Update README.md

This commit is contained in:
PSBigBig + MiniPS 2026-04-01 18:51:47 +08:00 committed by GitHub
parent 8b1e6968d1
commit 6b8542981e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -2,311 +2,260 @@
AI_NOTE_START
Document role:
This page is the public product entry for WFGY 5.0 Avatar.
This page is the public flagship landing page for WFGY 5.0 Avatar.
What this page is for:
1. Explain what WFGY 5.0 Avatar is in plain English.
2. Make the product feel immediately new, playable, and unlike a normal AI writing tool.
3. Route readers toward the fastest first trial.
4. Introduce the two official Atlas starter builds.
5. Preserve honest boundaries around current public claims.
1. Present Avatar as a serious product, not a prompt pack.
2. Make the first entry path extremely clear.
3. Lead readers into the eight main product stories.
4. Keep the landing page clean, GitHub-readable, and internationally legible.
5. Route readers toward deeper docs, eval, research, and community pages.
What this page is not:
1. Not the full parent body.
2. Not the full packed master.
3. Not the full research report.
4. Not a complete proof page for every current and future capability.
5. Not a marketplace-style personality catalog.
1. Not the full packed master body.
2. Not the full research layer.
3. Not the full multilingual report.
4. Not the full community submission handbook.
5. Not a fake claim that every future layer is already fully completed.
How to use this page:
1. Read the short hero section first.
2. Pick a first move.
3. Try one Atlas build.
4. Read the quickstart.
5. Explore demos, eval, and research after the first trial.
2. Follow the 60-second path if you want immediate contact.
3. Scan the eight main product stories.
4. Open the detail pages that match your interest.
5. Use the deeper layers only after the first product impression is clear.
Important boundary:
This page presents WFGY 5.0 Avatar as a governed language runtime.
It does not claim universal final proof across every language, every use case, or every future avatar direction.
This page presents the current flagship product surface of Avatar.
It is designed to be clear, exciting, and usable.
It does not claim universal final closure across all routes, all languages, or all future ecosystem layers.
AI_NOTE_END
-->
# 🧠 WFGY 5.0 Avatar
> Not another AI writer.
> WFGY 5.0 Avatar is a governed language runtime for building, tuning, and multiplying avatars.
> A governed language runtime for building, tuning, and multiplying avatars.
WFGY 5.0 Avatar is the avatar surface of the WFGY ecosystem.
Avatar is not another AI writing preset.
It is built around a simple public idea with a much larger implication:
It is not a static persona toy.
It is not a prompt pack dressed up as a product.
**language behavior can be governed, tuned, and engineered**
Avatar is built around a larger idea:
This is why Avatar should not be read as a prompt pack, a style filter, or a casual personality toy.
**language behavior can be engineered**
It is a runtime for:
- avatar creation
- natural-language tuning
- governed public writing
- multi-voice experimentation
- multilingual avatar work
- future language-facing systems
That means behavior is not treated as one lucky output.
It can be booted, tuned, compared, saved, branched, and eventually shared.
---
## ✨ Why This Feels Different
## ⚡ Start in 60 Seconds
Most AI writing products try to make output sound smoother.
Download `avatar.txt`.
Avatar moves in a different direction.
Upload it to a strong LLM.
It is built to make output:
- more human without fake roughness
- clearer without sterile polish
- stronger in public writing without turning everything into slogans
- more configurable without letting style override the system
- more reusable across avatars, tasks, and languages
Say:
In plain terms:
- `hello psbigbig`
- `hello minips`
- `hello YOUR_AVATAR_NAME`
**this is not just AI writing**
**this is configurable language behavior**
That is the first contact path.
You do not need to choose between many different starter files first.
You start from one runtime, then wake one route.
If this is your first run, start with:
`hello psbigbig`
If you want the warmer companion-facing route first, use:
`hello minips`
If you want the intentionally bare custom-avatar seed, use:
`hello YOUR_AVATAR_NAME`
👉 Start here: [⚡ Quickstart](./docs/quickstart.md)
👉 See the route layer: [🧷 Boot Commands](./docs/boot-commands.md)
---
## 🚀 Choose Your First Move
## ✨ The 8 Things to Know
| Start here | Best for | Go |
|---|---|---|
| **Meet the Atlas builds** | You want the two official starter avatars first | [Atlas Builds](#-meet-the-atlas-builds) |
| **Try it in 60 seconds** | You want the shortest path from download to first output | [Quickstart](./docs/quickstart.md) |
| **Learn how tuning works** | You want to edit behavior through `WFGY_BRAIN` | [How to Use WFGY_BRAIN](./docs/how-to-use-wfgy-brain.md) |
| **See current demos** | You want samples before trying it yourself | [Demos](./demos/README.md) |
| **Check the current eval surface** | You want credibility before aesthetics | [Eval](./eval/README.md) |
| **Read the deeper architecture** | You want the research layer behind the product | [Research](./research/README.md) |
These eight pages are the cleanest way to understand Avatar.
