EC COKE.CAPITAL

Independent systems research · London

Systems that answer to evidence.

Secure AI, cryptographic trust, privacy infrastructure and deterministic worlds — built from first principles.

Enter the proof atlas
ThesisAcross every system

Capability is easy to claim.
Authority is harder to earn.

The recurring question across this body of work is not “How powerful can it be?” It is: what can it prove, what can it touch, what survives failure — and who can leave?

AuthorityBounded
EvidencePortable
StateReplayable
FailureVisible
Selected systemsThree recurring problems

Proof, privacy
and playable worlds.

A blind-embossed DealSeal mark beside a fountain pen and a printed hash DealSeal · portable signed evidence
Trust / 01

Cryptographic trust · Identity continuity

A receipt for what was actually agreed.

DealSeal turns deals, predictions and promises made on X into signed records of who said what, which rule was fixed, and what evidence later showed.

What exists

Private deal receipts, opt-in X Chat key continuity and no-money public claim records. Canonical terms and off-chain wallet signatures; no custody or settlement.

Open DealSeal ↗
Evidence / 02

Secure AI · Market systems

Intelligence without a master switch.

Vega separates forecast, paper execution, real-fill policy, host, recovery and AI commentary authority. One signal cannot quietly authorize the next system.

VEGAEvidence core
ForecastProduces scored evidence
PaperSimulates execution
PolicyGrants no approval by default
HostOwns deployed lineage
RecoveryProves return to safety
AIExplains; never enforces

Defensible claim Evidence and deployment discipline — not a new ML algorithm, guaranteed edge or autonomous trading approval.

Guarded research system
WarGrid real-time strategy game showing vehicles moving through a forested map WarGrid · deterministic movement build
Worlds / 03

Game engines · Deterministic simulation

A world should tell the same story twice.

OpenRTS treats rules, content, commands and randomness as explicit contracts. WarGrid is the playable proving ground: economy, construction, movement, combat, replay and multiplayer state all meet at a canonical tick boundary.

Boundary

The game and engine are separate bodies of evidence. A playable WarGrid release does not, by itself, prove every OpenRTS milestone.

Play WarGrid ↗
Built surfacesInterfaces are part of the proof

Systems you can see,
not just describe.

NOSE minimum-disclosure marketplace prototype
NOSEMinimum-disclosure market
LocalForge private local coding interface
LocalForgeBound to this Mac
AnonTalk private audio matching interface
AnonTalkAnonymous voice matching
Full field indexImplemented · audited · researched

The wider body
of work.

Every entry names its evidence level. “Research” is not passed off as production; a historical release is not treated as current uptime.

MethodA repeatable operating discipline

Build for the
hostile reading.

The strongest systems in this portfolio share a method: map authority before capability, make claims answerable to evidence, and keep a way back.

  1. Map

    Find the real authority.

    Separate source, runtime, operator, model and recovery truth before making a claim.

  2. Constrain

    Give every actor less power.

    Models propose. Deterministic policy decides. Consequential actions remain bounded and visible.

  3. Prove

    Test the failure path.

    Replay, reconcile, restore and inspect. A green screen is not a recovery story.

  4. Release

    Ship with a route back.

    Immutable releases, exact checks and rollback state turn deployment into evidence.

Independent systems research

Build the layer
that must be true.

Talk on X ↗ Return to atlas ↑