Local 27B LLM agent autonomously buys A4 paper on Amazon in a single run
fuzhongkai · reddit · 2026-09-23
A Redditor ran a fully local agent stack through a real-world browser task: buying discounted A4 paper on Amazon.
- Setup: Qwen3 27B running locally, TensorSharp for local inference + agent runtime, and a Playwright skill for browser interaction.
- Human input: only twice — logging into Amazon and approving the final purchase. The agent handled searching, comparing products, navigating, and generating/executing Playwright code in a single run.
- Notably local: apart from Amazon traffic, inference, agent state, code generation and tool execution all stayed on the local machine. The video is 8x sped-up; the real run took 24 minutes.
- Limits: TensorAgent on iPhone isn't ready — local inference speed and iOS browser restrictions remain blockers. Code is on the TensorSharp GitHub.
Related event: Local Qwen 27B Agent Completes Autonomous Amazon Purchase(2 posts)→
More from coding & agent
- Dell expands AI Data Platform: agents have a data-context problem, not a model problem — DavidLinthicum · 2026-11-12
- LLM controls vocal synth to perform Ave Maria in 4-part harmony — AAAzzam · 2026-11-12
- ArcGrav: a cross-model Agent Control Language to cut token overhead — ArchitecturalGravity · 2026-11-12
- Claude watches competitors around the clock and emails me what matters — GlenBradley · 2026-11-12
- Developers say Gemini 4 Carbon is notably better at favicon design and micro-interactions — AI_Andrew · 2026-11-12
- v0 now supports ChatGPT Business subscriptions alongside Plus and Pro — OpenAIDevs · 2026-11-12