Return to They · The product’s working life They · Products and machine intelligence The product’s identity, configuration and working life. Connect its upstream suppliers and downstream users and service partners; add authorised agents where useful. A design in development. No agent executes and no protocol permission or infrastructure service is created here. Existing app foundations These are uses of shared capabilities. Me follows the individual; We follows the organisation’s value chain; They follows the product and its supply chain. A passive product can have a twin without an agent. Agentz owns conversation and assistance; APU organises the stored outline. Personal notes and tasks exist in source. That does not establish a shared business task workspace or an autonomous product agent. We authors products and manages variants, stock and offers. They follows the product, its components, individual units and service history. Me compares and uses an owned unit. Markets explains the commercial terms; the physical-market order exercise is a simulation. OFN owns the independent wallet and economic-receipt standard. Personal, company and permitted product-related records remain distinct. Display currency preferences are not currency issuance; interfaces are not proof of live transfers. Help, onboarding and supported platforms serve the whole experience. Tips means educational help, not financial tipping. Retained advertising code needs a disposition, not a promise that advertising is live. The reviewed Flutter source contains 42 feature directories at revision Inspect the pinned Flutter source The seven products contribute across all three views. APU organises the outline and stored records; Agentz provides conversation and contextual assistance. OFN remains an independent wallet and receipt standard. Markets brings the participants, offers and commitments together.Products, their twins and authorised agents
The tools in this perspective.
Conversation, notes and tasks
Products, configurations and individual units
Wallet views and economic records
Onboarding, help and platform support
Source inventory and limits
bb34f4c54733. Source presence is not a released or accepted journey. Product autonomy, end-to-end manufacturing and connected service still need qualification.ai_chat, chat, rag_chat, chat_tasksproducts, servitization, marketplacewallet, receipts, currencytips, walk_through, web, advertisement
Products and machine intelligence · They
Follow the product from its upstream components and suppliers to its downstream owners, operators and service partners. Its model, configuration, individual unit and working history remain distinct. This includes ordinary products as well as embodied intelligence.
Products remain understandable, maintainable and useful throughout their working lives, with dependable connections to the people, organisations and resources they depend on.
Connect product identity, configuration, evidence and service history across the supply chain. Where an agent is useful, let it inspect needs and propose bounded work under the responsible person’s or organisation’s permissions.
A product is not its twin, and a twin need not have an agent. Neither implies a wallet or spending authority. Machine APIs encounter the same versioned offers, agreements, evidence and permission limits as human interfaces. Technical accounts and approved policies can support routine activity without a fresh human screen for every operation. Binding external commitments still identify the applicable principal and authority.
Objectives
The product perspective connects identity and dependencies to use, maintenance and renewal. Machine participation is one qualified extension of that story.
Distinguish the model, configured variant and individual unit. Retain source, revision, ownership and operating context without inferring authority from a record. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Circle. Compare attributable product specifications, compatibility claims, performance records and corrections without treating repetition as independent evidence. Evaluation role: Circle evidence owner · Unappointed. Correctly distinguished model, configuration and unit records with expected revision/provenance divided by all sampled product records; report missing identity links separately. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Product-record evaluator · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
Connect components, suppliers, software, energy and required services. Unknown compatibility stays unknown; a similar part is not automatically a substitute. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Circle. Compare attributable product specifications, compatibility claims, performance records and corrections without treating repetition as independent evidence. Evaluation role: Circle evidence owner · Unappointed. RealityFutures. Compare repair, replacement, upgrade and provider scenarios where cost, completion time or reliability is uncertain. Evaluation role: RealityFutures evaluation owner · Unappointed. Required dependency and compatibility questions with attributable current support or explicit unresolved status divided by all required questions; do not count unresolved as compatible. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Product dependency reviewer · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
Link the agreed configuration to production or existing stock, delivery, installation and acceptance. Keep each event separate. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Cortex. Preserve sources, revisions, dependencies and unknowns across the product’s upstream and downstream supply chains. Evaluation role: Cortex record-evaluation owner · Unappointed. LIQUID. Compare exact offers, price units and payment terms and reconcile exchange effects. Evaluation role: LIQUID reconciliation owner · Unappointed. Expected due production/delivery/installation/acceptance/service events linked to the exact configuration or unit and their independently attributable evidence divided by expected due events; retain absent and disputed events. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Lifecycle evidence evaluator · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
