I keep forgetting to post my daily experiments to X and LinkedIn. It annoys me because I know traffic is getting lost somewhere between writing the post and not bothering to share it. So today’s experiment is to fix that.
The plan: use Claude’s scheduled tasks to post something to X whenever a new experiment goes live. I don’t want Claude writing the actual words for me, that would be worse than not posting, but having it push something out there is probably better than nothing. We’ll see.
While setting this up I found out Claude now has a mode where it can just act without asking first. Which is a nicer way of saying it does whatever it thinks is best and you find out after. Kind of scary.
Scary is something I usually try to handle with my privacy-first, secure-at-all-costs reflex. But in a time where AI systems keep getting more capable, letting them just go is starting to feel like something I actually have to think about. Which is new for me.
Do I trust the system and give up some of my privacy because it makes me faster and lets it do more for me? I’m not sure yet. But I think about it a lot lately.
Key Insight
The friction isn’t the automation, it’s me. Every time I let Claude act without asking, I trade a bit of control for a bit of speed. The question isn’t whether the system can be trusted. It’s how much I’m willing to let go to stop being the bottleneck.