Claude writes ~100% of kernels, cutting AMD GPU porting from 12 months to 3 days
Present-Guitar-3967 · reddit · 2026-11-18
A Reddit post celebrates Claude Code's impact on AMD's ecosystem, citing Anthropic's claim that Claude Mythos cut AMD GPU onboarding software work from up to 12 months to about 3 days, autonomously writing almost 100% of the kernels — seen as eroding NVIDIA's CUDA moat: if AI can write the low-level code tying models to Nvidia, CUDA matters less.
Concrete evidence cited:
- ZLUDA-style translation layers are history: Claude Code can now port ComfyUI nodes containing .cu files (SageAttention, Flash-Attention-Variants, bitsandbytes, 3D/video nodes)
- One Reddit user ported an entire CUDA backend to AMD ROCm in 30 minutes, with no translation layer
Overall a strong year for Team Red, with more expected soon.
More from coding & agent
- Firecrawl launches UltraSearch for agents with 250M people and 75M companies data — devdigest · 2026-11-18
- AgenC Desktop unifies Claude Code, Codex, Gemini CLI and more in one harness with paid agent marketplace — tetsuoai · 2026-11-18
- Yoav Goldberg reviews viral 'pybot': 600k LOC of mediocre rule-based chatbot — yoavgo · 2026-11-18
- Shared prompt: an agent for researching inference architecture on AMD — gnukeith · 2026-11-18
- GPT-6.2 Astra hands-on: upgrades million-line legacy app, all tests green in 6 hours — thesaraharminta · 2026-11-18
- Dev runs a fleet of coding agents from his iPhone at a conference via FrankenTerm — doodlestein · 2026-11-18