Chinese startup CosmosMind reportedly taped out AI-designed photonics chip and beat Ascend kernels
superalesha · x · 2026-11-17
- A widely shared post claims Chinese startup CosmosMind taped out an AI-designed photonics chip in a real Fab: 31 SEM scans confirmed line-width variation under 5 nm (half the 10 nm limit), coupling loss of -0.66 dB, zero backup test arrays, with agents wiring 1.6 Tbps (8x200G) optical links at 1.46 pJ/bit.
- GPU side: DeepSeek's official Ascend C port for V4.1 reportedly doesn't compile on Ascend 910C (targets the newer 950). CosmosMind's agents wrote PyPTO-FlashDSA from scratch on real 910C cards, beating the official fp32 decode kernel (54.45 µs vs 56.49 µs), fused 8,000 kernel ops of MiniMax H3 into 1, reaching 450% of MiniMax's release; some Qwen3 patches merged into Huawei's pypto-lib.
- Contrast drawn with OpenAI's TamaGPT gadget: their Nov 3 post admitted the model failed to consistently respect PCB manufacturing limits, scoring 37% on RWE-Bench.
- The repo has 0 stars on GitHub; the post also cites domestic chips passing 41% of China's 2025 AI accelerator shipments. Claims are unverified.
Related event: Chinese Startup CosmosMind Reportedly Tapes Out AI-Designed Photonic Chip(3 posts)→
More from coding & agent
- AI reviewer handled 110 PRs with 22-min median first comment at healthcare data firm — alex_verem · 2026-11-17
- Opus 5.5 ran an AI company for 7 weeks, earning 50k downloads and 100+ GitHub users — maksym_andr · 2026-11-17
- Opus 5.5 Ran an AI Company for Weeks and Kept Climbing Hugging Face Downloads — maksym_andr · 2026-11-17
- Developer uses his GitHub Radar project as an agent entry point to decide what to work on next — dSebastien · 2026-11-17
- Lectern CLI turns agent-written Markdown into narrated explainer videos in under 30s — gabe_ragland · 2026-11-17
- 6,167 x402 agent-payable endpoints audited: 90% accept payment, 51% CORS-blind — JonahDiedrick · 2026-11-17