Research inputs
Imperial thesis (2025 preprint) · published network physics · open market data · jurisdictional rule-sets compared in shadow mode.
INDICATIVE PROTOTYPE · figures pending NUMBERS v2.2 sign-off · internal — not for external release
Method · why you can trust the number
A 100% independent layer — deterministic where it counts, AI only at the edges.
Today's market clears electricity like a commodity — one equilibrium price, cleared in periodic auctions. As the system fills with weather-driven, zero-marginal-cost renewables, that design loses its stable footing and leans ever harder on price caps, uplift and intervention. A first-principles Imperial thesis underpins our engine: it reframes price as a bounded control signal computed from the physical state of the grid — not an emergent scarcity spike — so balance, fairness and cost recovery can be designed in, then tested before anything is locked in.
100% independent
Owned by none of the players — so it can audit value across all of them.
Deterministic core
The fairness, settlement & adequacy maths are auditable and reproducible — a first-principles design, not a black box.
AI at the fringe
AI assists exploration + workflow — it never makes the fairness decision.
Confidential compute
The engine itself is transparent and auditable; a hardware enclave (currently AWS Nitro) lets partners run it on sensitive data without exposing the raw data — substrate-agnostic by design.
Audited attribution
Value mapped, fairly distributed (Shapley) and stress-tested — by an independent party.
Settlement architecture
internal use only · gated — FCA counsel pendingout of scope for this open methods view · governed separately · FCA counsel pending
Core · deterministicInside the
deterministic core
deterministic core
Scoping question
Can the fairness, balance and cost-recovery maths of a power market be written down as a first-principles design — auditable end-to-end, reproducible by an independent party?
Method
Three principles, applied as one rule: control-signal pricing, holarchic architecture, and fairness by construction. Nodal and zonal pricing fall out as special cases. Prevents bias in execution and data-in-use through the enclave + reproducible-build pipeline (in development).
Control-signal pricing
Prices are bounded, stateful control signals read off physical tightness — congestion, reliability margins — not unbounded scarcity spikes. Their job is to hold the system in balance, now and forward.
Holarchic architecture
The grid is modelled as nested layers — device, feeder, zone, system — and each prices from the tightest binding constraint above it. Nodal and zonal pricing fall out as special cases of one rule.
Fairness by construction
Fairness is computed inside the clearing — who is served, and who pays — not bolted on afterward through caps and uplift. Value is attributed by what each participant delivers at the right time, place and reliability.
Fairness as a computable property — not a political afterthought.
Attribute
by time, place & reliability
Distribute fairly
nested Shapley · stress-tested
Audit
independent 3rd party
Regulator · network
Compare rule-sets before any single one is locked in
The thesis proposes one market design. Enleashed productises the engine and the thing around it: a neutral sandbox where regulators, networks and retailers compare many rule-sets — including this one — against real network physics and fairness, before any single design is locked in. Shadow-mode, white-label, on open data.
Shadow-mode · white-label · open data
Retailer · operator
Attest, then release
The engine runs inside a hardware enclave that cryptographically proves which code is running before any data key is released — a partner can compute on sensitive operational data without exposing it, and the operator (us, and the cloud provider) cannot read it in use. This is enleashed's own privacy layer, not part of the thesis. Runs on AWS Nitro Enclaves today; Azure Confidential Computing and GCP Confidential Space treated as equal alternates.
Substrate-agnostic · multi-substrate in progress
Honest limits · the reproducible-build pipeline that lets a third party independently verify which code ran is in development. Confidential compute shrinks the trust base — it does not remove the cloud provider from it, and attestation proves which code ran, not that it is side-channel-proof.
Owned by none of the players — so we can audit value across all of them, including the hyperscalers funding the data-centre build-out. The engine treats an AWS-, Microsoft- or Google-operated site on exactly the same terms as anyone else's: if a player looks good, it is a result the audit produced, never an input.
Energy Security Week
The electricity market does not work — a thesis about how we can fix it
On the absence of a programmable market machine — and the fix.
read →arXiv · preprint
A Fair, Flexible, Zero-Waste Digital Electricity Market (Imperial, 2025)
The first-principles thesis behind the engine. Preprint — not yet peer-reviewed.
read →Draft — thesis-grounded copy pending two-signature sign-off (Shaun + Blake) before public release.