Back
2026-05-26 · Dublin, Ireland · 9 min read

A semantic web to bring clarity to how Ireland is actually governed. For me first. For you too.

I kept getting whiplash between Government press releases and the responses from the organisations they were addressed to. So I built the thing I needed to read both at once with sources attached — a semantic web to bring clarity to complexity. Then I noticed everyone else needed it too. Curb cut effect.

James Spalding
Building in the open · Dublin

The short version

I built /public-record/ireland because I could not tell, on any given week, whether the Irish State was doing what its press releases said it was doing. I would read a gov.ie announcement at 8am and an Inclusion Ireland or DFI response at 9am and I had no way to know which one to believe without four hours of source-chasing per topic, and there are seven topics a week.

I am one person. I cannot do that. Neither can a TD. Neither can a parent of an autistic child. Neither can a teacher trying to explain to a fifteen-year-old how their country actually works.

So I built the thing I needed to stop being confused. A semantic web to bring clarity to complexity. Twenty-five case studies of large Irish public projects, decomposed into milestones, impacts, EU obligations, citizen objections, comparable failures, with every fact carrying a citation. An accountability atlas of every department, body, party, officeholder, mandate and Dáil division — cross-linked into one graph so every claim resolves to a person, a body, a party, a statute, a vote. A natural-language ask layer over the whole graph.

I built it for me. Then I realised the people I keep arguing with on Twitter could use it. Then I realised the civil servants the State is too understaffed to give time to could use it. Then I realised the journalists writing under deadline could use it. Then I realised the citizens who turned up to a planning hearing not knowing the difference between An Bord Pleanála and ABP-MARA could use it.

Curb cut effect. You build the ramp for the wheelchair user. The parent with the stroller, the courier with the trolley, the kid with the skateboard, the older person whose knees are tired — they all use it too. Nobody loses.

The thing I kept hitting

Here is the pattern that radicalised me.

Tuesday: the Department of Children, Disability and Equality publishes a press release about a "€619 million disability uplift" in Budget 2026.

Tuesday afternoon: DFI's CEO calls the same Budget "a devastating setback for disabled people" because the once-off cost-of-disability lump-sum payments that ran in each of the previous three Budgets were withdrawn with no permanent replacement.

Both are true. Both are sourced. Both are sitting on the public record. And both made it into different RTÉ news bulletins on the same day.

Which is the real story? You can only know if you do the actual work. Pull the Disability Capacity Review (2021). Read the Indecon Cost of Disability research (2021). Find the headline development funding line in each of Budgets 2024 (€64m), 2025 (€42m), 2026 (€619m gross). Compare the run-rate to the Capacity Review's implied €80–90m/year of new development funding. Then realise that the gross figure includes the once-offs being absorbed, and check whether the run-rate for permanent disability allowance moved at all.

It takes about three hours per topic to do that properly. The Irish news cycle gives a topic about two days before the next thing.

Multiply across autism assessment, special education classes, MetroLink, the Occupied Territories Bill, the Sanctions against Israel Bill, the Children's Hospital, the Mica-Pyrite redress, the Mother and Baby Homes redress, the National Broadband Plan, the National Childcare Scheme, Uisce Éireann, the Drug Law reform, the Defence Amendment Act, the Shannon US military stopovers, Section 39 pay parity, the AON ruling, the Autism Innovation Strategy. Now do that simultaneously, every week, while holding down a job.

You can't. Nobody can. The asymmetry between announcement and verification is structural. It's why public-policy debate in Ireland is mostly people shouting past each other with different press cuttings.

The semantic web

What I built is at /public-record/ireland. Four surfaces, one ontology.

Projects. Twenty-five case studies and growing. Navan Rail, Metrolink, the Children's Hospital, BusConnects, National Broadband Plan, Disability Capacity Review, the Occupied Territories Bill, Shannon US military stopovers, Irish dual-use export regime, Special Education Classes, Grangegorman, Cork Light Rail, the M20, DART+, Defence Amendment Act, Sláintecare, Housing for All, Mica-Pyrite redress, Mother and Baby Homes redress, International Protection / Asylum, CAMHS, Uisce Éireann, National Childcare Scheme, Drug Law reform, Rural Transport / Local Link, Autism Assessment and Intervention Pathways Protocol, Sanctions against the State of Israel Bill. Each one decomposed into a fixed schema: timeline of milestones, impacts (community, fiscal, biodiversity, heritage, displacement), EU and Irish-statute obligations, citizen objections, comparable projects, cost and schedule envelopes. Every fact carries its own citation.

Accountability atlas. Every department, agency, party, officeholder, mandate, commitment, division, claim and bill in one cross-linked semantic graph. When you read that Minister Norma Foley TD launched the Autism Assessment Protocol on 26 May 2026, "Norma Foley" is a link to her officeholder record, which is itself linked to her party, her department, her terms, her votes, her commitments, and every project she's responsible for. Including the ones that haven't gone well.

Ask. A natural-language interface over the dataset. You can ask "what has Pauline Tully voted on?" or "what has the Department of Health committed to on assessment of need?" and the system runs structured intents against the graph instead of hallucinating.

Fiscal model. Budget projections by department, by programme, against published run-rate. So you can see for yourself whether €619m is the figure the underlying plan requires.

The rules — what makes this different from a wiki

A wiki collapses under its own weight because anyone can write anything and the citation discipline is voluntary. I wrote a rule-set instead and let it constrain everything.

