Here's another, although simpler.
441 meshes, 41k tris, 3D camera build, build with Three.js, one-shot. DeepSeek-V4.1-Flash did an amazing job for one-shot. It is still missing a few parts, some connections etc. Maybe with more iterations, via harness i can get it to be more
WeChat has open-sourced WeKnora, an enterprise-grade LLM-powered knowledge framework.
It combines RAG, ReAct Agents, MCP, Skills, sandboxes, and Auto-Wiki. Wiki Mode is the standout: It turns raw enterprise documents into a self-maintaining, interlinked Markdown Wiki +
DeepSeek v4.1 Flash vs GLM 5.3 Flash
I asked both models to recreate the Psychonauts 2 and Metroid Prime game menus as closely as possible to the originals. Attached are full-sized videos from both models for both games. All were done in a single shot, both at Max thinking,
DeepSeek V4.1 Flash is free.
And it just outscored Claude Opus 5 on software engineering. The numbers made me read twice: 74.2 on SWE-bench. Opus 5 scored 74. GPT 5.6 scored 73. 88.1 on CyberGym. Ahead of both. The honest part: it loses on raw knowledge. Opus scores 56.3
You have reached the end of the archive
All of deepseek