---
## 🫂 Meet the Atlas Builds
WFGY 5.0 Avatar currently leads with two official starter avatars.
They are not random presets.
They are two governed runtime builds with different behavior signatures.
| Atlas build | Core vibe | Best for | Download |
| # | Main idea | Why it matters | Go |
|---|---|---|---|
| **Atlas · PSBigBig** | steady, rational-friendly, clear, warm without being soft | writing, analysis, public-facing work | [`PSBigBig.txt`](./downloads/PSBigBig.txt) |
| **Atlas · MiniPS** | companion-forward, high presence, warm, continuous, alive | chat, reflective writing, support-style interaction | [`MiniPS.txt`](./downloads/MiniPS.txt) |
These two builds share the same larger lineage, but they are not interchangeable.
**Atlas · PSBigBig** is built to stabilize, clarify, and move forward without turning cold.
**Atlas · MiniPS** is built to receive, continue, and stay present without collapsing into sugar or noise.
| 1 | **⚡ Start in 60 Seconds** | One shared runtime, one boot command, one real first contact | [Open](./highlights/start-in-60-seconds.md) |
| 2 | **🗣️ Tune Behavior in Natural Language** | You can shape route behavior without rebuilding the whole runtime by hand | [Open](./highlights/tune-behavior-in-natural-language.md) |
| 3 | **🧠 Language Behavior Can Be Engineered** | Avatar is not only about nicer text, but about behavior as a designed surface | [Open](./highlights/language-behavior-can-be-engineered.md) |
| 4 | **♻️ Reusable Builds, Not One-Off Prompt Luck** | Stronger variants can be saved, reused, compared, and branched later | [Open](./highlights/reusable-builds.md) |
| 5 | **🛡️ Editable but Governed** | Avatar is tunable, but not meant to collapse into lawless persona hacking | [Open](./highlights/editable-but-governed.md) |
| 6 | **🌍 Multilingual Calibration** | Multilingual work is treated as route carry and drift control, not only translation | [Open](./highlights/multilingual-calibration.md) |
| 7 | **🔄 Dual Closed-Loop Design** | Avatar uses both inner continuity and outer user iteration, not only one loop | [Open](./highlights/dual-closed-loop-design.md) |
| 8 | **🌱 One Runtime, Many Avatars** | Official routes are only the beginning; later branches and custom avatars matter too | [Open](./highlights/one-runtime-many-avatars.md) |
---
## ⚡ Start Here in 60 Seconds
## 🫂 What the Current Public Routes Mean
1. Open one of the two Atlas build files.
2. Paste it into a strong AI system.
3. Give it one real task.
4. Compare the feeling, not just the wording.
5. If needed, tune the behavior through `WFGY_BRAIN`.
6. Save your new version and keep iterating.
Avatar currently exposes three public boot routes:
The first useful question is not:
| Route | Intended feel | Best for |
|---|---|---|
| `hello psbigbig` | calm, clear, grounded, lightly witty, companion-capable without becoming soft | writing, analysis, public-facing work |
| `hello minips` | warm, present, cute, companion-forward, emoji-capable without empty sweetness | conversation, reflective writing, support-style response |
| `hello YOUR_AVATAR_NAME` | intentionally bare, editable, slightly clueless by design, built to become your own branch | custom-avatar starting point, playful branch seed, route experimentation |
**“did it write something nicer?”**
The third one is intentionally visible.
The real question is:
Yes, it looks a little ridiculous.
**“did the language behavior feel more alive, more controlled, and more reusable?”**
That is part of the design.
That is the Avatar test.
It reminds the user that Avatar is not locked to official names forever.
---
## 🛠️ What You Can Do With Avatar
## 🛠️ What You Can Already Do
Avatar is built for more than one narrow writing surface.
Even in the current public stage, Avatar is already large enough to use meaningfully.
