Local Qwen 27B agent logs into Amazon and buys paper autonomously in one run
fuzhongkai · reddit · 2026-09-22
Reddit user fuzhongkai gave a local Qwen 3.8 27B agent, running on the open-source TensorSharp inference engine, access to his Amazon account with one instruction: log in, find the cheapest/discounted A4 paper, and buy it. The agent completed the whole task in a single 24-minute run, with the human only stepping in to log in and approve the final purchase.
Stack:
- Qwen 3.8 27B, fully local inference
- TensorSharp inference engine + agentic runtime
- Playwright skill for browser automation
Searching, comparing, navigating Amazon, and preparing the order were all autonomous; apart from accessing Amazon itself, inference, agent state, and tool execution stayed local. The video includes the full raw reasoning/tool trace (sped up 8x), and the project is open source on GitHub. The author notes the iPhone TensorAgent version still lags due to local inference speed and iOS browser restrictions.
More from coding & agent
- ZenML puts TypeSafe's Jev in production as an evidence reviewer for its SRE agent — strickvl · 2026-09-22
- Developer builds CitizenAgent: giving AI agents their own email, phone number and crypto wallet — jomic01 · 2026-09-22
- Seven guards that made our agents boring enough to trust — arthaudm · 2026-09-22
- AI memory system built with Jev claims 10x speed, 6x lower cost — realferrari · 2026-09-22
- Claude usage is shifting from 'write this' to context-heavy workflows — Thick-Session7153 · 2026-09-22
- riderless: A Zero-Token Decision API on Gemma 4 Hits 93.1% Accuracy on One RTX 5090 — Pale-Soil-2524 · 2026-09-22