Dev builds open-source local coding harness on Qwen 27B to replace OpenCode
xornullvoid · reddit · 2026-08-27
A developer shares warpdrv, an open-source (AGPL) harness for running local LLMs, built over 90% locally under their supervision with Qwen 3.x 27B — free, no telemetry, Windows/Linux. Its Server Manager also works with Claude Code/OpenCode.
Key features: Just-in-Time code review guardrails before tool calls; three-way chat where agents and user both talk in sub-agent threads, with nestable conversations; voice dictation annotations; built-in MCP tools (filesystem, web fetch, code graph, to-dos); all AI messages reviewed by a second adversarial AI; voice-chat with TTS; work modes for planning/building/researching/reviewing; custom-compile llama.cpp from any branch (CUDA/ROCm/Vulkan). Hardware: Strix Halo 128GB + RTX Pro 5000 48GB.
Lessons learned: one chat per feature/bug keeps quality high; in large codebases, a script-generated file tree in CLAUDE.md beats verbose feature descriptions; early decisions matter more as the codebase grows — serious apps need senior-dev guidance, unattended vibe-coding only works for small throwaway apps; routinely reject file-read requests for files you know are irrelevant to avoid polluting context.
Related event: Developer Builds Open-Source Coding Harness warpdrv on Local Qwen 27B(2 posts)→
More from coding & agent
- Dev bills $20,000 for an app built with $200 of Claude Code — IgorCarron · 2026-10-15
- Dan Shipper: run product and labs in parallel, even with a labs team of one — danshipper · 2026-10-15
- Gemini 4 Argon wrote 5,000+ lines of pure Python to build its own ray-tracing 3D engine — aitrendz_xyz · 2026-10-15
- Minecraft fully ported from Java to Rust, file by file, server and client included — Vjeux · 2026-10-15
- mitsuhiko builds a full Python IDE for iOS in an afternoon with new AI models — mitsuhiko · 2026-10-15
- How to get Gemini 4 Argon to build games that actually look decent — chongdashu · 2026-10-15