TTS API and MCP: how to add speech synthesis to your project
TTS and TTS HD: capabilities, the /v1/audio/speech endpoint, MCP tools and getting a key.
Vibe coding prompts: a request structure that produces working code
A vibe coding prompt is a short spec, not a plea to 'build me a site'. We break down a six-block structure, show a weak and a strong request on the same example, and give ready-made phrasings for follow-up edits — for the NeuralSpace vibe coding section.
Code Mode: how AI agents call tools through generated code
A Code Mode walkthrough: the agent writes JavaScript, the sandbox orchestrates tool calls, and the model sees only the final summary — fewer tokens, lower latency.
MiniCPM-V 4.7: Open 35B MoE Multimodal Model Runs on Phones with 3 GB RAM
OpenBMB has launched MiniCPM-V 4.7 under Apache 2.0. The 35B-A3B Mixture-of-Experts model activates 3B parameters per token and decodes at over 13 tokens/sec on smartphones within 3 GB of memory. Inside the architecture, visual compression, and benchmark numbers.
Gemini 4 Argon is now live on the site
Gemini 4 Argon is already in the chat, the Code section and the public API: 1M token context, understands images, audio and video, reasons and searches the web. How to try it.