Blog
Notes on typed graphs, openCypher and turning a vault into storage that agents can use.
Who knows, who decides, what the agent may do: a team map in tg
The tools half of "From Islands to Archipelago": a team ontology for people, teams and agents, typed edges for ownership, knowledge and decisions, seven queries that route people to each other, and a skill that teaches agents to do the same.
Declaring document elements: one shape, one reading
TGS 0.5 settles how a type declares the blocks inside its documents. YAML declares and XML occurs. Attributes have four keys and a reserved id, nested tags have a scope, counts are per container, and every shorthand has one normal form. With a checklist.
Explainable blocks: an ontology for the inside of your documents
A type can now declare and explain the XML blocks its sections hold, such as a decision's options, a requirement's rules or a runbook's rollback. Blocks with an id become graph nodes that code can point at, the rest stays prose, and tg checks both.
One big file or a graph of small ones?
About sixty papers and benchmarks on agent context: what long context costs, what the 2026 AGENTS.md studies found, whether XML beats Markdown, where graphs help and where they lose, and the layered design the evidence points to.
Bring your own spec framework
tg now traces Spec Kit, Kiro-style specs, BMAD, Get Shit Done and ADRs to code and tests, next to OpenSpec. tg init asks which one you use and installs it if you have none.
Specs that know where their code is
We linked 239 OpenSpec requirements to their code, tests and docs with typed annotations. An audit on the way found 13 drifted specs and 11 real bugs; one command now keeps the three in step.
A slip box that answers back
Build a Zettelkasten as a typed graph, connect it to your books and articles, compose it with a library ontology, and ask it a dozen questions, from the unprocessed inbox to the ideas that came from fiction.
A shared space for agents and humans
Why coding agents need the same documented, checked knowledge as the people they work with, and what that looks like in plain files.
Give Claude Code a memory in five minutes
Install the tg CLI, let tg init wire up Claude Code, write a first typed link and see what the agent gets.
Ontologies you can download, and combine
Five ready-made ontologies as plain Markdown (Zettelkasten, books, requirements, prompts and agents, and an OKF catalog) and the rules for putting several in one graph: a shared core, multi-label notes and mixins.
One vocabulary for the whole team: using the code ontology
Six intent types and eleven edge types that tg init installs as a note and a TGS schema, and how developers, architects, product people and agents use them.
Your codebase has two graphs. You only maintain one.
Typed graphs for developers: what to model, what to leave to the machine, why specs are code now, and a small ontology for intent and code.
An ontology you can edit by hand, and hand to an RDF tool
Typed Graph Schema: types, edge types and templates in plain YAML notes, exported to and imported from W3C SHACL, and published as an open specification with a JSON Schema and conformance examples.
Getting your vault OKF-ready
Four habits that make notes conformant with Google Cloud's Open Knowledge Format, a one-command export that keeps typed edges, a conformance check, and reading any OKF bundle as a graph.
Your vault as a property graph
Typed, signed edges with properties, typed nodes and schema notes, openCypher queries and graphs that look like what they mean, all in plain Markdown.
Your vault as RAG storage
The sidecar serves your vault to programs and agents: full Cypher on LadybugDB, local vector search over notes and edges, GraphRAG that follows links and cites sources, and MCP.
tg: your docs and your code as one typed graph
A CLI that checks design docs against code, links functions to docs with typed edges, queries both with Cypher and sets up coding agents safely.
tg and lat.md: what is the same, what differs
Same folder format and the same check verdicts. What tg adds, where lat.md is stronger, and how to switch in four commands.
Why a typed graph for your codebase, and why you still write the docs
Typed links between design and code, what tg adds to lat.md, and why intent cannot be generated from the source.
A vault that teaches itself
Download the demo vault: a small research lab with the plugin preinstalled, a guide to typed graphs and a hands-on Cypher course.