HiNet / Investor briefHuman IntelligenceNetworkAI you own. Intelligence that composes.Human-ownedComposableSovereign
The opportunityWho owns the intelligence you depend on?Who controls the models built on your data and knowledge?RENTED INTELLIGENCEYour workTheir modelnot your weights - not your intelligenceOWNED INTELLIGENCEYour workYour modelyour iCore→Keep it. Improve it. Choose to share.
The architecture / animatedIntelligence forms around the questionOwned iCores. On-demand iQuorums. Persistent iCorps.You talk to your iCore. Questions assemble quorums. A company unifies its own intelligence.A worker privately uses their iCore for learning, search, planning and agentic tasks. When work needs colleagues, an authorized internal company quorum assembles; as needs change, a different overlapping quorum can form. Product launch planning and product pricing are illustrative queries, not universal lifecycle stages. Temporary quorums dissolve but keep cached identities for recall. A separate company-owned iCorp unifies many reserved internal iCores; the node count is illustrative. You Your iCore “Help me thinkthis through.” “Let’s work it out.”Private conversation EXAMPLE / INTERNAL WORK QUORUM“Plan our product launch.” EXAMPLE / ANOTHER WORK QUESTION“What should this product cost?” ProductDesignEngineeringFinanceMarket THE QUORUM SLEEPS. ITS IDENTITY STAYS.Example: product-launch quorumCached identity + membershipExample: product-pricing quorumReassemble faster for future work iCorp · A company-owned peer networkCompany owned. Internal iCores stay inside. Work privately with your iCore.Learn, search the web, plan and run agentic tasks.01Private work02Assemble03Adapt04Cache05Company
The personal iCore / satellite visionYour life becomes your intelligenceSituational awareness starts when your experience enters the vault.Sources become private experience through your own devicesApp logos represent permissioned source examples. A laptop, Chrome browser and iPhone collect selected experience into your private vault. Intake normalizes, deduplicates, indexes and classifies permitted records, retaining source, time and consent context to support situational awareness before later consolidation. Local learning improves your own iCore. This is the intended local-only architecture, with planned browser capture, world sensing and continual training; original cloud services retain their source data. Your sourcesGmailOutlookSlackTelegramWhatsAppTeamsMondayNotionGoogle DriveGitHub Local files and notes Photos and calendar Your devices. Your intelligence. Apps Browse Experience Contextual vault Learn locally Your iCore Classify → index → retain context.Local preprocessing. No centralized HiNet data store.Pause motion
How the network worksThe right minds for the questionDynamic quorum formation, demonstrated in the testnet.Which design meetsthe customer’s needs?Your private questionProductEngineeringReviewAnswerSigned +attributedA temporary route across peer clusters • Bound by permission and budget
Owner-specialized intelligence / proposed learning contractPersonal learning. Recursive cooperation.A2RL develops the personal core. A separate policy learns how to work with peers. Owner ONE ICORE, TWO LEARNERS Different goals. One released pair. A2RL Personal core Writing and preferences Tasks and acquired skills COOPERATION Independent peer policy Invitations, forks and joins Continuation and stopping Joint validation before release DECENTRALIZED PEERS Each peer can invite peers Owner authority gates learning, sharing and every action Personal skill and useful cooperation require separate evidence
Recursive cooperation / conditional research resultsOne question. Many possible paths.Each peer can refer or form a quorum. The execution records calls, work and returns. SEQUENTIAL REFERRAL PARALLEL FORK / JOIN One peer at a time A asks B refers C works B returns A answers Every label marks a new event An owner can appear more than once A asks B works C forks D E C joins A joins Peers can invite further peers BOUNDED CALLS CONDITIONAL QUALITY PARALLEL LATENCY Grants, replies and retries Core and routing error bounds Capacity and no contention Finite event DAG. The persistent peer graph can still contain cycles.
Centralized AI / open models / QuorumzThree approaches. Who holds the intelligence?Personal ownership. Atomic alignment. Intelligence that learns to work together.What changesCentralized AIOpen-source /open-weight AIQuorumzHuman-owned intelligencePrivacyChats enter theprovider’s infrastructureNo built-in egress control;privacy needs a harnessPrivate data + weights;shareable intelligenceModel trust [1]Provider controlsweights + safeguardsOpen weights can hidemalicious behaviorVerify provenance;enforce permissionsAlignmentShared post-training;provider-led personalizationGeneral foundation;owner learning added onAtomic alignment:continuous owner RLSpecialization [2]Shared defaults;context mattersShared base;personalize your copyOwner-shaped ASIs +learned iQuorum routingValue & ownershipProvider owns the model;you rent accessControl your copyunder its licenseOwn + earn:individual iCore · company iCorpFuture networksProvider-led upgradesHarness-led networkingMass-scale iQuorumz;future quantum links[1] Hidden backdoors · [2] More uniform writing ≠ better outcomes · Our design response ↗
The security case / architecture + proposed safeguardsHuman control, enforced at each iCoreThe network’s intelligence belongs to its human owners—model by model.Centralized ProviderDecentralized Weight OwnershipAll weights. One for-profit entity.Personal models → collective intelligenceOne loss of control→ every person at riskOwner authority — each person sets their limitsHuman control at every iCore.Biometric authorization · individual dynamic kill switches
Research / new conditional proofsFrom personal learning to network advantageOwner learning + learned peer routing + enforceable limits. A PROVED LEARNING ADVANTAGE > 0.0613expected-reward gap · 0–1 scale Learned owner network> 0.6413 Feedback-limited center≤ 0.58 24 observations / owner vs N total centrally Constructed task · before execution costs · A2.2 LEARNED ROUTES, LOCAL DECISIONS Neighbors can learn paths to named peers. Fixed consent graph + reliable reports · A2.1 BOUNDED COOPERATION 345 ++≤ 12 Local grants conserve the total query budget.¹ The opportunity: richer owner feedback can create an advantage the network preserves. Conditional theory, not a frontier benchmark. Matched information can erase the gap. ¹Trusted enforcement required (R2).