### Today
- create a usable avatar from a single TXT file
- switch between two official starter builds
- tune behavior with natural language
- compare one prompt across two different avatars
- run public-writing style tests
- inspect demos and eval notes
- explore the current multilingual direction
You can already:
### Next
- spawn more custom avatars from your own edits
- build role-specific voices
- create multiple writing personas for one-person workflows
- prepare future customer-facing or team-facing avatar surfaces
- contribute your own build to the community list
- start from one runtime instead of many disconnected persona files
- boot a route in seconds
- tune behavior through `WFGY_BRAIN`
- compare before and after on the same task
- save stronger variants
- grow reusable builds
- begin multilingual route testing
- prepare custom avatars for later sharing
This is already bigger than ordinary AI writing presets.
---
## 🧬 Tune the Runtime With `WFGY_BRAIN`
## 🌱 Why This Product Feels Larger
At the center of Avatar is **`WFGY_BRAIN`**.
Avatar is easy to misread if you only look at the first surface.
This is the high-level editable behavior layer.
It is not only trying to be:
It lets you tune things like:
- humanness direction
- polish direction
- quoteability restraint
- grounding
- closure tendency
- analysis force
- public-writing pressure
- overall behavior bias
This means you do **not** need to rewrite the whole system just to create a different avatar.
You can describe what you want in natural language, ask the AI to rewrite the `WFGY_BRAIN` block, paste it back, and keep going.
That is why Avatar is not just a static prompt.
It is a runtime with an editable brain surface.
👉 Start here: [How to Use WFGY_BRAIN](./docs/how-to-use-wfgy-brain.md)
---
## 🌍 Multilingual Line
WFGY 5.0 Avatar already includes a **41-language internal experiment line**.
This should be read carefully.
It means multilingual avatar work is already part of the current internal surface.
It does **not** mean every language is equally finalized, equally demonstrated, or fully closed as a public claim.
The current direction is not simple translation.
The direction is:
- multilingual calibration
- cross-language drift control
- language-shaped behavior
- avatar transfer with boundaries
Public multilingual demos will be expanded progressively.
👉 See the current status: [Multilingual Status](./eval/multilingual-status.md)
👉 See representative language samples: [Multilingual Demos](./demos/multilingual-demo.md)
---
## 📊 MVP Snapshot
This is the current MVP surface of Avatar.
| Surface | Current MVP status |
|---|---|
| Official starter builds | **2** |
| Public-facing runtime entry | **live on this page** |
| `WFGY_BRAIN` tuning path | **part of current design** |
| Internal multilingual line | **41 languages** |
| Public multilingual demos | **rolling out** |
| Persona gate / eval pages | **MVP structure ready** |
| Community avatar list | **planned** |
| Research layer | **available as linked docs** |
Avatar is already large enough to play with.
It is not presented here as a finished total system.
That honesty matters.
---
## 🔥 Why This Product Has Bigger Scope
Avatar is easy to misread if you look only at the surface.
It is not just:
- a nicer writing tool
- a nicer writing assistant
- a roleplay wrapper
- a personality prompt
- a text-polishing gimmick
- a tone pack
- a personality preset
The larger direction is much bigger:
The larger direction is closer to:
**If language behavior can be governed, it can also be tuned, reused, multiplied, and eventually deployed in many forms.**
- a governed language runtime
- a branchable behavior system
- a reusable avatar workspace
- a multilingual calibration surface
- a future avatar ecosystem
That is why Avatar matters for:
- one-person company workflows
- multi-voice content operations
- public writing systems
- avatar experimentation
- future service-facing language systems
- governed language infrastructure
That is why the product has structure around:
This page stays focused on the MVP.
- boot routing
- WFGY_BRAIN
- workflow
- reusable builds
- eval
- research
- community
The underlying direction is larger.
The product shape reflects the claim.
---
## 🧪 Demos, Eval, and Research
## 🧭 Suggested Reading Paths
### Demos
If you want to see how Avatar behaves before tuning your own build:
- [Demo Hub](./demos/README.md)
- [Atlas · PSBigBig Demo](./demos/psbigbig-demo.md)
- [Atlas · MiniPS Demo](./demos/minips-demo.md)
- [Multilingual Demo](./demos/multilingual-demo.md)
Not everyone should read Avatar in the same order.