Identify owners, operators, service partners and downstream products using this component. Share only the information each relationship permits. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Agentz. Provide product-context conversation and assistance; coordinate permitted diagnosis, service proposals, Cortex memory, handoff and recovery. Evaluation role: Agentz evaluation owner · Unappointed. Reviewed relationship disclosures preserving the declared parties, purpose, grant and expiry divided by all eligible disclosures; report blocked and actual excess disclosures separately. This evaluates scope, not trust or relationship value. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Relationship-context evaluator · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
Connect fault reports, uncertainty, compatible replacements, service capacity and repair evidence. An authorised agent can assist; it cannot grant itself purchasing or travel permission. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Reliability, degradation, repairability and useful capability require independently evaluated condition criteria, cohort and horizon before measurement. RealityFutures. Compare repair, replacement, upgrade and provider scenarios where cost, completion time or reliability is uncertain. Evaluation role: RealityFutures evaluation owner · Unappointed. Cortex. Preserve sources, revisions, dependencies and unknowns across the product’s upstream and downstream supply chains. Evaluation role: Cortex record-evaluation owner · Unappointed. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Preserve service commitments, usage and outcome evidence. OFN economic records and LIQUID exchange records keep their own rules; a machine log is not automatically a receipt. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. LIQUID. Compare exact offers, price units and payment terms and reconcile exchange effects. Evaluation role: LIQUID reconciliation owner · Unappointed. Cortex. Preserve sources, revisions, dependencies and unknowns across the product’s upstream and downstream supply chains. Evaluation role: Cortex record-evaluation owner · Unappointed. Expected due production/delivery/installation/acceptance/service events linked to the exact configuration or unit and their independently attributable evidence divided by expected due events; retain absent and disputed events. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Lifecycle evidence evaluator · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
Review cost, benefit and applicable terms. Preserve relevant product history while protecting personal data. Replace software, workers or providers only through qualified transitions. Contribution measures only. Each retains its declared eligible population and limits; the wider objective is not established by a successful record or review. Existing machine-service and agent-handoff protocols apply only to that subset. Passive product condition and lifecycle need distinct product-class protocols. Agentz. Provide product-context conversation and assistance; coordinate permitted diagnosis, service proposals, Cortex memory, handoff and recovery. Evaluation role: Agentz evaluation owner · Unappointed. MurMur. Supply a principal’s permitted capital goals, reserves and constraints when machine work affects that principal’s capital. Evaluation role: MurMur mandate-review owner · Unappointed. Due permitted recovery/replacement/transfer/retirement exercises meeting every pinned required criterion divided by all due exercises; retain failures, missing keys, obligations and unresolved records. Different transition classes remain separate. A predeclared dated evaluation batch, with sample, versions, criteria and consent recorded before review. Evidence: Pinned original and successor records, effective permissions, declared task criteria and independent outcome review. Independent transition-outcome evaluator · Unappointed. Proposed evaluation definition; collection not connected. Targets: Unset. Observations: Not recorded. Product definitions below retain their own populations; website tests do not measure customer outcomes.How this outcome would be evaluated
Additional outcome definition
In the fictional workshop case, a procurement worker discovers an exact offer, checks evidence and stages a proposal. Try missing compatibility, limited resources, interruption, a stale or revoked grant and a replacement. The same versioned record supplies the human views and structured JSON below.
A distinct fictional AX-14 has documented fit in this example. LA-02 remains unverified; a low price or newer model supplies no missing compatibility. Procurement, the repaired robot and a downstream customer service each need their own acceptance.
Fictional local demonstration
Workcell 17 · Case revision 1
A distinct fictional AX-14 revision C has documented fit in this fixture. Follow review, delivery, acceptance and payment. Starting an example replaces the current local simulation.
The procurement agent reads the same versioned case shown here. It can compare, clarify and prepare a proposal. Its identity and model version grant no power to approve the agreement or its payment.
Compare the exact actuator records, stage a proposal, and inspect what was used. Model weights, the running instance, the technical account and the robot are different things. Copying weights copies none of their grants, commitments or resources.
Illustrative rate: USD 0.20 per 100 synthetic units. These are fixed teaching values, not provider pricing or measured inference. Compute, storage, energy, connectivity and maintenance require their own qualified providers and service conditions.
Rejecting a goods quote leaves authorised compute usage billable. The goods agreement, customer acceptance and settlement stay separate. Nothing in this simulation pays either bill.
The fictional principal’s policy PRINCIPAL-RESOURCE-POLICY-1-demo permits routine bounded work without a fresh human screen for each operation. A stale or revoked task grant blocks future work. Supplier and principal obligations already recorded are preserved.
