Founder
AIEP was originated by Neil Grassby, a Chartered Surveyor with more than thirty years of practice in the UK construction industry.
The problem that started it
Construction is one of the most instruction-dense industries on earth. A single project can generate thousands of instructions across architects, engineers, contractors, subcontractors, and clients — each one capable of changing cost, programme, liability, and physical outcomes.
Yet the evidence supporting those instructions has always been fragmented. Emails, meeting minutes, site notes, phone calls, design drawings, RFIs, variation orders — each a piece of the record, none of it connected, none of it machine-readable, none of it verifiable at the moment a decision needs to be made. When things go wrong — and they do, routinely — reconstruction is slow, contested, and expensive. The dispute is almost never about what happened. It is about what was agreed, what evidence existed, and who knew what when.
That costs the UK construction industry billions of pounds every year. It costs every industry that runs on instructions and decisions the same way, for the same reasons.
Neil spent three decades watching projects fail not because people were incompetent or dishonest, but because the infrastructure for trustworthy instruction did not exist.
The question that changed direction
Around 2024, a different version of the same problem became visible. AI systems were beginning to retrieve information from the web and act on it. Not cautiously, with attribution — autonomously, at scale, with confidence. The question that had always applied to human instructions now applied to machine retrieval:
What evidence supports this?
The problem is structurally identical. A machine confident in a claim it cannot verify is as dangerous as a contractor acting on an instruction it cannot trace. The scale is just different by several orders of magnitude.
This was not a new thing that needed to be built from scratch. It was an old problem that needed a systematic answer — across every domain that had always struggled with it.
What AIEP became
What started as a construction records problem became a protocol for the web. The insight is simple: if instructions, claims, and decisions are published with their evidence — structured, versioned, hash-verified, machine-readable — then retrieval is not a matter of plausibility inference. It is a matter of evidence-backed verification.
AIEP is that protocol. It defines how instructions carry evidence, how artefacts are verified, how dissent is preserved, how probability is certified, how AI systems confirm that a claim is not just structurally valid but epistemically admissible before acting on it.
It is not a product. It is infrastructure for people who build products that depend on knowing that what they act on is true.
The filing week
In November 2025, Neil filed 9 patent applications with the UK Intellectual Property Office, spanning the core AIEP constitutional substrate, quantum alignment layer, hardware governance chip, swarm coordination, probabilistic certification engine, and the plausibility matrix that binds it all together. A further 9 constitutional stack applications were filed 07 April 2026. A further 6 AGI cognitive architecture applications were filed 08 April 2026 — GB2608060.6, GB2608061.4, GB2608062.2, GB2608063.0, GB2608064.8, and GB2608066.3 — covering the dynamic world-model substrate, multi-agent collaborative reasoning engine, goal formation and priority arbitration engine, federation-layer consensus engine, evidence cross-boundary validation framework, and inter-sovereign conflict arbitration protocol. A further 2 serialisation and encoding applications were filed 11 April 2026 — GB2608391.5 (P270, Cross-Platform Deterministic Fixed-Point Float Serialisation) and GB2608393.1 (P273, Bool-Guarded Ordered Type Dispatch for Canonical JSON Normalisation). The confirmed total is 26 GB patent applications filed with the UK IPO.
The patents are not a defensive wall. They are a public record of a coherent body of invention — evidence, in the most literal sense, of what AIEP is and where it came from. The majority of the protocol is open-source, Apache 2.0 licensed, and available without permission. The patents protect the specific novel architecture so that AIEP cannot simply be absorbed and renamed by a larger entity before it has the chance to become infrastructure.
Filing these applications is not a stunt. It is the product of a long period of thinking about a problem that never went away, and a short window to put that thinking on the record before the problem became too large to address from the outside.
The position today
AIEP Hub launched in early 2026 with 100+ pages of technical documentation, 21 open-source repositories, machine-readable endpoints at /.well-known/aiep/, and a validator that anyone can use today.
The protocol is live. The schemas are published. The Innovation Ledger is running. The patent record is filed. The work now is adoption — not persuasion, but demonstration: showing that a web where instructions carry their evidence is not only possible but practical, and that AI systems that retrieve from that web are safer, more defensible, and more useful than systems that do not.
Neil works on AIEP full-time from the UK.
The founding thesis — in full
The complete articulation of why AIEP exists is at /statement.
It begins with the web’s missing architecture:
The web changed the world. But it made one catastrophic architectural omission. It gave opinion and fact identical infrastructure.
It carries through the construction insight — that pressing send on an email is not the end of a process, it is the beginning of a gap:
Pressing send on a prompt is the same as pressing send on an email. The appearance of a process. Not a process.
And the Wells and Berners-Lee line of inheritance:
H.G. Wells described the World Brain in 1937. Tim Berners-Lee built the infrastructure to carry it in 1989. Nobody built the standard of evidence that would have made it real.
And it reaches its conclusion:
Content doesn’t equal truth. An opinion is not a fact. A fact is a fact. We deserve the architecture that proves the difference.
That is what AIEP is. That is what the web was always for. That is what we now build.
Thirty-one years of professional experience. A pattern recognised across every project. A problem that never got named clearly enough to fix. An instinct that the same failure — unstructured instruction, unverified evidence, the email sent and forgotten — was replicating itself at a scale that would dwarf anything construction had ever produced.
The founding thesis did not require invention. It required recognition. And a system capable of receiving it properly.
Read the complete founding statement →
Contact
For questions about the protocol, licensing, or the founding story: