What is phantomcrowd-simulacra?

PhantomCrowd Simulacra, also called EchoHerd in its README, is a tool that simulates how a piece of content, such as a post, article, or marketing message, might spread and change as it moves through an online audience. Instead of just measuring likes or clicks after the fact, it creates a population of simulated persona agents that read your message, react to it, and talk to each other, generating synthetic conversation chains that show how your original meaning might get amplified, ignored, or twisted over time. You start by pasting in your seed message, then choosing or generating a population of agents, each with its own backstory, social connections, attention span, and memory. As the simulation runs, agents respond to your content and to each other's reactions, and the tool tracks how the conversation drifts away from your original wording, plus how quickly interest in the topic tends to fade. You can pause a simulation midway to inject a correction or a different angle and see how that changes where the conversation goes next. The README lists use cases such as testing marketing slogans and press releases before launch, modeling how policy language might land with different groups, previewing reactions to a product announcement or company policy, and studying rumor and narrative spread for academic research. Every agent can be tuned with settings like influence weight, skepticism, and memory decay rate, so you can explore what-if scenarios, for example testing what happens if your most influential simulated agent turns skeptical of your message. The system runs entirely on your own machine using a local, Ollama-compatible language model such as Mistral, Llama 3, or Qwen, so no data or content leaves your network and no cloud API is required. Results are shown on a web dashboard with a node map, a timeline you can scrub through, a sentiment gauge, and per-agent conversation logs, and simulation runs can be exported as JSON, CSV, or animated GIF for later review.