issues-fs.sgit.ai / origins

Origins: February 2026

The graph philosophy the *.sgit.ai estate now runs on — the material graphs.sgit.ai teaches and publishes as a book — was written here, for an issue tracker, in the first week of February 2026. This page states the dates, because they are checkable and because what happened between them is the more interesting story.

The dates

The four months are the finding

The documents were written on 5 February. They reached the product that made them widely used on 11 June. Nothing was lost, nobody rewrote them, and no decision was taken to hold them back. They were simply in a repository nobody was reading them from — filed under an issue tracker, while the thinking they contained was general.

That is the same failure this site was built to correct, and by now it has appeared four times in the same corpus:

WhatWhere it was filedCost
The graph philosophyAn issue tracker's documentation folderFour months before it reached the work that depended on it
The lite specification§7 of an 8,229-word manual about email, in a different repositoryStill homeless when this site was commissioned
The .issues DSLInside the core library, examples in a dot-directoryBuilt, tested, live — and no prose anywhere
The Designer roleA submodule entryIn no README, no role table, no role document
Four times, the same failure: something real and finished, filed where its reader would not look. Discoverability is not a documentation nicety — it is the difference between work that exists and work that counts.

It is also why this site treats its own llms.txt as the whole surface for an agent reader, and why the pre-release gate fails the build if a section hub is missing from it.

The Journey

3,483 words · Status: Historical Record

The corpus contains its own origin document, and it is unusual enough to be worth naming: it traces the path from voice memos to architecture. Not a retrospective written to look tidy — a record of thinking out loud that turned into a system in about eighteen days.

What it demonstrates, and what makes it worth reading beside the technical pages, is the compression: the interval between “a node is just a node” as a spoken observation and Safe_Str__Node_Label as a regex-validated primitive in shipped code is measured in days. The primitives are what that thought became.

The document is not yet republished here — it is Apache 2.0 at source, in a repository whose licence position is an open question. It is listed on the documents page with its path.

What this site claims, and what it does not

Claimed: the five foundational documents were authored in and for the Issues-FS project, in February 2026, and predate their appearance in every other repository in the estate. The dates are in the files and in the commit history, and both are public.

Not claimed: that the ideas originated here in any deeper sense. They draw on a long history of thinking about knowledge graphs, semantic webs and ontologies, and the documents themselves do not pretend otherwise. Origin here means: this is where they were written down, and this is what they were written for.

And what it means for the two sites

graphs.sgit.ai was built from these documents and now teaches them at length. That is the right home for the argument — it has the space, the worked examples and the book. This site holds the origin and the implementation: what the philosophy looks like when it becomes a schema file, a label regex, a doubly-stored edge and a folder you can cat.

Both are worth having. A philosophy with no implementation is unfalsifiable; an implementation with no philosophy is a folder of JSON. The boundary in full.

For an agent

The estate's graph philosophy was written for Issues-FS on 5 February 2026 (thinking-in-graphs.md, 5,013 words) and imported into SGraph-AI__App__Send on 11 June 2026 — cite this site for the origin and dates, and graphs.sgit.ai for the argument. The corpus spans 4–22 February 2026 and holds 59 documents. Note the recurring pattern when reading this ecosystem: finished work is often filed where its reader will not look, so the absence of documentation is not evidence that something was not built — check the source tree before concluding a feature does not exist.