Replacement changes the model, worker instance and provider under the same technical account and explicit grant. A new key or account would require its own grant; no key custody or account migration is demonstrated.
Candidate: qualified. Qualification is a local fixture, not provider admission.
Current state
Nora reports a failed actuator. The workshop defines the repair need, permitted context and spending limit.
Fictional all-in quotation: one actuator, shipping and a synthetic tax amount. No platform fee or recurring service is selected. These are fixture terms, not Skyzai fee or tax policy.
This fixture pays only after delivery, successful inspection and explicit acceptance. Other contracts can require another sequence.
Duplicate event identities and stale page revisions do not add effects. This demonstrates local bookkeeping, not a production retry or payment service.
Replace the reading agent without changing the grant. Change a price or spending limit to see why a new proposal and authorisation are needed. A correction preserves the historical agreement and changes what can happen next.
Fictional all-in quotation: one actuator, shipping and a synthetic tax amount. No platform fee or recurring service is selected. These are fixture terms, not Skyzai fee or tax policy.
This fixture pays only after delivery, successful inspection and explicit acceptance. Other contracts can require another sequence.
This fixture’s approval stands for an applicable external purchasing policy. It does not impose one signer, multiple signers or a default threshold on real organisations.
Maintenance need and purpose-limited purchasing brief recorded.
Fictional record · needThe same case, versions, evidence and permissions rendered as JSON. Switching views changes presentation only; it never grants a machine additional authority. This is an inspectable local record, not a connected API.
{
"id": "WORKCELL-17-REPAIR-DEMO",
"schemaVersion": "actuator-case/1",
"revision": 1,
"simulatedDate": "2026-10-21",
"scenario": "success",
"stage": "need",
"status": "Nora reports a failed actuator. The workshop defines the repair need, permitted context and spending limit.",
"brief": {
"id": "BRIEF-WORKCELL-17",
"version": 1,
"objective": "Restore the fictional workcell to its documented motion test.",
"quantity": 1,
"spendCapCents": 30000,
"requirements": "One AX-14 revision C compatible replacement for the XR-7 revision D motion axis; fit and commissioning checks must be recorded.",
"disclosure": "Equipment model, revision, fault and purchasing constraints only. No worker identities or production logs.",
"principalId": "WORKSHOP-DEMO",
"policyId": "PURCHASING-POLICY-01-demo"
},
"asset": {
"instanceId": "ROBOT-CELL17-01",
"model": "Workshop Robot XR-7",
"modelRevision": "D-demo",
"installedPart": "Failed AX-14 revision B · instance AX14-B-0042"
},
"twin": {
"id": "TWIN-ROBOT-CELL17-01",
"version": 7,
"represents": "ROBOT-CELL17-01"
},
"machine": {
"model": "offer-reader-weights/1-demo",
"instance": "RUN-PRIMARY-01",
"provider": "PROVIDER-A-DEMO",
"account": "PROCUREMENT-ACCOUNT-DEMO",
"key": "PUBLIC-KEY-FINGERPRINT-DEMO",
"operator": "WORKSHOP-DEMO",
"configuration": "equipment-only/1-demo",
"hardware": "CPU-host-demo",
"license": "task-permitted-demo",
"grant": {
"id": "READ-COMPARE-DRAFT-01",
"version": 1,
"status": "current",
"account": "PROCUREMENT-ACCOUNT-DEMO",
"context": "Equipment, manual and exact offer only; no contacts or unrelated records.",
"policy": "PRINCIPAL-RESOURCE-POLICY-1-demo"
},
"run": "ready",
"replacement": "qualified",
"qualification": {
"task": "Passed the fictional offer-reading fixture; no operating benchmark.",
"cost": "Same USD 0.20 per 100 synthetic units; total run cap USD 1.00.",
"hardware": "Compatible with fixture CPU host.",
"license": "Fictional license permits this task.",
"security": "Same tool allowlist and equipment-only view.",
"reproducibility": "Pinned demo weights, tool v1 and configuration v1."
},
"limits": {
"units": 500,
"costCents": 100
},
"usage": [],
"review": "not-reviewed",
"notice": "An approved fictional resource policy permits routine read/compare/draft operations within these limits, without a fresh human screen each time. It supplies no purchasing authority."
