I like social media, most of the time. But sometimes...things get dark. It started with a question I couldn't shake: why does my timeline make me feel worse even when I agree with what's being said?

The answer, I think, isn't the opinions. It's the texture. There's a difference between "this policy is cruel and here's why" and "lmao get fucked." Both might be directed at the same target, both might even be correct. But one of them is critique, and the other is just...enjoying someone else's pain.

That's spite. And spite is exhausting, even when you're not the target.

So I'm building Kindling.

So what is Kindling?

It's a regenerative layer for feeds that tries to remove cruelty without removing controversy. Plug in the URL of any feed, and we remove the stuff that passes muster with regular moderation, even targeted moderation, but is just plain nasty.

The thesis is simple: you can be angry about injustice, you can mock the powerful. You can even drag billionaires & CEOs. That's attacking up. But taking glee in a stranger's personal tragedy? Celebrating someone's cancer diagnosis because you disagreed with them politically?

That's spite dressed up as commentary. And Kindling is designed to recognize the difference.

The key word is designed. This isn't keyword matching. It's not a blocklist. It's a model that tries to understand context—who's being discussed, what the power dynamics are, whether the post is punching up or down, whether there's genuine engagement or just schadenfreude.

See why. Turn the dial. Your feedback fans or smothers the flames.

Every filtering decision explains itself. You'll see something like: "Hidden: detected glee at suffering (0.78). View anyway?"

You set the threshold. You see the reasoning. You can always override. The algorithm doesn't get to be a black box. If it's going to make judgment calls about your feed, you deserve to see its work. The decisions are complex, the positionalities murky. When you're dealing with culture and context, you have to be transparent about why a decision was made.

Where we are today.

Early. Very early. I'm building this in public because I think the conversation about how we filter matters as much as what we filter.

If you're interested in being a tester—especially if you're skeptical, especially if you have edge cases you think will break it—I want to hear from you. DM me on Bluesky (@chaosgreml.in) or reply here.

This is going to be an ongoing conversation. Welcome to The Fireside. Pull up a chair.

—Bryan