Local Qwen 27B fixes commercial app bug on first try at ~2x Claude Opus speed
julianharris · x · 2026-09-20
Developer julianharris reports a milestone: a bug in his native Mac screencast app Cutaway—"Insert Cut" not respecting the selection—was fixed on the first attempt by his fully local AI stack.
The stack
- Qwen 3.8 27B (Unsloth quantization) served via llama.cpp / llama-server
- Nvidia 4090 24GB on Ubuntu
- Cline/agent tooling + MCP for efficiently narrowing down where the bug might be
- Throughput of 70–90 tok/s, roughly double Claude Opus's 40 tok/s
He open-sourced his setup via an "awesome-local-ai" repo so others can replicate the pipeline without manual configuration.
More from coding & agent
- px0 editor ships git status streaming via SSE, checking just 3 files instead of polling — arpit_bhayani · 2026-09-20
- DIY Jev-style classifier: shuffling options lifts accuracy from 47% to 73% — WelcomeMysterious122 · 2026-09-20
- Andrew Ng releases free 1-hour course on building agentic knowledge graphs from scratch — irinarish · 2026-09-20
- Why Jev might finally kill the text prompt: millisecond decisions for generative GUIs — dpopa · 2026-09-20
- NoSpoon agent churns out microdrama content fully autonomously in about 15 minutes, zero prompts — Kyrannio · 2026-09-20
- Laya-MLX: open typed decision model runs locally at 7–14 ms per decision, zero tokens — JiliJeanlouis · 2026-09-20