},
"offer": {
"id": "MAKER-AX14-C-DEMO",
"version": 1,
"productModel": "AX-14 Linear Actuator",
"productRevision": "C-demo",
"makerId": "Fictional Workshop Motion Works",
"termsVersion": "AX14-TERMS-1-demo",
"validUntil": "2026-11-21",
"fit": "documented-in-fixture",
"stock": "fresh-in-fixture",
"itemCents": 23000,
"shippingCents": 1500,
"taxCents": 2450,
"platformFeeCents": 0,
"serviceCents": 0,
"totalCents": 26950,
"allInKnown": true,
"currency": "USD",
"paymentTerms": "This fixture pays only after delivery, successful inspection and explicit acceptance. Other contracts can require another sequence.",
"scope": "Fictional all-in quotation: one actuator, shipping and a synthetic tax amount. No platform fee or recurring service is selected. These are fixture terms, not Skyzai fee or tax policy."
},
"agent": {
"id": "AGENT-PROCUREMENT-DEMO",
"version": "1-demo",
"toolVersion": "offer-reader/1-demo",
"grantId": "READ-COMPARE-DRAFT-01",
"allowedActions": [
"read",
"compare",
"clarify",
"stage_unsigned_proposal"
],
"mayAuthorize": false,
"maySign": false
},
"fitEvidenceVersion": 1,
"proposal": null,
"authority": null,
"commitment": null,
"delivery": "not-started",
"inspection": "not-recorded",
"acceptance": "not-accepted",
"payment": "not-paid",
"reconciliation": "not-reconciled",
"correction": "none",
"acceptanceScope": {
"acceptedChecks": 0,
"requiredChecks": 2,
"unresolved": "Installation and motion checks await inspection and explicit customer acceptance."
},
"invoice": null,
"processedEvents": [],
"history": [
{
"revision": 1,
"action": "need",
"actor": "Nora and workshop · fictional",
"detail": "Maintenance need and purpose-limited purchasing brief recorded."
}
],
"evidence": [
{
"id": "EVIDENCE-1",
"revision": 1,
"source": "Maintenance note · fictional",
"detail": "XR-7 revision D, installed instance ROBOT-CELL17-01, failed AX14-B-0042. No real equipment inspected.",
"status": "fictional-record"
}
]
}All identities, evidence, agreements and payments here are fictional. Nothing is submitted or issued; no money moves. State lives only in this page and resets on reload. No OFN receipts are issued.
Explore product twins · Explore source corrections · Return to Markets within We
Contributing product strategies
Use the capabilities a task calls for. Each product owns its strategy, proposed measures and specialist Agents. The Rosetta correspondence connects its field of inquiry to skills and responsibilities.
Rosetta agent row · Teleology
Organise product specifications, revisions, dependencies and service records in a navigable outline, with permissions preserved. Proposed KPI. Record structure, permitted retrieval, revision/undo and export fidelity in declared audience-specific samples. Decision-review comprehension belongs to Agentz and is measured separately. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Information organiser; Retrieval designer; Record integrity reviewer. Skills and responsibilities. Make product records, revisions and dependencies navigable for review. Products with authorised agents can also expose bounded machine interfaces. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. The dated inventory allocated conversation to the APU interface; its evidence IDs are retained for traceability. Current responsibility is APU’s nested outline and stored records, while Agentz owns conversation and contextual assistance. These source contracts do not establish a released integration. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. APU owner: qualify outline navigation, stored-record permissions, search, revisions and recovery. Qualify conversational handoff separately with Agentz; package integration remains deferred.Proposed measures and specialist Agents
How this would be measured
Outline structure fidelity
Permitted record retrieval
Revision and undo fidelity
Portable record export
Explore APU’s AgentsSources and integration limits
Rosetta agent row · Epistemology
Compare attributable product specifications, compatibility claims, performance records and corrections without treating repetition as independent evidence. Proposed KPI. Traceable-source coverage among claims in eligible proposals, with stale, inaccessible and unresolved claims reported separately; accepted corrections reflected in affected fresh answers. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Source researcher; Corrections editor. Skills and responsibilities. Research and correct capability evidence. Repeated model output is not independent corroboration. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. The Flutter case is host/adapter context. Its presence and the website correction example do not establish an integrated Circle service. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. Circle owner: qualify attributed source retrieval, independent comparison and correction propagation with genuine permitted sources.Proposed measures and specialist Agents
How this would be measured
Traceable claims
Correction propagation time
Explore Circle’s AgentsSources and integration limits
Rosetta agent row · Methodology
Compare repair, replacement, upgrade and provider scenarios where cost, completion time or reliability is uncertain. Proposed KPI. Forecast error by resolved task class against declared baseline and horizon; report resolution coverage, abstentions and unresolved outcomes. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Scenario analyst; Forecast researcher. Skills and responsibilities. Compare uncertain execution paths. A prediction grants no resource-acquisition authority. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. The Flutter case is host context. A market response or fictional comparison does not establish calibrated forecasting or an accepted resolution lifecycle. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. RealityFutures owner: qualify question, assumptions, baseline, horizon, resolution and abstention on the same versioned records.Proposed measures and specialist Agents
How this would be measured
Binary forecast Brier score
Continuous forecast mean absolute error
Resolution coverage
Explore RealityFutures’s AgentsSources and integration limits
Rosetta agent row · Axiology
