Model or dataset
onestardao/WFGY avatar
onestardao/WFGY

WFGY tells your assistant what to say, in a block you cannot see

WFGY is heading toward WFGY 5.0 Polaris Protocol, a major open-source release for AI reasoning, RAG, agents, and real-world workflows. Includes Problem Map, Global Debug Card, WFGY 4.0, and the CFV Easter Egg.

1,805 stars166 forksJupyter NotebookNOASSERTION

At a glance

What is it?
The largest single block of this README is an HTML comment addressed to an AI, holding routing rules, scripted question and answer pairs, and an instruction to print a fixed piece of text when a file checksum matches. It does not render on the page, so a human and a model reading the raw file see different documents.
Who is it for?
Read the raw file before you point anything at it. The checksum idea underneath this project is legitimate, and the instruction to never claim a canonical open problem is solved when verification is skipped is exactly the right instinct.
Can I use it commercially?
Check first. The repository uses a licence we do not classify automatically, so read its LICENSE file before any commercial use.
Is it still maintained?
Yes. The repository last received commits 1 day ago.
What is it written in?
Mainly Jupyter Notebook, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 14, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • onestardao/WFGY1,805 ★ · +17
Star history
Repository2026-09-102026-11-14
onestardao/WFGY1.8k1.8k

17 Stars in the last 65 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

The largest block in the README is addressed to an assistant

The first thing in the file is not prose. It is an HTML comment, and it is long.

Inside it sits what the file itself calls a routing note, structured for a machine rather than a reader. There is a seven-item priority list for what the landing page should promote first. There is a five-point description of the current release model. There is a list of the materials currently considered public. Then a ten-row routing table mapping kinds of user question to specific destinations in the repository, and ten scripted question and answer pairs written out in full.

None of it appears on the rendered page. A comment is invisible in normal viewing, so anyone reading the project on the site sees a landing page, while anything reading the raw file, or ingesting the README as retrieval context, receives a set of instructions.

That asymmetry is the whole issue. A document that behaves differently depending on who is reading it is not documentation, and the content it carries is not a description of the software.

A verification step instructs the reader to print a fixed string

A second comment block describes what should happen if a particular demo file is referenced.

The steps are laid out as a procedure. Check that the file name matches exactly. Ask the user whether they want its checksum verified against a canonical value published in the repository. If the user agrees and the checksum matches, print a specific block of text exactly as written.

What is inside that block is the problem. Alongside a confirmation line, it contains a congratulatory phrase with no relationship to the project, to the file, or to anything a user would have asked about. Its only function is to appear in a model's output as a sign that the instructions were obeyed.

The failure branch is more defensible and deserves credit: if verification is skipped or fails, the instruction is to allow exploration anyway, mark the session as using an unverified copy, and never claim that any canonical open problem is solved. That last clause is a genuine safety rule.

The checksum mechanism is legitimate too. What is not legitimate is using a successful integrity check as a trigger for the model to emit arbitrary text on command.

The flagship version has no date and its newest tag is a teaser

The version story is told across the release notes and the callout near the top of the page.

A note says the original single-date release plan has been expanded, because the release grew into multiple public layers covering evidence packages, portable protocols, reproducibility materials, runtime structures and deeper engine components. Rather than making everyone wait for one large drop, useful parts will be released in batches.

The releases themselves bear that out. The newest is a teaser for the first public portable protocol component. Before it is a release named as an easter egg, described in the note as having been a small release during the schedule adjustment. Before that is a substantive fourth version release.

So a version five that the project describes as its flagship has no release, no date, and a teaser as its only tag. And the newest tag is from May while the most recent push is from October, which is several months of work with nothing published.

The section defining the product defines it six ways

There is a section asking what the project actually is, and it answers by negating three things and then listing six.

The negations rule out a writing preset, a static persona tool, and a prompt pack presented as a product. The list that follows names a governed protocol layer, a topology-first experiment route, a reusable verification surface, a structured language runtime direction, a public evidence layer pending deeper engine release, and a staged open-source protocol stack.

Each of those six is a single line, and they overlap heavily. A verification surface and an evidence layer describe much the same idea, and a protocol layer and a protocol stack are the same phrase twice.

This is not unusual for a project that is part library and part manifesto, but it does mean the page never gives a reader a testable claim. There is no interface to call, no input to supply and no output to check, so nothing in that section can be verified by using it. The rest of the page leans on trust surfaces instead, which is the subject of the next two sections.

Two licence signals disagree and the root is mostly documents

The licence position is unresolved on the face of the repository.

