Engineer tests offline coding agent with Qwen3.5 9B in Himalayan workspace
A Principal Engineer at Thomson Reuters tested running an offline coding agent on their laptop in the Himalayas, where network outages are common. Using OpenCode with the open-weight model Qwen3.5 9B (hosted via Ollama), they successfully fixed a bug in a Python script without internet access. The agent read local files, proposed edits, and ran tests—all offline—though dependencies and model…
Key points
- Thomson Reuters engineer ran **OpenCode** with **Qwen3.5 9B** offline on a Mac, fixing a Python bug without internet
- Setup required pre-downloading the model (6.6 GB) and project dependencies; memory usage peaked at 12–16 GB
- Agent proposed edits, ran tests, and passed all five unit tests—verifying offline capability for small tasks
The setup required downloading the model and project dependencies ahead of time. The engineer noted memory constraints (12–16 GB for Qwen3.5 9B) and emphasized focusing on small, test-driven tasks. While the test passed, the engineer acknowledged limitations: larger projects or unfamiliar frameworks may not perform as well. The walkthrough includes two additional exercises—a CSV report and log analyzer—for further offline testing.
Coding From the Himalayas With an Offline AI Agent
Latent Space · 27 October 2026
Loading the full article…
This text was published by Latent Space and written by Rittika Jindal. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Author intentionally breaks browser agent and finds debugging harder than detecting flaws · 1 src
- Bito releases Governor to cut coding-agent costs by up to 70% · 1 src
- Wired tests six AI agents, ranks Gemini Spark top for news briefings · 1 src
- OpenAI’s dots score 8.8 in hands-on test, rank third among five agents · 1 src
- Developer builds voice-driven AI agent with ledger, multiple PCs and Claude Code · 1 src
Comments
via GitHub Discussions