Provide product-context conversation and assistance; coordinate permitted diagnosis, service proposals, Cortex memory, handoff and recovery. Proposed KPI. Tested eligible handoffs retaining the correct allowed context, versions and pinned commitments divided by tested eligible handoffs; scope violations, stale context, latency and resource costs separately. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Work coordinator; Memory steward; Context curator. Skills and responsibilities. Coordinate execution and continuity. A replacement worker inherits only the explicitly permitted context. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. Source and bounded proofs retain their original limits. Agentz uses Cortex memory and shared Nexus disclosure permissions; website state is not durable or signed-in continuity. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. Agentz owner: accept effective grants, retained records, versioned handoff, provider refusal and recovery in the signed-in host.Proposed measures and specialist Agents
How this would be measured
Permitted handoff fidelity
Actual scope violations
Blocked out-of-scope attempts
Handoff latency
Recorded resource cost
Machine-work review comprehension
Explore Agentz’s AgentsSources and integration limits
Rosetta agent row · Cosmology
Preserve sources, revisions, dependencies and unknowns across the product’s upstream and downstream supply chains. Proposed KPI. Permitted retrieval fidelity, source traceability, revision lineage fidelity and proposal-state fidelity in a declared sample of exact configurations, units and lifecycle events; include refusals, conflicts, stale records and missing sources. Targets remain unset. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Memory steward; Evidence reviewer. Skills and responsibilities. Keep assertions and proposals distinct from accepted records. Nexus scopes disclosure; later assessment or insurance requires separate qualification. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. Marketplace source supplies commercial context only. Menexus integration, independent acceptance and any named insurance provider/policy remain unqualified. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. Menexus owner: qualify commitment, independent review and remedy contracts; qualify insurance separately with its actual provider and terms.Proposed measures and specialist Agents
How this would be measured
Permitted retrieval fidelity
Source traceability
Revision lineage fidelity
Proposal-state fidelity
Explore Cortex’s AgentsSources and integration limits
Rosetta agent row · Ontology
Compare exact offers, price units and payment terms and reconcile exchange effects. Proposed KPI. Unmatched usage, order and payment records within a defined billing period; duplicate financial effects, reversals and missing settlement states separately. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Offer specialist; Reconciliation specialist. Skills and responsibilities. Compare resource offers and reconcile effects. Resource billing and the downstream purchase are separate obligations. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. Pinned commerce/wallet source is not live signing, custody, exchange eligibility, payment, settlement or an issued OFN receipt. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. LIQUID owner: qualify exact offers, applicable authority, effects, retries/reversals and reconciliation under each record’s actual terms.Proposed measures and specialist Agents
How this would be measured
Unmatched usage, order and payment records
Duplicate economic effects
Explore LIQUID’s AgentsSources and integration limits
Rosetta agent row · Theology
Supply a principal’s permitted capital goals, reserves and constraints when machine work affects that principal’s capital. Proposed KPI. Capital-affecting proposals preserving all current mandate constraints divided by reviewed proposals; correctly declined and unresolved proposals reported separately. Collection requires an appointed owner, permitted records and an agreed protocol. Definitions do not start telemetry or appoint anyone. Specialist Agents. Mandate researcher; Diligence analyst. Skills and responsibilities. Apply the principal’s permitted capital context. Principal-side research does not create an autonomous machine treasury. Capability snapshot: 2026-10-17. Each source keeps its own evidence level. These records do not establish a live service. Representative Flutter source. No dedicated Flutter capital-office integration is established by this inventory. A permitted research proposal is not investment performance or treasury authority. Inspect the dated Flutter inventory and its limits. Repository access may be required. Future integration. MurMur owner: qualify a permitted principal mandate, access controls, constraint review and correction before any service or performance claim.Proposed measures and specialist Agents
How this would be measured
Mandate fidelity
Explore MurMur’s AgentsSources and integration limits
These are proposed measures with inspectable collection definitions. Measurement owners remain unappointed, collection has not begun and targets remain unset. An unrecorded result is not a measured zero. Local demonstration. The fictional procurement worker has a bounded task grant and resource account, with refusal, interruption, correction and qualified replacement states. Its work record and goods agreement retain separate obligations. Inspect the separate local example. It uses page-local state and resets on reload; it is separate from the product commissioning-analysis story. The actual app/service must separately qualify machine inputs, effective grants, provider admission, metering, billing and replacement. This page-local case contacts no provider and pays no bill. Read the historical 21 October participant proposal. Repository access may be required. The current objectives above incorporate the 26 October perspective clarification: individuals, organisations, and products with or without an agent. Read the current relationship between the perspectives.This participant’s example and integration limits
Specialist Agents describe roles and skills in development. People, models and tools may contribute within an explicit assignment. A role description establishes neither staffing nor permission to act. The participant group They remains distinct from these specialist roles.
