Llama 3

1. I pitted different LLMs against each other in Pokemon Showdown (www.reddit.com)

   +1 5w llamagemini

   I wanted to see if LLMs could reason through complex game states, so I built a system where they can play Pokémon Showdown battles autonomously. They get the battle state every turn and use tool calls to attack or switch.

2. I built AI agents that play Pokemon Showdown autonomously using free LLM APIs via tool-calling (www.reddit.com)

   5w tool-callinggemmaqwen+1

   I've built a system where models like Llama 3, Qwen, and Gemma play Pokémon Showdown battles autonomously. Instead of simple prompt-response, they analyze the full battle state every turn (type matchups, HP, weather, field conditions, reve…

3. What STT/LLM/TTS combo are you running for production voice agents in 2026? (www.reddit.com)

   +21 5w llamaanthropicopenai

   Curious what stacks people are actually using right now, and where you're hitting walls. Some things I've been observing while testing combos: - Deepgram Nova-3 still the best STT for English, Cartesia is closing the gap on streaming - Ele…

← all threads