Show HN: Brontosaurus, a voice-driven generative AI canvas (thomasdhughes.com)

Brontosaurus, a newly unveiled web-based generative AI canvas, allows users to create digital widgets simply by speaking their requests. Utilizing OpenAI's gpt-oss-120b model, hosted on Cerebras' infrastructure, Brontosaurus offers impressive performance, generating content at 3,000 tokens per second—far surpassing the speed of typical AI chat applications. The innovative design emphasizes human-AI collaboration, enabling a seamless back-and-forth experience where users can easily modify or create widgets in real-time. This platform is significant for the AI/ML community as it embodies a shift toward more interactive and collaborative AI systems, moving away from traditional agent autonomy models. Technical elements include multi-agent orchestration, with two agents—the Conductor and Builder—coordinating tasks like creating, editing, and arranging widgets based on user prompts. Although still in development, Brontosaurus aims to enhance its capabilities with a virtual file system for persistent storage and potentially explore more powerful models. The founder actively encourages community feedback and ideas, reflecting a growing trend of open collaboration in AI development.