Mitch's Multifarious Three-Headed Hydra Chatbot is an interactive demonstration that shows how large language models generate text on a per-token basis. Users submit a question and watch three models produce a collaborative answer one word at a time. The models are Claude from Anthropic, Gemini from Google, and Llama from Meta.
Each model reads only the text generated so far before contributing its next word. The interface displays color coding so that the contribution of each model is visible. The demonstration highlights that the models have not been instructed they are participating in a turn-taking exercise. It illustrates that separate thought processes can occur for each individual token without any persistent consciousness across the full response.
The experiment addresses the common assumption that LLMs maintain something akin to consciousness throughout a session. By forcing each model to operate with no memory beyond the preceding words, the demonstration shows that such persistence is unnecessary for coherent output. Caveats note that models could theoretically plan ahead for future tokens when selecting the current one and that the setup does not prove the absence of consciousness, since humans could perform the same exercise while remaining conscious.
Three-Headed Hydra Chatbot is an AI & ML product. It focuses on demonstrating that LLMs generate text token-by-token without persistent consciousness or long-term thought across words. Three-Headed Hydra Chatbot is a consumer product aimed at AI enthusiasts and researchers. Three-Headed Hydra Chatbot is free to use. It runs on the web.
Mitch builds and maintains Three-Headed Hydra Chatbot, and the product first shipped in 2026. The project is developed in the open on GitHub with 6 commits in the last 90 days. Among its 4 catalogued features are Multi-Model Collaboration, Token-by-Token Generation, and Model Attribution.
Latest indexed changes and source events
Mitch's Multifarious Three-Headed Hydra Chatbot verified by the PulseGate indexer
Other apps tracked under the same category.