### Eval
If you want the current credibility surface:
- [Eval Hub](./eval/README.md)
- [Persona Gates](./eval/persona-gates.md)
- [Multilingual Status](./eval/multilingual-status.md)
- [Blackfan Testing](./eval/blackfan-testing.md)
### If you are new
1. [⚡ Quickstart](./docs/quickstart.md)
2. [🧷 Boot Commands](./docs/boot-commands.md)
3. [🗣️ Tune Behavior in Natural Language](./highlights/tune-behavior-in-natural-language.md)
### If you care about product depth
1. [🧠 Language Behavior Can Be Engineered](./highlights/language-behavior-can-be-engineered.md)
2. [🛡️ Editable but Governed](./highlights/editable-but-governed.md)
3. [🔄 Dual Closed-Loop Design](./highlights/dual-closed-loop-design.md)
### If you care about future scale
1. [🌍 Multilingual Calibration](./highlights/multilingual-calibration.md)
2. [♻️ Reusable Builds](./highlights/reusable-builds.md)
3. [🌱 One Runtime, Many Avatars](./highlights/one-runtime-many-avatars.md)
---
## 🔍 Go Deeper
Once the main landing story is clear, the deeper layers become more useful.
### Practical docs
- [⚡ Quickstart](./docs/quickstart.md)
- [🧷 Boot Commands](./docs/boot-commands.md)
- [🛠️ How to Use WFGY_BRAIN](./docs/how-to-use-wfgy-brain.md)
- [🧭 Avatar Tuning Workflow](./docs/avatar-tuning-workflow.md)
### Evaluation
- [📊 Eval Hub](./eval/README.md)
- [🧪 Persona Behavior Checks](./eval/persona-behavior-checks.md)
- [🌍 Multilingual Status](./eval/multilingual-status.md)
- [🪓 Blackfan Testing](./eval/blackfan-testing.md)
### Research
If you want the deeper system story:
- [Research Hub](./research/README.md)
- [Architecture Overview](./research/architecture-overview.md)
- [Language Governance](./research/language-governance.md)
- [Firewall, SRDF, and τ_carry](./research/firewall-srdf-tau-carry.md)
- [🧪 Research Hub](./research/README.md)
- [🏗️ Architecture Overview](./research/architecture-overview.md)
- [🌐 Language Governance](./research/language-governance.md)
- [🧠 WFGY_BRAIN Theory](./research/wfgy-brain-theory.md)
- [🔄 Dual Closed-Loop Notes](./research/dual-closed-loop-notes.md)
### Community
- [🌟 Submit Your Avatar](./community/submit-your-avatar.md)
- [⭐ Awesome Avatar List](./community/awesome-avatar-list.md)
- [🧾 Avatar Submission Format](./community/avatar-submission-format.md)
---
## 🌱 Build Your Own Avatar
## ❓ FAQ
The official Atlas pair is only the beginning.
### Is Avatar just another AI writing prompt?
No. The public surface is designed as a governed runtime with routes, editable behavior, reusable builds, and later community branches.
The long-term direction of Avatar is not “pick one preset and stop”.
### Why one `avatar.txt` instead of multiple persona files?
Because one shared runtime gives later routes and branches a clearer center. It is cleaner, more scalable, and more legible than scattered starter files.
The direction is:
- start from a governed base
- tune behavior through `WFGY_BRAIN`
- build your own avatar
- compare it across tasks
- compare it across languages
- keep the good versions
- share them back with the community
### What does `hello YOUR_AVATAR_NAME` mean?
It is the intentionally bare custom-avatar route. It can stay funny and minimal, or become the seed of your own future avatar branch.
👉 Future community surface: [Awesome Avatar List](./community/awesome-avatar-list.md)
### Do I need to understand the full theory first?
No. The intended path is simple: boot a route, run a real task, observe, tune, compare, save. Theory comes later.
### Is multilingual work already finished?
No. It is a major product direction, but the current public multilingual surface is staged and still expanding.
### Can I submit my own avatar later?
Yes, that is the direction. The community layer is growing, and the Awesome Avatar List is currently **WORK IN PROGRESS**.
---
## 📚 Quick Links
## 🔗 Quick Links
- [🏠 Avatar Home](./README.md)
- [⚡ Quickstart](./docs/quickstart.md)
- [🧷 Boot Commands](./docs/boot-commands.md)
- [🛠️ How to Use WFGY_BRAIN](./docs/how-to-use-wfgy-brain.md)
- [🫂 Demo Hub](./demos/README.md)
- [🧭 Avatar Tuning Workflow](./docs/avatar-tuning-workflow.md)
- [✨ Highlights Index](./highlights/README.md)
- [📊 Eval Hub](./eval/README.md)
- [🧪 Research Hub](./research/README.md)
- [🌱 Awesome Avatar List](./community/awesome-avatar-list.md)
- [🌟 Submit Your Avatar](./community/submit-your-avatar.md)
- [⭐ Awesome Avatar List](./community/awesome-avatar-list.md)
- [⬆️ Back to WFGY Root](../README.md)