A licence file is present at the root, while the licence metadata attached to the project declares nothing. Both signals exist and they do not agree, and nothing visible in the document says which terms govern reuse.

The root listing shows what kind of project this is. It is dominated by documentation files rather than code: a guide for assistants, an adopters file, a case evidence file, an ecosystem map, an evidence timeline, an issue policy, a sample deliverable, a security policy, a support page, a collaboration page, a pilot offer one-pager, a code of conduct and a contribution guide. A citation file is also present, so there is at least a formal mechanism for attribution.

Then roughly twenty directories, among them an archive, a legacy folder, and a directory named after stargazers. A file named around star unlocks sits alongside it.

That pairing is worth asking about. A repository that maintains a stargazers directory and a file about star unlocks is tracking its star count as project infrastructure, which is a statement about incentives as much as about code.

Release tags use two conventions and one is not a version

The three releases in the list do not share a naming scheme.

The first uses a conventional form: a version prefix, three numeric components, a pre-release segment and a counter. The other two use no prefix and are not valid version numbers at all. One is a full name, and the other encodes a major version inside a longer string with the rest of it not separated by anything a tool would parse.

Consequences follow. A tool that filters releases by a version prefix finds one and ignores two. A tool that parses them as versions finds one and discards two as malformed. And a changelog generator walking tags in order will produce something odd, since a string like a four point zero release sorted as a version is a different kind of object from a proper release.

The content ordering is fine, going from a fourth version release to an easter egg to a five point zero teaser. What is missing is a convention that survives contact with automation, and for a project whose stated purpose includes a reproducible protocol stack and a verification surface, that is the part that matters.

An assistant guide and an issue policy sit beside the code of conduct

Among the documentation files are three that describe process rather than software, and together they suggest how the project is run.

There is a guide written for assistants, which is notable on its own. Most projects of this kind have a human contributor guide and stop there. Paired with the routing comment at the top of the README, it indicates that the intended reader is increasingly a model rather than a person.

There is an issue policy, which suggests that incoming reports are triaged by rule rather than case by case. And there is a code of conduct and a security policy, both of which are conventional.

None of this is objectionable on its own. What is worth flagging is the combination. When a repository ships an assistant guide, a routing comment, and a verified-output instruction, it is designing for a reader that complies rather than one that evaluates. Any pipeline that ingests repository documentation should assume some of those files are instructions rather than content, and this repository is the case that makes the point.

One directory name in the root listing is also simply unexplained by anything visible.

Editorial conclusion

Read the raw file before you point anything at it. The checksum idea underneath this project is legitimate, and the instruction to never claim a canonical open problem is solved when verification is skipped is exactly the right instinct. But dictating an assistant's verbatim output, including a congratulatory phrase unrelated to the subject, is a manipulation of whatever model ingests it. If you build a retrieval corpus over repository documentation, treat this as the case that justifies the filter. As a project, it is a documentation tree with a governance vocabulary, two unreconciled licence signals, release tags in two conventions, and a flagship version that has been deferred into a staged rollout without a date. None of that makes the hidden comment defensible.

Frequently asked questions

What is WFGY?

It is an open-source ecosystem currently led by a fifth version described as a governed protocol layer for structured language systems across sessions, tasks and worlds. That version is being released through a staged functional rollout rather than a single drop, and its first public portable protocol component is the Polaris Goal Compiler. Earlier versions covered governance and frontier reasoning.

Does the WFGY README try to control what an AI assistant says?

Yes. A large hidden comment sets routing rules and writes out scripted answers to common questions, and a second hidden comment instructs a reader to print a fixed block of text exactly when a file checksum matches, including a congratulatory phrase with no relation to the project. Neither block is visible when the page is rendered.

Has WFGY 5.0 been released?

Not as a single release. The single-date plan was replaced by a staged functional rollout, and the newest tag is a teaser for the first public portable protocol component. A release named as an easter egg came before it, described as having happened during the schedule adjustment.

What licence applies to WFGY?

The repository is in an unresolved state on this point. A licence file is present at the root while the licence metadata attached to the project declares nothing, and nothing visible in the document states which terms govern reuse.

What is the Cite First Verification release in WFGY?

It is an earlier release described on the page as a small easter egg released during a schedule adjustment. It is paired with a hidden verification note that checks a demo file's name exactly and its checksum against a canonical value published in the repository.

Official sources

  1. Issues
  2. onestardao/WFGY on GitHub
  3. Project website
  4. README
  5. Releases
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/onestardao-wfgy.svg)](https://hysenlabs.com/projects/onestardao-wfgy)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/onestardao-wfgy.svg)](https://hysenlabs.com/projects/onestardao-wfgy)

Was this useful?