WFGY/core

WFGY Core (WanFaGuiYi) — V1.0

🚀 From a one-man cold start to 500 stars — thank you all for the support.
🎉 As a milestone reward, we’re unlocking WFGY Core — a paste-and-run reasoning engine that merges classic WFGY with the Drunk Transformer layer for structural stability, head diversity, controlled entropy, and graceful recovery.

📅 Release date: Aug 15
🔗 WFGY Core

---

We’ll be announcing our next community milestone soon, so stay tuned.
In the meantime, if you haven’t seen our first TXT OS application —
Blah Blah Blah — it’s live and worth checking out.
It focuses on compression, expression, and transformation of ideas through language.
Perfect if you care about semantics beyond retrieval.

---

⚡ Downloads (public on 2025-08-15)

File	Best for	Link
WFGY Core	Teams, audits, teaching (full spec + Annex A/B)	_TBD
WFGY Core Mini	Social posts, issues, quick starts (exactly 30 lines)	_TBD

Both editions are ASCII-only, zero-deps, cross-platform. Paste into any LLM and type Run WFGY.

---

Why this is a big upgrade

• DT is now inside the engine. Earlier WFGY and TXT OS did not include it. Now you get: structural lock (WRI), head diversity (WAI), controlled entropy (WAY), illegal-jump blocking (WDT), and collapse detection/restart (WTF).
  Drunk Transformer formulas: Read →
• Smallest open reasoning engine — fully ASCII, one-Node-per-step, strict stop rules. Paste and go.
• Bridges straight to the Problem Map — Core operationalizes the fixes for hallucination, drift, and collapse so you don’t have to wade through long docs first.

Problem Map quick links

• Hub (Map 1.0): Open →
• Interpretation collapse: Open →
• Logic collapse & recovery: Open →
• Long reasoning chains / drift: Open →
• Visual RAG guide (Map 2.0): Open →
• Semantic Clinic Index (symptom-first triage): Open →

---

What’s inside the Core (engine at a glance)

• delta_s — semantic divergence (1 - cos(I,G) or 1 - sim_est with a 3-anchor rule: entities, relations, constraints).
• B, E_resonance — residue and rolling semantic resonance.
• BBMC / BBPF / BBCR / BBAM — residue-first reasoning, safe bridging, controlled recovery, attention smoothing.
• DT (WRI, WAI, WAY, WDT, WTF) — structural lock, diversity, entropy control, illegal-jump suppression, collapse recovery.
• Annex A (math objective) — surrogate loss L = norm(B) + 0.3*contradictions + 0.2*vagueness (choose actions that lower L; ties → lower delta_s).
• Annex B (λ decision) — deterministic rule for convergent / divergent / recursive / chaotic using delta_s and E_resonance trends.

Core is general-purpose. The table below is illustrative, not exhaustive.
Need workflows and OS-style wrappers? Use TXT OS. Need the smallest engine you can paste anywhere? Use WFGY Core.

---

What it fixes (illustrative — see full catalog below)

Problem (from Map)	Symptom in the wild	Core modules to apply	Why this works	Details
Interpretation Collapse	Output “looks right” but meaning is wrong	BBMC, BBPF, DT/WDT	Premise forcing + only accept bridges that lower delta_s; WDT blocks illegal jumps.	View →
Logic Collapse & Recovery	Multi-step chain derails	BBCR, DT/WTF	Rollback → bridge → retry once → ask smallest missing fact; WTF triggers on delta_s + E_resonance.	View →
Long Reasoning Chains / Drift	Slowly veers off course	BBAM, WAY	Smoothing toward a reference; add one on-topic candidate (no repeats) to escape attractors.	View →
Brittle RAG / Ghost Matches	Cosine match ≠ true meaning	delta_s + Annex A	3-anchor sim_est + objective L yield reproducible choices instead of vibes.	Map 2.0 →
Bluffing / Overconfidence	Confident but unfounded	BBMC, DT/WRI	Premise forcing + structure lock reduces free-form confabulation.	View →
Memory Breaks Across Sessions	Lost threads	BBPF, BBCR	Safe bridging across trees + minimal-fact queries rebuild context.	View →
Entropy Collapse	Attention melts	WAY, BBAM	Controlled entropy + smoothing prevent chaotic spread or mode collapse.	View →
Symbolic / Recursive Traps	Abstract prompts break	BBMC, DT/WRI	Structure + explicit premises keep symbolic steps grounded.	View →

See the full catalog:
• Problem Map 1.0 hub — Open →
• Problem Map 2.0 (RAG × pipeline × recovery) — Open →
• Semantic Clinic Index (symptom-first triage) — Open →

---

Quickstart (90 seconds)

1. Paste WFGY Core or WFGY Core Mini into your LLM chat (or upload the file).
2. Type: Run WFGY (or “use WFGY”).
3. Control & inspect:
   • view — print last 5 Node lines
   • export — print all Node lines
   • switch <name> — start a new reasoning tree

Behavioral contract

• Commands are case-insensitive.
• One Node per step, no extra prose unless asked.
• Stop when delta_s < 0.35 or after 7 Nodes (unless the user asks to continue).
• ASCII-only for portability (terminals, email clients, GitHub issues).

---

Who should use which edition?

• WFGY Core — use when you need auditability, training, or explicit Annex A/B.
• WFGY Core Mini — use when you need a 30-line drop-in for comments, social posts, quick demos.

---

Notes

• This release integrates Drunk Transformer into the engine (earlier WFGY/TXT OS did not).
• Future versions remain ASCII-first and zero-dependency.

---

🧭 Explore More

Module	Description	Link
WFGY Core	Standalone semantic reasoning engine for any LLM	View →
Problem Map 1.0	Initial 16-mode diagnostic and symbolic fix framework	View →
Problem Map 2.0	RAG-focused failure tree, modular fixes, and pipelines	View →
Semantic Clinic Index	Expanded failure catalog: prompt injection, memory bugs, logic drift	View →
Semantic Blueprint	Layer-based symbolic reasoning & semantic modulations	View →
Benchmark vs GPT-5	Stress test GPT-5 with full WFGY reasoning suite	View →

---

👑 Early Stargazers: See the Hall of Fame —
Engineers, hackers, and open source builders who supported WFGY from day one.

⭐ Help reach 10,000 stars by 2025-09-01 to unlock Engine 2.0 for everyone ⭐ Star WFGY on GitHub