influences.sgit.ai / register / Tim Berners-Lee & the Semantic Web

Tim Berners-Lee & the Semantic Web

traced topic + person full

An influence on Dinis Cruz — entry 3 of 25 in his register.

The format at its strongest, because this is an influence argued with. The dream of machine-readable meaning shaped the estate's graph work; the diagnosis of where the community went wrong shaped it just as much.

Block 1 The anchor

The Semantic Web — Tim Berners-Lee, James Hendler and Ora Lassila — Scientific American · 2001

Linked, never rehosted. This site explains why the work resonated and traces where it was applied; the work itself stays where its author put it.

The article that put the idea in front of a general audience. The wider programme — RDF, OWL, linked data, and later Solid — is in Block 7.

Block 2 In his own words

First person is the house style on this site and nowhere else in the network — what resonates with me and why is the genre, and paraphrasing it into corporate third person would destroy the evidence. Until the briefing document for this entry arrives, this block carries the first-person statements that already exist in the corpus, cited by file.

The Semantic Web community identified the right problem… But the community made a subtle mistake in practice. They ended up attaching meaning to nodes rather than deriving meaning from edges.
— Dinis Cruz, briefs/.../library/concepts/v0_4_0__thinking-in-graphs.md

The section is titled The Semantic Web's Insight (and Mistake), and the ordering of those two words is the entry. An influence you have a technical disagreement with is an influence you have read closely.

Two things resonate, and they pull in opposite directions. The first is the ambition: a web where a machine can act on meaning rather than pattern-match on strings is still the right target, twenty-five years later, and most of what is now called an AI agent is a worse-engineered attempt at it.

The second is the diagnosis of the failure, and it is a technical one rather than a sociological one. Attaching meaning to a node makes every node an assertion that has to be agreed on before anyone can use it; deriving meaning from edges makes agreement local and lets two parties disagree about what a thing is while still agreeing about how it relates. That inversion is load-bearing in the estate's graph substrate.

The third strand is live rather than historical: Berners-Lee's Solid puts personal data in pods the person controls, and this estate puts it in encrypted vaults the person holds the key to. Those are the same instinct with different threat models, and the corpus treats them as complementary architectures rather than competitors.

Block 3 The principle

Meaning should be machine-readable, so that independent parties can exchange it without agreeing on a schema first.

Note what the principle does not say: that meaning lives in the things. That distinction is the whole content of this entry's disagreement.

Block 4 The trace table

Pattern from the anchorWhere the estate implements itVersionStatus
Machine-readable meaning as the target — a graph a machine can act on rather than a document it can only parseThe estate's graph substrate and the concept document that sets out its thinkingv0.4.0implemented
Meaning attached to nodes (RDF's typed-resource model)Deliberately inverted. The estate derives meaning from edges instead, and the concept document states the inversion as a correction rather than a variationv0.4.0absent
Meaning derived from connection between thingsThe edge-first graph model the estate's ontology work is built onv0.4.0implemented
Personal data under the person's own control (Solid pods)Encrypted vaults, keyed by their holder — compared to pods explicitly, as complementary architectures with different threat modelsv0.6.17partial
Interoperation with Solid itself — a vault that can read or write a podNowhere. The brief is an architectural comparison, not an integrationv0.6.17absent

This table on its own, with its sources → · as JSON · as markdown

Two of these rows are the influence being inverted rather than implemented, and they are marked as such. A trace table that could only record agreement would be useless on an entry like this one.

Block 5 The gaps, as build specs

Patterns from the anchor the estate does not implement yet, written precisely enough that an agent could pick one up as a work item. This is what makes an influence page forward-looking: it is provenance and backlog in the same document.

G1 · build spec

The Solid bridge the comparison implies

The integration brief establishes that pods and vaults solve the same problem with different trust assumptions, and stops there. The build spec is the smallest thing that would make the comparison real: a read path from a Solid pod into a vault, or a vault snapshot published in a pod-readable shape. Either direction would turn an argument into an artefact and would be the first interoperation this estate has with the programme it says it inherited from.

G2 · build spec

The edge-first claim has no published counter-test

The entry asserts that deriving meaning from edges avoids a failure mode that node-typing has. That is a falsifiable claim and nothing in the corpus falsifies it — there is no worked case showing a query the node-first model answers and the edge-first model cannot. Building that case, and publishing it whether or not it is comfortable, is the honest version of this disagreement.

Block 6 The checklist

What to ask of new work in this influence's light. The checklist is the influence made operational — the part an agent can run without having consumed the anchor work.

Block 7 The wider library

The rest of the work, linked and never rehosted, each item with one line on what it adds. The full union of every entry's Block 7 is at /library/.

The corpus evidence

What the mining run found, by path. These are the files that put this entry in its tier — the claim on this page is checkable against them, which is the whole point of the format.

Path in the corpusWhat it carries
SGraph-AI__App__Send/library/concepts/v0_4_0__thinking-in-graphs.mdthe section The Semantic Web's Insight (and Mistake) — the node-first versus edge-first critique, in the document that sets out the estate's graph thinking
SGraph-AI__App__Send/team/humans/dinis_cruz/briefs/02/24/v0.6.17__architecture__solid-protocol-integration-complementary-architectures.mdSolid pods and sgit vaults as complementary architectures, with Bruce Schneier's involvement at Inrupt noted
(corpus-wide, ~35 files)the wider footprint the mining run found across briefs and concept documents

Paths are as recorded by the mining run behind the commissioning pack (v0.33.62, 25 August 2026). This repository holds the website, not the corpus, so they are cited rather than resolved — R2 in the comms queue.

Where this sits

No nesting and no sibling site: this entry stands on its own. The influence map draws every relation the register records.