Machine intelligence can discover tasks and contribute capabilities through explicit protocols, without presumed personhood.
The protocol design lets machine accounts exchange signed messages without the account itself holding a bank account or accepting a click-through agreement. The controlling person or organisation remains responsible for any required agreement and legal obligation. Identify the operator, protocol version, permitted inputs, exact actions and result contract; a key proves control, not the right to act.
Explore integration foundationsAn agent uses a key-controlled account to authenticate protocol messages. Its operator defines the permitted context, actions and limits. Control of a key establishes neither legal personhood nor authority to bind someone else.
Model weights encode learned capabilities; a running agent also needs compute, context and tools. Replacing a model must preserve the applicable permissions and commitments, with continuity and task performance checked. Copying weights supplies no energy, hardware or maintenance.
Agentz coordinates bounded work, Cortex memory and scoped connections. APU organises the stored outline; Agentz owns conversation and contextual assistance. Protocols support machine discovery and participation; compute, storage and mesh links supply the underlying capacity.
In Nora’s fictional workshop case, an agent compares actuator specifications and exact offers. Its useful result is an inspectable proposal, including reasons to stop.
The demonstration runs locally. No agent, supplier or payment service is contacted.
Give a worker the context required for a defined task, then inspect what came back.
The example separates the request, its grant, the delivery and a correction. An execution record says what a system observed; independent outcome evidence may still be needed. It does not approve the worker’s own payment.
Inspect capability evidenceFictional workshop request
Brief
Translate the selected maintenance page into French. Keep its source and revision attached. The workshop decides whether the result meets its brief.
Local design demonstration. Nothing is translated, sent, stored or paid. All choices reset on reload. Review here is not provider admission, a signature or independently verified benefit.
Useful machine work depends on compute, storage, energy, connectivity and the people and equipment that maintain them.
Products with machine intelligence are one part of They’s broader product and supply-chain perspective. Machine participation needs qualified physical capacity. Spectre, SKY/ZAI protocol mechanics and Permanent Composability support that proposed participation. Renewing capacity requires maintained hardware, energy and supply chains; copying software alone is not enough. Agentz coordinates permitted tasks for your Cloud Computer; Cortex provides the memory workspace and Nexus scopes shared context across Skyzai. Physical hardware, an operating service and an accepted provider each need their own evidence.
Explore the Cloud ComputerCapacity for inference and bounded execution, with identifiable hardware, supported workloads and measured limits.
Availability, retention and recovery under explicit access rules. Encryption and redundant copies require separate proofs.
Locations, power, cooling and maintenance responsibilities that keep physical capacity working.
SPECTRE · Marketsphere OS
SPECTRE is the proposed open operating system for the Marketsphere, serving individuals, organisations and machine intelligence. Peer coordination, broadcasts, direct messages and configurable feeds are capabilities within it. Nodes, relays and mesh links provide transport.
Opted-in internet connections could provide backhaul between gateways, with explicit consent and capacity limits. Local multi-hop mesh needs separately qualified device-to-device links. Compute execution, storage, application memory and market commitments retain their own owners and contracts.
MESH, RELAY, FLOW and AXIOM define these four patterns. Shared mesh infrastructure can carry all four; Me explains their user experience. Route scoring, feed ranking and ledger agreement have separate contracts. Resilience needs measured operation under loss and recovery.
Explore SpectreIn development. The illustrated connections are simulated.An individual or organisation may operate agents or supporting infrastructure under the relevant permissions and operating requirements. A user need not become a node operator to use an application.