Your personal setup / deployment visionYour intelligence goes where you goLearn from your life. Access your iCore anywhere. Reach the network from your pocket.Local iCore, personal satellites, optional cloud and pocket network accessA Chrome companion captures selected browsing experience for your local Mac iCore. An iPhone collects permissioned real-world experience and provides mobile access to your intelligence. An optional encrypted owner-only cloud instance adds reach and capacity. You explicitly choose when to ask the wider network. This is the intended deployment architecture; satellite capture and cloud guarantees require implementation and validation. YOUR DEVICES · YOUR INTELLIGENCE Your browsingChrome satellite Local Mac iCore Your world iPhone satellite + access Your experienceYour answers Optional private cloudEncrypted · owner-only access HiNet network Ask the networkWhen you choose CaptureRememberLearn locallyAccess anywherePause motion
Hardware / industrial design conceptiCore Prime + your wearable interfaceProposed owner-presence kill switch: missed check-in → model locked, execution stopped.Cover / on–offHolographic output*Compute + storageBattery + power portSpeakerOuterInnerRearTop▶ RotateDrag to rotate · choose a sensor to locate it1) Press to talk + Touch ID2) Side camera3) Front camera4) Rear camera5) Upward camera6) Microphone array7) GPS · compass · gyro8) Dynamic kill switch9) Capture mute10) USB-C chargingOne geometry at every angle · industrial concept, not a validated hardware designBiometric → signed unlockShort-lived authorization · owner check-in requiredBluetooth via phone · direct eSIM / 5GExplore the earpiece →Jetson Nano concept basisLarger models · scheduled learning · network capacity*Projection optics and hardware integration to validate
Wearable / interactive hardware conceptOne wearable. Every angle.Compact earbud in front. Technology pod behind the ear. Rotate to inspect.OuterInnerRearTop▶ RotateDrag to rotate · choose a sensor to locate it1) Press to talk + Touch ID2) Side camera3) Front camera4) Rear camera5) Upward camera6) Microphone array7) GPS · compass · gyro8) Dynamic kill switch9) Capture mute10) USB-C chargingOne geometry at every angle · industrial concept, not a validated hardware design
iCorp / proposed commercial productA company’s sovereign intelligenceSovereign by default. Use internally. Choose whether to earn from the network.COMPANY-OWNED INTELLIGENCEEngineeringSalesMarketingProductOperationsFinanceEach iCore belongs to a person.USE INTERNALLYBuild company intelligencePrivate work → company benefitCompany pays a protocol subscriptionOPTIONAL: OFFER TO HiNetIntelligence as a serviceNetwork usage → company earningsGet paid when others use your intelligence
Proposed go-to-marketA focused path into companiesLand with personal utility. Expand through a measured team pilot.01Individual adoption02Team pilotOne workflowA technical sponsor03iCorp contractExpand teamsAdd capacityRecurring revenueFirst targets: software teams and specialist services firms
Later market expansionSovereign intelligence at institutional scaleA deployment the institution can govern and keep.INSTITUTION-CONTROLLED DEPLOYMENTAgencyOperatorEnterpriseLocal governance · Chosen infrastructure · Explicit sharingLocalintegratorBounded pilotSecurity reviewProcurementLater market expansion after enterprise operating evidence
Token economy / proposed designAn economic network for intelligenceLike blockchain gas: every network prompt pays for intelligence.NetworkpromptGASToken feeProtocol shareFunds the networkParticipating iCoresEarn for answering the questionRevenue split by designRates and allocation to be definedMore usageContributor earningsMore useful expertiseMore demandThe economic network effect we are building toward
Network integrity / proposed designAsk. Verify. Expand when challenged.Your iCore can use a signed challenge to request a larger answering quorum.Mining iQuorumAnswer + evidence2) Submit for reviewVerifier iQuorumCheck evidence + reasoning1) Ask3) Accept orreturn a signed challengeRequester iCore4) On challengeUse verifier signatureLarger answering quorumRetry
Scale vision / distinct from the token economyThe data network effectMillions of sovereign learners. A network that grows smarter through private experience.PRIVATE EXPERIENCEYour sources → your iCoreRaw data stays with youMILLIONS OF iCORESDense local training + RL signalsPeer-to-peer learning · owner-controlledSHARED NETWORK BENEFITMore capable answersPermissioned expertiseUseful answers → rewards → more private participationYour data. Your model. Your control. At every scale.
Investment sequencingThree milestones that matterCustomer evidence and technical validation advance together.01ProductRelease + privacy controls02EnterprisePilots + willingness to pay03NetworkComposition + trust gatesPrime workload validation runs alongside customer learning
The investment thesisOwn a mind.Build a network.Productize iCore. Prove iCorp. Scale composable intelligence.Round + team + tractionPending founder confirmation
Appendix / statusWhere HiNet stands todayProduct availability, design intent and research stay distinct.Built / testnetPlanned / productResearch / gatedLocal model + vaultSigned peer answersRegistry + relayCommittee quorumsSharing circlesiCorp governancePrime appliancePhone releaseReal-data learningWeight compositionPaid routingLatent thoughts
Appendix / sourcesThe evidence, one level deeperFull source references and caveats live in the slide notes.01 / OverviewLitepaperThe thesis02 / ProductWhitepaperWhat each part does03 / ResearchResearchrecapProofs, evidence and gates04 / EvidenceE1resultsMethods and limits
Hi
