iCreatAI News
Skip to content
Some English translations are not ready yet. The site is still in English. Untranslated news and event text is shown in its original language until background processing completes.
Trending storyUpdating

本地AI智能体无网络逆向串口设备

1 reports1 sources42 min agoUpdated

Understand this story

AI overview

作者bluepeter将一台未知设备接在树莓派串口上,让两个本地运行的LLM智能体在无互联网、无文档、无人工协助的条件下自行查明设备身份、逆向协议并在其上运行程序。据其实验记录,Qwen3.8-Flash-Next在约5.5小时内识别出该设备是运行TUTOR 1.3的Motorola 68000单板训练机,并完成内存映射、命令集解码和自写程序运行。 复现该实验需要OpenCode CLI、两个兼容OpenAI接口的模型服务器(如llama.cpp的llama-server),以及一台装有systemd-networkd的Linux机器才能启用隔离环境;模型仅使用本地权重,端点需自备。隔离环境通过install-jail.sh脚本安装,属可选项。上述识别与运行结果来自作者自述。

AI overview based on source reports · Updated 16 min ago

Coverage timeline

Follow the coverage to explore different perspectives.

Nov 22
  1. Hacker News AI
    Local AI agents investigate an unknown machine attached to a serial port

    An experiment wired an anonymous device to a Raspberry Pi's serial port and had two local LLM agents, with no internet access or documentation, identify it and reverse-engineer its protocol. In about 5.5 hours, Qwen3.8-Flash-Next discovered the device was a Motorola 68000 single-board trainer running "TUTOR 1.3," mapped its memory via deliberate bus errors, decoded its monitor command set, and ran its own program on the hardware, with a second independent LLM acting as an adversarial critic that rejected unsupported claims. Three full runs with verdicts, notes and raw logs are published under an MIT license, along with the harness and instructions to replay it.

Activity for this story

Not enough continuous observations to show a trend yet.