Vibe coding with Chinese models: role-splitting Kimi, GLM and DeepSeek in one workflow
rrtsehc · reddit · 2026-10-18
A developer shares a full workflow for role-splitting Chinese models in vibe coding:
- Kimi K3: UI design, architecture planning, algorithms. Slow and pricey but strong design sense and depth on complex logic; Kimi writes a detailed spec, faster models build from it, with GLM 5.3 reviewing critical logic
- GLM 5.3 Flash: actual code writing, best balance of intelligence, price-performance, long context and parallel tasks. The main plan is broken into an ordered task pool with background sub-agents building in parallel, using file locks, draft staging and an agent communication log; an audit sub-agent runs every 10 tasks
- DeepSeek V4.1 Flash, Hy4 preview, Space Bunny: feature ideas and code audits. Different models catch different bugs, so repeated cross-checks are run since they're free on WorkBuddy
Qwen 3.8 MAX/Flash haven't been tested yet.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23