5.3 KiB
WFGY OS — TXT Build Beta
繁體中文版介紹預計一個小時內更新,連結會在這邊
The first plain-text operating scaffold—fork it, shape it, call it yours.
This README is continuously updated
Last update: 2025-07-02 (Beta launch)
Please check back for newer docs and links.
📥 Download
🌍 Download WFGY OS · HelloWorld.txt (v1.0)
Hosted on Zenodo — an open-access platform supported by CERN.
This release is cryptographically verified, secure, and free of any external code or malware.
The DOI never changes; every new release appears at the same URL.
⚡ Quick Start (⏱ under 1 minute)
1. Download HelloWorld.txt from the link above
2. Upload / paste it into any LLM chat
3. Type hello world (or wfgy console)
→ pick a language → OS boots immediately
Bug hotline: Telegram
@PSBigBig• Tested: ChatGPT (o 3 / o 4o), Claude-3 Opus, Phi-3-mini Untested ≠ unsupported — open a GitHub Discussion for issues & ideas.
🚀 Road to 10 K Stars
Current engine WFGY 1.0 Semantic accuracy ↑ 22.4 % | Reasoning success ↑ 42.1 % | Stability ↑ 3.6 ×
Stretch goal → 10 000★ before 2025-08-01 unlocks WFGY 2.0 (GPT-4-Turbo simulation) for everyone—free. New stats: accuracy ↑ 36.7 % | success ↑ 65.4 % | stability ↑ 5.1 × All stargazers will be credited in the changelog.
🔑 Key Points
| Feature | Why it matters |
|---|---|
| Plain-text only | No executables, no network calls, zero malware risk |
| Semantic Tree memory | Records reasoning nodes—not chat logs |
| ΔS + BBCR guard | Detects semantic turbulence; self-corrects before hallucinating |
| Four core modules | BBMC BBPF BBCR BBAM govern residue, progression, correction, attention |
| MIT-licensed & forkable | Copy the file, edit the language, publish your edition |
🗺️ Roadmap
| Date | Milestone |
|---|---|
| 2025-07-02 | Beta — DOI released on Zenodo |
| 2025-07-07 | v 1.0 — cross-platform tweaks & packaged TXT apps |
TXT apps are plain text; “app” is only a friendly label.
🤝 Contributing & App Hub
- Fork this repo, create your own
.txtOS or app. - Upload finished apps to the WFGY Zenodo community (link arrives with v 1.0).
- Submissions pass an automated check (license · ASCII-only · safety).
- Curated entries will appear under
/apps.
📂 Repository Layout
/OS core TXT builds & changelogs
/apps community TXT apps (opens 2025-07-07)
/docs white paper & diagrams
Project home → https://github.com/onestardao/WFGY Direct OS → https://github.com/onestardao/WFGY/tree/main/OS
No auto-update — always grab the newest TXT manually.
⚖️ License
MIT License — © 2025 The WFGY Project
🕹️ Hidden Tip
Type logo inside the console to view the ASCII logo.
❓ FAQ (11 quick answers)
Click to expand
1 · How does WFGY give AI memory?
Semantic jumps (high ΔS) create nodes in a Semantic Tree—topic, module, tension—giving a recoverable reasoning path.
2 · What is ΔS and how does it stop hallucination?
ΔS measures semantic tension; when too high, BBCR reroutes logic or asks for confirmation, preventing confident nonsense.
3 · How can a single TXT file do all this?
Logic, boundaries, and memory rules live in natural language. The model reads and obeys—no code runs.
4 · Why call it an OS, not a prompt?
It manages memory, logic, and boundaries—like an operating system manages processes. Reboot, patch, extend via text.
5 · What do the four modules do?
BBMC minimise residue · BBPF progress paths · BBCR correct collapse · BBAM modulate attention & tone.
6 · Semantic Tree vs standard memory—can it recover forgotten info?
Standard memory stores snippets; the Tree stores logical context, so reasoning can be rebuilt after token drop.
7 · How does the BBMC formula help reasoning?
B = I - G + m*c^2 quantifies deviation from ground truth, letting the model self-correct across turns.
8 · How do I verify WFGY isn’t fake?
Paste the TXT into any LLM, run kbtest, or ask how memory works—answers use the embedded logic.
9 · Can WFGY integrate with agents or workflows?
Yes—load the TXT as the reasoning core, then layer external tools or APIs.
10 · Commercial use?
MIT—free for commercial or personal projects; keep copyright & disclaimer.
11 · How do I fork or customise WFGY?
Copy HelloWorld.txt, edit the rules, rename, publish. The AI follows any coherent structure.