“Here, your words matter - to readers who always listen”

Alternatively go and find a nice railway bridge to stand under so you can shout loudly and hear your voice echo back at yourself. No code required, and it gets you out of the house. You might even meet an actual human on the way.

This seems like a surefire way to build a private echo chamber. I doubt a simulated audience is going to challenge your thoughts or help find what you are trying to say.

Can I include an asshole AI? Or one that nitpicks my grammar? Also need someone to explain how I am wrong about nearly everything and should give up writing? Otherwise it would feel super fake.

Hey team! I find journaling for a fictive audience to be more effective personally; since it forces me to try digest my thoughts for an external listener.

Then why not add the actual fictive audience through LLMs? That's how this was born. Feel free to leave your thoughts/feedback here.

I think it's pretty neat. It'd be cool to have recurring commenters with different personas who focus on different things. I wrote a test post and almost all of the comments focused on one detail to the exclusion of others. With different personas/focuses you'd get more variety in responses. I also think responding to comments would be a natural next step. Or if those commenter personas are persistent, you could @-mention them in a follow up.

This is actually genius and obvious and also feels really, really weird, all at the same time.

I do the opposite of this (release stuff publicly and don't have comments enabled) but this looks really nice for a specific kind of audience. I'm curious to see how it'll evolve and be received, good luck!

Given the rapidly diminishing quality of discourse on the open internet the last decade or so, I understand.

I recently restricted comments on my blog or 15 years to existing subscribers only. It took me a while to accept that after removing the random spam, then the racist, misogynist, homophobic, and other lowlife commenters, I was left with 10-15% of discussions worth reading.

Pretty sad state of affairs, and it’s clear it’s degrading faster every day.

This feels like one of those moments where you're witnessing history be made in real-time. And I hate it.

It takes approximately zero effort to see how this could be both monetized and used for harm. This episode of Black Mirror is writing itself and congratulating itself in the comments.

I appreciate you sharing this and it looks made with care. It also seems like a cool take on how to use LLMs.

This isn’t for me, but thanks for sharing.

A good extra use-case could be to test an article out and tease out any flaws before releasing it to a live audience.

this could be good to help someone steelman their arguments, highlight implicit assumptions, identify communication that's confusing, etc. could be interesting to get reactions from a random sampling of synthetic personalities from within a target audience set.

What happens when you trigger the models censorship function?

Does it comment that it can't comment on that?

Honestly, fine. I think some commenters have a point (on both sides). But really y'all should just try talking to yourself if you think you'll like this.

OP, pay no attention to the haters. Brilliant idea, especially given the slop out there. For my journaling, I've been experimenting with similar.

That being said, whatever I do I'm going to keep completely in-house, as in offline with local LLMs only, because privacy.

This is sarcastic, right? I refuse to believe people are doing this for real.

Thanks. I hate it.