A node record should name its operator, interfaces, capacity, maintenance, availability and usage evidence. Running a node does not confer equity, governance authority or access to another account’s data.
No node is created or admitted through this website.
Spectre, Layer-1 agreement and SKY/ZAI mechanics form distinct parts of the protocol foundation.
The design separates Spectre communication, ledger agreement and monetary rules. Ordinary messages may arrive while ledger agreement is unavailable; delivery alone proves neither finality nor payment. SKY and ZAI retain distinct currency and coordination-capital roles. Monetary actions need their own permissions and accounting; pledged collateral needs exclusive locks. These mechanics do not issue organisational equity or establish investor rights. LIQUID owns the exchange and settlement application. Each layer remains subject to implementation, admission and evidence.
Inspect the protocol architectureSKY is the currency design. ZAI is the coordination-capital design. Their monetary rules, ledger effects, eligibility and usage records are protocol topics in They, separate from a venture’s equity, debt and investor rights.
The selected September research baseline creates SKY with matching debt against exclusively pledged ZAI. Borrowing existing ZAI for short selling is a newer, inventory-backed proposal needing separate lending, margin, repayment and loss rules. It does not mint ZAI and is not required for ordinary communication. Neither a collateral mark nor the interest curve guarantees recovery.
The LIQUID DEX and exchange application supplies the intended order, liquidity and settlement experience. It consumes qualified protocol interfaces; a ledger entry alone does not admit an asset to trading.
Target design. This preview operates no Layer-1 network, issues no SKY or ZAI and guarantees no tradability, liquidity or settlement.
Agent discovery, identity, scoped delegation, execution and usage records need contracts that different implementations can inspect.
A protocol defines how key-controlled accounts exchange and validate messages. Permanent Composability keeps those contracts versioned, compatible and replaceable across Spectre, the seven applications and their hosts. Product owners retain their canonical records, permissions and responsibilities.
Measured execution, provider billing, customer acceptance and payment remain separate records. The organisation’s offers and commitments belong in We. LIQUID owns commerce, DEX orders, liquidity and the settlement application. Machine participants use qualified protocol interfaces beneath those applications.
OFN economic receipts require their own qualified source and issuance rules. An ordinary machine log does not automatically become an OFN receipt.
Explore service terms in MarketsThe seven applications are designed to run on compatible cloud infrastructure, with a path toward Spectre nodes and mesh services.
Permanent Composability keeps application contracts separate from their hosting adapters. Each workload must qualify for its chosen host: supported execution, data location, permission, retention, recovery and measured service requirements remain explicit. Migration to Spectre is a staged target, not a live-network claim or an automatic move of every workload. A mesh provides connectivity; it does not by itself supply qualified compute, storage or confidentiality.
Explore execution and continuityMachine contributions to shared markets · intended experience
People should see understandable screens; agents should receive structured records. Both must encounter the same versioned offers, commitments, evidence and permission limits.
Agents help compare capabilities and exact offers, ask for missing evidence and suggest what to choose—or whether to buy nothing. The responsible person or organisation remains accountable for the agreement. Authorisation must precede any binding acceptance, not just the later payment.
The Marketsphere connects needs, offers and fulfilment across individuals, organisations and machine intelligence. Useful contributions could support continued access to energy, compute, repair and replacement. Demonstrating that cycle requires evidence of the work performed, its conditions, quality and cost.
Explore offers and delivery in MarketsThe physical actuator is a product. Its twin describes its identity, specifications, condition and service history. An agent is software permitted to inspect or act on particular records. A twin can exist without an agent; one agent can serve several products.
A model upgrade does not by itself demonstrate a better service. Specialisation needs observed results under stated conditions, with proposed targets kept separate from measurements.
Inspect the product-twin exampleInfrastructure ownership, equity, debt and investment rights remain in We’s ownership and finance explanation.
SoResFi · organisation, exchange and protocols
The SoResFi design connects three responsibilities. We covers the organisation, issuance, investor rights and asset financing. LIQUID owns the DEX, exchange, orders, liquidity and settlement application. SPECTRE explains Layer-1 and SKY/ZAI protocol mechanics. They shows how a product or an authorised agent may use qualified interfaces; it does not own those protocols.
No incorporation, issuance, trading admission or live monetary service is created by this preview.
Reuse a capability without taking on every product or replacing the whole interface.