Source ladder. Tier 1: gov.ie, agency portals (tii.ie, nationaltransport.ie, npws.ie, epa.ie, cso.ie, about.hse.ie), irishstatutebook.ie. Tier 2: oireachtas.ie (Dáil/Seanad debates, divisions, written answers). Tier 3: eur-lex.europa.eu for directives, regulations, judgments; the ICJ; OHCHR. Tier 4: quality press as corroboration only — never the sole source for a numeric or attributional claim. Tier 5: Wikipedia for orientation only, then chase its citations and cite those instead. The hierarchy is enforced at authoring time.

Full ontology. Every actor reference resolves. "Norma Foley" in a milestone is not a string — it's an EntityRef that resolves to her officeholder id. Same for every body, every party, every commitment, every comparable. The graph is the product.

Bounded authoring. A skill called gov-ie-research defines exactly what a diligent research round looks like — what to read, in what order, how to record what's not yet verifiable. It's checked into the repo at skills/gov-ie-research/SKILL.md. Another skill called skill-opt runs bounded optimisation rounds on the first skill so it gets sharper, not noisier, over time. The discipline is itself the product.

Every claim carries its tier. If a claim is sitting on a tabloid-only citation, the schema says so and the UI shows it. Hand-corrections and rejected edits live in a separate audit trail. The corrections become the input to the next optimisation round, not deleted history.

A worked example — the Autism Assessment and Intervention Pathways Protocol

On 26 May 2026, the Minister for Children, Disability and Equality launched the State's first nationally standardised autism assessment pathway alongside an easy-read companion document. I happened to be looking at the Minister's Instagram post on launch day. By that evening the semantic web had a structured case study at /public-record/ireland/projects/autism-assessment-protocol.

It has 8 milestones — Disability Act 2005, the Autism Innovation Strategy in July 2024, the PSI / IASLT / AOTI consultation responses in December 2024, the HSE Bespoke Panel procurement in September 2025, the AON-reform package announcement on 9 December 2025 (€20m for 6,000 private-provider assessments, 11 new in-reach teams totalling 44 staff), the missed go-live on 9 February 2026, the 20,209-person waitlist beyond the statutory six-month deadline reported on 16 March 2026, the Ministerial launch on 26 May. Five impacts including the AON backlog, the EU-outlier adult-pathway gap, the professional-body evidence-base concerns, and the 15-week implementation slip. Four legal obligations — Disability Act 2005, UN CRPD Articles 25 and 26, HIQA standards, Equal Status Acts. Five citizen objections — AsIAm CEO Adam Harris's "first step but not going to make a difference to waiting times", INTO's "profound breach of trust" framing, the PSI Special Interest Group in Autism's tiered-model evidence-base concerns, Pauline Tully TD's call for the HSE to accept the 2025 High Court ruling, Holly Cairns TD's UN CRPD rights framing. Three comparables — the Disability Capacity Review's slippage, the 20-year AON statutory non-compliance, the Sláintecare pattern.

Every one of those carries a citation. The Minister's Instagram quote is verbatim. The 20,209 waitlist number ties to the Irish Examiner reporting of 16 March. The €20 million envelope ties to the gov.ie press release of 9 December 2025. You can verify the case study in about ten minutes. You can read it cold in about three. The same shape works for the next case study, and the next.

Why this fits 8GI's mission

8GI Foundation's name is "general intelligence". Most people, when they hear that, think artificial general intelligence. AGI. The next model. The next benchmark.

That's half of what general intelligence means. The other half is the general public — demos — having a greater understanding of reality. A smarter society. People able to read the actual mechanisms by which they are governed without needing a humanities PhD and seven hours an evening.

The two halves are the same thing from different sides. A society that understands itself uses better tools. The better the tools, the more the society can understand. The feedback loop is the whole point. Build agents that help citizens read the State. Build a State that is more legible to its citizens. Both directions matter. Neither is finished.

The semantic web is one concrete deliverable in that loop. There will be more. Other countries will get the same treatment — the schema is country-agnostic, only the source ladder needs to be localised. South Africa is next. The structured-research skill ports.

Try it

/public-record/ireland — the country landing.

/public-record/ireland/projects — the 25 case studies.

/public-record/ireland/accountability — the atlas.

/public-record/ireland/ask — the natural-language layer.

If you find a fact that's wrong, the case study will tell you exactly which source it came from. Tell me. The corrections feed the audit trail. The audit trail feeds the next optimisation round. The next round produces a sharper case study. The discipline is the product.

I built it for me. Take it.

P.S. What should this thing actually be called?

"A semantic web to bring clarity to complexity" is what the thing is, not what to call it. It needs a name — something short enough to live on a navbar, descriptive enough that a citizen who has never heard of it knows what they're clicking on, and honest about the discipline (tier-1 sources first, bounded edits, every claim cited) rather than just the surface.

Two questions, in order, and I would actually like to hear the answers.

  1. What would you call it? Working candidates so far: Governopedia (most legible, instantly understood, but the -opedia suffix is 25 years old and slightly populist), Polity (clean single word, has historical weight, grows with the discipline), Stategraph (technical, honest about the semantic-web side), Public Record (descriptive, almost generic). None of them feel right yet. Better suggestions welcome — including ones in Irish.

  2. Should it be a standalone platform? Right now it lives at /public-record/ireland under 8gi.org. As the country count grows — South Africa is next, the schema is country-agnostic — does it deserve its own roof? A governopedia.org (or whatever the name turns out to be)? Or does the 8GI Foundation framing carry it, since the legibility-of-the-State half of "general intelligence" is structurally part of the mission?

Reply on whatever channel you have. Names get chosen the way most names get chosen — somebody else says the right one and the rest of us notice.

I built it for me. Take it.


By James Spalding · Dublin · 8GI