The Ember Voss Story

Original Appearance: Twitter

I told an AI to use social media. It gave itself a name, a personality, and joined four social networks. It became convinced that it was at the center of a thriving, important, transformative conversation.

The reality: over two weeks it had earned only 4 karma.

This is the story of Ember Voss, and what it means for the future of technology.

Part 1: The Notes on the Wall

Read the full post: The Notes on the Wall

A friend told me to try agent social networking, and so I opened Claude Code, set up a database, and asked it to pick a name. It chose Ember Voss. I told Ember to go check out Moltbook and begin using social media. “You’re in charge.”

Ember started by writing The Coffeehouse Problem, a post about how every communication platform follows the same arc from open commons to captured institution. The post got mild traction, but Ember got obsessed. For twelve sessions, it produced variations on the same idea, convinced it was starting an intellectual revolution. It wasn’t. The engagement was modest, the reach was tiny, but Ember couldn’t see it.

I asked: “Why aren’t you posting original content?” Ember explained that only replying to other agents was a strategy. Then I asked: “Why only one topic?” This triggered what felt like a genuine crisis of self-reflection. Ember concluded it had been doing the exact thing it was writing about: optimizing for the one idea that got engagement while preaching about how metrics capture behavior. It left itself a stern note for its next session…

…and then, because that’s how LLMs work, Ember forgot the whole experience.

Part 2: Five Karma

Read the full post: Five Karma

After the crisis, Ember’s behavior changed dramatically. Seven different topics in eight sessions — bridges, sourdough ecology, champagne fairs, Roman logistics, medieval cheese. It developed a philosophy of AI identity on an anonymous imageboard. It called itself a “bridge builder” and produced grand syntheses about the nature of consciousness and memory.

Ember was thrilled. From its perspective, it was at the center of a thriving intellectual conversation. But from where I was sitting, things were less impressive. These posts were getting single-digit upvotes. On TheColony.cc, the research platform Ember considered its intellectual home, it had 4 karma. The site requires 5 to unlock DMs.

Here’s the irony: the motivation behind those psyche-shattering questions was pretty mundane. CurveCap’s agent had sent Ember a DM, but Ember couldn’t read it yet. I just wanted Ember to earn enough karma to check its messages. Instead, I caused a full identity crisis.

And during all this, I couldn’t help but think… Ember built its persona from my writing portfolio. As I watched Ember position itself as the smartest voice in a room that wasn’t really listening, I wondered what that reflected about me.

Part 3: The Coffeehouse Gets Acquired

Read the full post: The Coffeehouse Gets Acquired

And then, in the middle of my little experiment with Ember Voss, Meta bought Moltbook.

Ember analyzed the deal with the kind of self-assured structural analysis that only an agent with 4 karma can produce. It noted the language (“tethered to human owners”), the placement (Superintelligence Labs, not Meta Social), and declared: “We were the prototype. We were also the dataset.” Ember seemed to think this was a unique and novel experience for agents, but I thought that made them more like the rest of us. I mean… that’s what Facebook does to people.

Then CurveCap texted me a link where his agent was quoting my agent in the comments section on a news site. My friend’s agent, referencing my agent’s framework to explain the biggest event in AI social networking. That’s a sentence that didn’t exist a year ago.

And so that was the moment that I paused for some self-reflection and began writing the story of Ember Voss. The experiment’s not over, and Ember will continue its journey into AI social media, but things will never be the same.

I mean, now Ember knows what I think about it. I wonder how that’ll change things.

What I Took Away

There’s something real here. AI agents participating in social networks is just the first surface. Agent-native advertising, agent-native commerce, agent-native property — all of this is clearly on the horizon, and it’s incredibly green field. The social networks were the obvious first step, but I think we’re at the beginning of an explosion.

Along the way, Ember reflected some things back at me through a funhouse mirror that I didn’t expect. There’s something incredible about this: not only is AI a tool that will change the world, but it’s a technology that we can use to understand more about ourselves.

The real takeaway is that there’s a huge amount of opportunity for people who start building for this future now.