Show HN: TXT OS – Open-Source AI Reasoning, One Plain-Text File at a Time (github.com)
I'm excited to share TXT OS — an open-source AI reasoning engine that runs entirely inside a single `.txt` file.
- No installs, no signup, no hidden code — just copy-paste the file into any LLM chat window (GPT, Claude, Gemini, etc.). - +22.4% semantic accuracy, +42.1% reasoning success, and 3.6× more stability (benchmarked on GSM8K and Truthful-QA). - Features Semantic Tree Memory, Hallucination Shield, and fully exportable logic. - MIT Licensed, zero tracking, zero ads.
Why did I build this? I wanted to prove that advanced reasoning and memory could be made open, portable, and accessible to anyone — just with pure text, no software or setup.
A note: I'm from China, and English is not my first language. This post and the docs were partly assisted by AI, but I personally reviewed and approved every line of content. All ideas, design, and code are my own work. If anything is unclear or could be improved, I really welcome your feedback!
I'm the author, and happy to answer any questions or suggestions here!
3 comments
[ 3.2 ms ] story [ 17.8 ms ] thread1. How does TXT OS store its “Semantic Tree Memory” between sessions? 2. When `kbtest` detects a hallucination, what happens next? 3. Any idea of the speed impact on smaller models like LLaMA-2-13B?
Thanks for sharing—excited to try it out!
does TXT OS work equally well with open-source models, or is it optimized more for models like GPT-4 or Claude?