QwenBend: Qwen writes and runs Bend programs inside its own forward pass using 53M hand-set weights
sull · x · 2026-11-02
@rifeash unveils QwenBend, where a Qwen model writes a Bend program, type-checks it against Bend2 rules using its own weights, and executes it within its forward pass.
No interpreter, compiler, or tool calls are involved—similar to the earlier qwenputer concept. The entire "CPU" consists of 53M hand-set non-zero weights inside the model, with all computation happening within the network itself.
Related event: QwenBend Runs Self-Written Bend Programs Inside Model Forward Pass(2 posts)→
More from Fun
- Developer ships v0.6.83 entirely via autonomous agents, zero human input — jasonkneen · 2026-11-02
- Lighthaven venue pricing sparks debate: $50-$125 per person, one evening event could hit $75K — mimi10v3 · 2026-11-02
- Open-source pixel-plays: browser games as isometric pixel dioramas, one JS file, zero dependencies — algo_diver · 2026-11-02
- Claude and ChatGPT both 'predict' OpenAI will fire Noam Shazeer before 2031 — ethanCaballero · 2026-11-02
- NYC took decades to build 2.9km of subway; Su mocks it against Guangzhou Metro — suchenzang · 2026-11-02
- Fully connected five-ellipse Venn diagrams weren't found until 1984, with help from GPT — MikePFrank · 2026-11-02