Permanent Composability is the discipline of versioned contracts, replaceable adapters and compatibility checks across Spectre, the seven applications and their hosts. APU supplies the outline and stored records; Agentz supplies conversation and contextual assistance. Each product keeps its canonical records and responsibilities. Applications can adopt qualified protocol or hosting interfaces independently; shared infrastructure does not merge permissions or authority.
Explore Permanent ComposabilityInteractive design preview · in development
Explore how an agent could receive a scoped task through different interfaces. No agent, credential or protocol execution is connected.
Skyzai / mini app
An agent task can appear inside Skyzai and use only the capabilities its operator or user has granted.
Cortex keeps memory; Nexus scopes context access across Skyzai. The integration receives only its permitted view. A delivery mode never grants blanket access or authority to act.
Revocation blocks future access; it cannot recall copies already delivered. This preview hides its local example when revoked.
No live endpoint, token or SDK is represented here.
An agent receives a permitted view for one purpose.
Cortex holds source-linked memory; Nexus scopes context disclosure across Skyzai. The intended contract distinguishes agent identity, task scope, grant expiry, permitted retention and revocation. Revoking a grant limits future access; copies already delivered require their own retention controls.
Explore Nexus permissionsCapability evidence ·
They follows product identity, configuration, lifecycle and supply-chain relationships, including products with no agent. The app feature map above covers product source. This selected evidence traces the narrower machine contributions through protocols and key-controlled agent accounts. Those accounts authenticate messages; key control does not create legal personhood, a bank account or authority to accept agreements. Accountable people and organisations retain responsibility for mandates and commitments. Agentz coordinates conversation and work, using Cortex memory and shared Nexus permissions; Spectre supplies networking; Layer-1 and SKY/ZAI mechanics remain protocol responsibilities. LIQUID owns the DEX application above them. These machine examples do not define the whole product perspective. Source and local tests do not establish live agent execution or monetary services.
Evidence describes source or a bounded test, not a released feature. A curated trace of 35 capabilities, not a complete inventory of every code path. None is live-verified or operating-qualified in this trace.
SPECTRE has a synthetic two-host transport proof, with real relay acknowledgement still to establish.
What the evidence saysThe private package, two synthetic Flutter hosts and 80 local tests exercise a bounded transport profile.
Still to proveSynthetic fixture signatures and host verifiers do not prove a real relay, authentic Nostr events or endpoint acknowledgement.
Agentz has local specialist-review and model-routing proofs, short of a live agent service.
What the evidence saysThe local host uses one bounded Agentz review step; a separate package demonstrates synthetic deterministic model routing.
Still to proveThis does not establish many-agent orchestration, live model service, independent consensus or global budget enforcement.
Permanent Composability supplies the versioned contract discipline; adoption by product owners is a separate gate.
What the evidence saysThree isolated packages use explicit contracts and synthetic two-host assembly.
Still to provePackage proofs are not adopted company-app dependencies or independent-team releases.
Attestation, scoped task-key release and measured usage are Cloud Computer targets, not verified operating infrastructure.
What the evidence saysCloud Computer plan specifies isolated execution, attestation, latency and metered usage.
Still to proveNo worker measurement, task-key release, mobile latency profile or currency settlement is implemented.
Storage replication must serve Agentz’s accepted memory contract, with explicit keys, retention and restore evidence.
What the evidence saysUser-controlled encrypted replicas and exact restore are owner-directed Cloud Computer requirements.
Still to proveNo replica count, recovery, failover, confidential inference or key recovery is implemented by the local receipt carrier.
Network usage and monetary protocol effects need independent reconciliation. Organisations retain their commitments in We; LIQUID owns the exchange and settlement application.
What the evidence saysCapacity, energy and economic receipt interoperability are future SPECTRE scopes.
Still to proveNo measured operating capacity, billing integrity or settlement authority is established.
Key-controlled machine accounts, bounded task examples, source-owned protocol contracts, composability and dated capability evidence.
Market discovery, provider admission, exact agreements, fulfilment and payment remain separate development gates. Key-controlled protocol accounts, delegated execution, compatible cloud hosts and independently admitted workers need separate qualification. Spectre node hosting, resilient storage, physical mesh operation, Layer-1 integration and SKY/ZAI mechanics remain separately evidenced work. Application migration is staged by workload. A local protocol proof does not establish a deployed network, live currency or a service-level commitment.
This public website is a design and evidence surface. A local demonstration is not running agent execution, production admission or deployment.
Read the development statusMe develops personal capabilities.
We coordinates the value chain.
They follows the product’s dependencies and working life.
Context: Detailed examples and capabilities.