chat.md

It's live on openrouter now. In my personal benchmark it's bad. So far the benchmark has been a really good indicator of instruction following and agentic behaviour in general. To those who are curious, the benchmark is just the ability of model to follow a custom tool calling format. I ask it to using coding tasks using chat.md [1] + mcps. And so far it's just not able to follow it at all. [1] https://github.com/rusiaaman/chat.md

Did some quick tests. I believe its the same model as Quasar. It struggles with agentic loop [1]. You'd have to force it to do tool calls. Tool use ability feels ability better than gemini-2.5-pro-exp [2] which struggles with JSON schema understanding sometimes. Llama 4 has suprising agentic capabilities, better than both of them [3] but isn't as intelligent as the others. [1] https://github.com/rusiaaman/chat.md/blob/main/samples/4.1/t... [2] https://github.com/rusiaaman/chat.md/blob/main/samples/gemin... [3] https://github.com/rusiaaman/chat.md/blob/main/samples/llama...

Hi HN, I was fed up with chat clients either not supporting MCP for certain models or not being hackable enough. I envisioned a markdown file that is a serial format of the chat conversation in which LLM directly streams its response. I "vibe coded" this vscode extension and found it to be useful. It may have some rough edges but works well in my testing so far. I'm excited to share the result of this exercise. - Edit past messages of user, assistant or tool responses and have the AI continue from any point. The file editor is the chat interface and the history. - LLM agnostic MCP support: no restrictions on tool calling on any LLM, even if they don't official support tool calling. - Press shift+enter to have AI stream its response in the chat.md file which is also the conversation history. - Tool calls are detected and tool execution res

Gemini 2.5 pro hasn't been as good as Sonnet for me. The prompt I have tried repeatedly is creating a react-vite-todo app. It doesn't figure out tailwind related issues. Real chats: Gemini: https://github.com/rusiaaman/chat.md/blob/main/samples/vite-... Sonnet 3.7: https://github.com/rusiaaman/chat.md/blob/main/samples/vite-... Exact same settings, using MCP server for tool calling, using OpenAI api interface. PS: the formatting is off, but '#%%' starts a new block, view it in raw.