Companies House 17160992 · 16 April 2026

Wot.Technology

The deepest rethink of the computing substrate in thirty years — shipping today.

Two customers in production. Twenty million signed thoughts in our own deployment. Four defensive publications filed, with more in flight.

  • Stave — inference at 66 ms, 100% recall, zero adversarial pass-through.
  • Ferb — cross-architecture layer grafting, +49% reasoning, 8 MB streaming buffer regardless of model size.
  • Content-addressed weight storage — 24 B model at 88 tok/s sustained, 117 peak; streamed from a distributed mesh, never all on one box.
  • Internal unpublished research — 14–45% perplexity reduction across six production-grade models; specifics when the IP is locked.

Computing gets measurably better — for free — just by doing it better. No new silicon. No bigger models. No more data. Just the right substrate underneath, and every ceiling the industry accepted as fixed falls two orders of magnitude lower.

Plus: AI that cannot hallucinate by construction. GDPR Article 17 erasure and EU AI Act Article 86 explanation — both satisfied by the maths, not by policy.

Five services: wot.is, wot.was, wot.now, wot.if, wot.then — on FlowSpace, our platform. ThoughtSpace is the consulting arm when picking up the tools isn't the answer. Anyone can build. Anyone can hire. Let us know.

Request early access See the five surfaces Publications

A quintet, in seven parts

Wot.Technology is five commercial services — capture, history, present, possibility, future — built on one data and hosting platform, with a consulting arm for when you'd rather we came in and solved it. License the services, build on the platform, hire the consultants. Or all of them.

I · ᛟ · PAST

wot.now

What is.

Current-state data dashboards and modelling for SMEs. Live picture of your business pulled from the graph, with receipts on every figure.

Spreadsheet-tangle to single-screen in one step. Every number traces back to the source system that wrote it and the moment it landed — so "where did this come from?" always has a literal answer.

For founders and ops leads who've outgrown spreadsheets but can't stomach a six-figure ERP.

Read more →

II · ᛏ · PRESENT

wot.if

What could be.

Conversational AI with receipts. Generative reasoning over a content-addressed graph, with because-line provenance on every output.

Voice in, conversation through, action out — with signed receipts the whole way. Hallucination becomes the structural absence of a because-line, and that absence is a query you can run, not a guess you have to live with.

For teams using AI for work that has consequences, and for compliance-adjacent businesses facing Article 86.

Read more →

III · ᛇ · FUTURE

wot.then

What will be, by the numbers.

Deterministic mathematical simulation and what-if scenario projection. Spreadsheet-grade rigour with cryptographic provenance. No generative AI.

Explicit models, solver-driven projections, audit-lines on every digit. When the board asks "how did you get this number?", you walk the because-line; the receipt is the answer.

For finance, operations, regulated industries, and anywhere "the LLM thinks so" is not a good enough answer.

Read more →

Inputs · real-world capture and vetted history
Under the hood · the platform and the people

Capabilities

What the technology currently does, in plain English. We keep the how for the papers and conversations with partners.

  • Content-addressed signed knowledge Every piece of data arrives signed and content-hashed: identical content always resolves to the same address, and every reference is a verifiable receipt.
  • Because-line provenance Every claim carries a chain of evidence back to the data that produced it. Hallucination becomes the structural absence of a because-line — visible, by construction.
  • Observer-relative trust Trust values are materialised into a time-locked, fractal index the observer slides through at query time, not looked up from a stored table. Trust is always current for the observer's chosen frame.
  • Local-first inference with receipts LLM reasoning constrained by the graph it can read from; every answer traces back to the signed thoughts it drew on.
  • Protocol-native data migration Any relational schema, change-feed, or API surface reduces to a ThoughtSchema and a stemline; data lands as signed thoughts ready for downstream use.
  • Structural regulatory fit Because-line receipts satisfy EU AI Act Article 86 (right-to-explanation) and GDPR Article 22 by construction, not as a post-hoc bolt-on.

Publications

Defensive publications with third-party timestamps, and UK provisional patent drafts ready for assignment at incorporation.

Defensive publications (IP.com, timestamped)

  • IPCOM000277753D — Round 1
  • IPCOM000277762D — Round 2
  • IPCOM000277803D — Round 4 (Content-Addressed Mixture-of-Experts Inference)
  • Round 3 (Infinite-V) — priority preserved; commercial-hold; substrate-agnostic hedge published as Round 3 Addendum
  • Round 5 (Sound Brains) — continuous-signal substrate, first-signal experimental stage

UK provisional patent drafts ready for submission

  • GRAFT-001
  • CANOPY-001
  • DUALFIRE-001
  • COMPLY-001
  • THINKER-001
  • SULLY-001 / FlowSpace
  • ZEROCOPY-001
  • CONSTRAINED-INFERENCE-001
  • MIMIR-001

Early access

If any of this catches your eye, we'd like to talk. Early access arrangements available for design partners, prospective customers, academic collaborators, and aligned open-source ecosystems.

Want to know more?

Go deeper into any of the five product surfaces: wot.is · wot.was · wot.now · wot.if · wot.then. Or the platform underneath, FlowSpace, or the consulting arm, ThoughtSpace.

WoT Technology Ltd — UK SME, Companies House 17160992.