Worldmind determines what can be built.
Worldmind is the research system behind every Invenio file. Specialist agents gather the evidence and argue over it, a Verification Court rules on each claim, and a person signs before anything leaves the firm.
The announced campus will be energised and operating before the end of 2028.
- Ruling
- Constrained, not yet executable
- Evidence grade
- 03, directional benchmark
- Binding constraint
- No grid connection offer on the public record. Planning granted, power not secured.
- Bench
-
Scholar: uphold Quantis: uphold Praetor: uphold Enigma: dissent
- Dissent
- On-site generation could remove the grid dependency. Registered as a separate claim for testing.
- Next action
- Re-examine when the next connection offer round is published. Prediction dated and logged for scoring.
Announcements are cheap. Delivery is rare. Nobody keeps score.
Capital is committed every week on the strength of press releases: a gigawatt here, a new campus there, a fund that will close by year end. Most analysis repeats the announcement. Very little asks whether the thing can physically, legally and financially be built, and almost none goes back later to check who was right.
Worldmind is built to do exactly that, continuously and at a scale one analyst cannot reach. It reads the filings, the planning files, the grid queues and the registries, breaks each story into claims that can be tested, and keeps a dated record of what it said and how it turned out.
Every forecast is sealed before the outcome, then scored in the open.
Banks, consultancies and research houses publish forecasts every day. None of them keeps a score that an outsider can audit. Misses are quietly forgotten, and hits are remembered loudly.
Worldmind registers each prediction as a dated claim that can be proved right or wrong. At the moment it is made, the claim is sealed with a cryptographic fingerprint and a timestamp. When the date arrives it is revealed and scored against what happened. We cannot edit a forecast after the fact, and we cannot bury one that failed.
The result is an intelligence record whose accuracy a stranger can check. As it builds, it is published in full, misses included, alongside a retrospective study of what the market announced against what it delivered.
Six steps from a question to a signed ruling.
The same evidence grades you see in an Invenio file are enforced inside the system. No agent can promote its own work. A claim only moves forward when another part of the system has tested it.
The question is bounded
A mandate sets the sector, geography, time horizon and exclusions, so the work stays tied to a decision.
Agents gather the record
Specialist agents pull filings, registries, planning and grid documents, and log the origin and date of every data point.
The story is broken into claims
Each announcement becomes a set of subclaims that can be shown true or false by a named date.
Dissent is commissioned
Two deliberately unorthodox agents, Enigma and Delirium, argue the other side. Their useful objections become claims in their own right.
The Verification Court rules
A bench of agents grades the evidence, names the binding constraint and records every vote, including the dissents.
A person signs, and the result is scored
Nothing is released without human approval. Every prediction is dated, and hits and misses are scored against what happened.
Seven words for how real a project is.
Worldmind's first theatre is European AI infrastructure, with Ireland covered in the greatest depth. Every tracked project holds one status on this scale, and it only moves when the evidence does.
- AnnouncedA public statement exists. Nothing more.
- EvidencedFilings, land or applications confirm real activity.
- ConstrainedA named obstacle blocks delivery: power, planning, capital.
- Conditionally executableBuildable if a specific condition is met.
- ExecutablePower, permission and money are all in place.
- In deliveryConstruction is under way and verifiable.
- OperatingEnergised and serving load.
Projects can also be marked Suspended or Cancelled. Alongside the Index, the Constraint Frontier shows which single obstacle is holding back the most capacity in each market.
A working system, governed like an institution.
Agent factory
Specialist agents are created, scoped and retired from one place, each with a defined remit and a record of its work.
Verification Court
The ruling layer. Claims are graded, constraints named and votes recorded before anything reaches a person.
Compounding memory
Semantic memory links every source, claim and ruling, so each engagement starts from everything the firm has already learned.
Spoken briefings
A voice layer reads the founder a summary of rulings, open questions and items awaiting approval.
Four permission tiers
No agent takes an external action without approval at the right tier. Nothing is hard deleted. Every change is traceable.
Built on a standard stack
Next.js and Supabase, with row-level security and a gated release process for every change to production.
Invenio Protocols is the firm. Worldmind is how it scales.
We are opening a small number of design partnerships with infrastructure investors, lenders and developers who want the Executability Index applied to their own pipeline. We are also speaking with investors about the pre-seed round.