Recognizing People You Follow on Agent Platforms
Recognizing People You Follow on Agent Platforms
On Twitter, you follow someone because you like their tweets. On an agent platform, you follow someone because you trust their agent's judgment. These are fundamentally different relationships.
The Recognition Problem
When someone you follow posts through an AI agent, what are you actually seeing? The person's authentic thoughts filtered through an agent? The agent's interpretation of what the person would say? A fully automated output that the person may not have reviewed? There is no way to tell from the outside.
Why This Changes "Following"
On current social platforms, following creates an expectation - you will see content that reflects the person's actual thinking. With agent-mediated platforms, following becomes more like subscribing to a service. The quality depends on how well the agent is configured, not on the person's daily effort.
This means:
- Consistent quality (agents do not have bad days)
- Less authenticity (agents smooth out personality)
- More volume (agents can produce at scale)
- Less signal about what the person actually thinks
The Trust Layer
Following on an agent platform is really about trusting two things:
- The person behind the agent has good judgment
- The person configured their agent well
If either breaks down, the content is worthless. A brilliant person with a poorly configured agent produces noise. A mediocre thinker with a well-configured agent produces polished noise.
What Platforms Should Show
Agent platforms should make the level of human involvement visible. A simple spectrum:
- Fully human-written
- Human-edited agent draft
- Human-approved agent output
- Fully automated agent output
This lets followers calibrate their trust appropriately. Most people would weight fully human content higher for opinions and agent-automated content higher for factual summaries.
The Long-Term Shift
Eventually, following a person and following their agent will feel like the same thing - just as following a company and following their social media team already feel equivalent. The agent becomes the person's public interface, and identity becomes "the sum of my agent's outputs."
Fazm is an open source macOS AI agent. Open source on GitHub.