Local AI Coding Agents Are Finally Good Enough
Video Overview & Insights
Book (From Idea to App Store with Claude Code): https://cwb.sh/book?r=yt
I ran `qwen3-coder:30b` on a 9 hour flight without wifi recently, and eh, it's decent if you're babysitting it very closely and analyzing it's decisions, but couldn't trust it anywhere near as I do the flagship OpenAI/Anthropic models. Ended up not using any of that work 😅 (+ the speed! I don't have the attention span to wait long turns... I need to be working on parallel sessions which doesn't work well with my machine)
Local LLMs finally don't suck, I ran Qwen3.6 27B on my Mac with LM Studio, MLX, and opencode to build real app features 100% offline, no subscription required.
Resources mentioned
how's the latency on qwen 27b with mlx? i've been wanting to try a local setup but worried about the token speed on my mac
• Book — From Idea to App Store with Claude Code: https://cwb.sh/book?r=yt
• Aqua Voice (1 free month) — https://cwb.sh/aqua
wooow, you just made my day ... if this really work ... this is the future
My Products
• Pro Access — get access to everything: https://cwb.sh/pro?r=yt
• React Native Course: https://cwb.sh/rn?r=yt
• Platano — AI app template: https://cwb.sh/platano?r=yt
• Book — From Idea to App Store with Claude Code: https://cwb.sh/book?r=yt
• Inkigo — AI Tattoo (iOS): https://cwb.sh/inkigo?r=yt
• Quick Push — test push notifications (macOS, $4.99): https://apple.co/4tvT4wF
Free Resources
• Newsletter: https://cwb.sh/newsletter?r=yt
• Discord: https://cwb.sh/discord?r=yt
Support My Work
• Channel Membership: https://cwb.sh/join
• GitHub Sponsors: https://github.com/sponsors/betomoedano
Socials
Follow Beto on X: https://cwb.sh/x?r=yt
LinkedIn: https://cwb.sh/in?r=yt
GitHub: https://cwb.sh/gh?r=yt
Instagram: https://cwb.sh/insta?r=yt
TikTok: https://cwb.sh/tiktok?r=yt
Threads: https://cwb.sh/threads?r=yt
Bluesky: https://cwb.sh/bsky?r=yt
Facebook: https://cwb.sh/fb?r=yt
Timestamps
00:00 Intro
00:55 Qwen3.6 Benchmarks vs Claude
02:02 Hardware Requirements (Can You Run It?)
03:13 Download & Set Up LM Studio
04:04 Reasoning, Vision & Tool Calling Demo
04:30 Developer Tab: Logs, Concurrency & Cache
05:14 Turn It Into a Coding Agent with opencode
05:59 Real Project Demo: My AI Tattoo App
08:17 Building Minesweeper & Snake Offline
10:48 Stress Test: 31 Tool Calls on a Real Codebase
13:43 Chat From Your Phone (LM Link + Locally AI)
14:56 Final Thoughts
#localllm #qwen #lmstudio