRs, resources)
It's so nice 'cause i can just jump in there and write some stuff down and then reference it to my agents and we're all on the same page!
I stopped writing my own prompts.
My projects have never felt more powerful. It really started with Grok Bot.
I gave it my Claude sub and decided to try letting it steer my Claude sessions instead of me. I gave it all the project context, yapped a little to iron out the idea, and then told it to go for it.
I saw the Claude session spin up on my Mac under the guidance of grok bot. I then read its prompt and it was ridiculously clear. Way clearer than I’ve ever been in any prompt that I’ve given an LLM before.
That’s when I decided I’m done prompting myself.
Since then grok bot has done my prompting and session steering for me. I’ve had it start verifying and re-prompting too when the output isn’t what I asked for.
It’s like having a software engineer on my team (and this is great cause I’m only a rookie)
Even if you don’t have grok bot, use your ai to create a prompt for another ai to do the task. Or ask your ai session to spawn a subagent to complete the task you had in mind and then verify the results! The ai’s prompt will be much better than any prompt you write.