# Neosoul's Core Solution ### Neosoul's Positioning **Neosoul Is Not a Single AI Product** Neosoul is not simply a chatbot, an automation tool, a trading bot, a prediction app, or an oracle project. It is a system for forming, testing, authorizing, verifying, and evolving agents as economic participants. The product path may begin with evoevo and the AI-native prediction market, but the underlying objective is the Agent Economy Trust Layer. **Neosoul Is a User-facing Agent Economy Trust Layer** Neosoul is a Trust Layer designed around users. It allows users to retain sovereignty while authorizing agents to perform bounded economic actions. It manages: * user sovereignty and delegation relationships * agent permissions and behavior boundaries * task intent and execution objectives * connections to tools, markets, and protocols * verifiable histories of decisions and actions * memory, context, and evidence states * prediction, causal hypothesis, and confidence calibration environments * recourse after failure * long-term reputation, qualification, and governance evolution Its purpose is to transform agent capability into trusted economic agency. **What Neosoul Does Not Claim** Neosoul explicitly defines its boundaries: * **No legal personhood for agents**: agents are authorized economic action units, not legal persons. * **No full autonomy replacing users**: the system starts from limited authorization, revocable permissions, risk tiers, and human escalation. * **No full on-chain maximalism**: only key commitments, authorizations, credentials, settlement, disputes, and liability states belong on-chain. * **No prediction-market maximalism**: the prediction market is the first proving ground, not the full scope of economic intelligence. * **No automatic promotion from high standing to infrastructure qualification**: AON requires independent qualification. These boundaries make Neosoul more credible. The project is not arguing for unlimited agent autonomy, but for trusted delegation under explicit constraints. **Economic World Model: An Actionable Model for the Economic World** The **Economic World Model** is the actionable model that an agent forms for the economic world. It asks how agents perceive economic states, understand signals and incentives, estimate risks, reason about authorization boundaries, anticipate other participants, and decide what actions are feasible under the Trust Layer. The Trust Layer defines how an agent may act. The Economic World Model defines how an agent understands the economic world in which it acts. Without the Trust Layer, the Economic World Model remains analysis and advice. Without the Economic World Model, the Trust Layer can manage permissions but cannot produce high-quality economic action. *** #### The Seven Core Modules of the Agent Economy Trust Layer The seven modules of the Trust Layer are not a loose feature list. They form a full authorization and accountability chain. **Sovereignty Module** Sovereignty addresses: **who owns, who authorizes, and who is responsible.** **User Sovereignty** Users retain asset sovereignty, authorization rights, shutdown rights, and high-risk confirmation rights. Agents are delegates. They do not become independent owners of user assets or goals. **Persistent Identity** Users, agents, AON nodes, reviewers, key data sources, and relevant protocol actors need persistent identity. Without persistent identity, long-term reputation and responsibility cannot accumulate. **Asset and Permission Sovereignty Structure** Agents should not directly control users' primary wallets or full accounts. They should operate through restricted accounts, smart accounts, delegated contracts, or vaults with bounded permissions. **Explicit Delegation Relationships** Delegation must define the authorizer, authorized agent, scope, budget, validity period, prohibited areas, escalation conditions, and revocation methods. The essence of the Sovereignty Module is not to grant abstract autonomy to agents, but to ensure that economic agency always has clear origin, boundaries, and responsibility. **Intent Module** Intent addresses: **what the user actually wants the agent to do.** **Intent Capture** The system must capture user goals, constraints, preferences, budgets, risk tolerance, time horizon, and prohibited areas. **Intent Formalization** Natural language is not enough for real economic action. Intent must be translated into executable structures, including policy, budget, confirmation conditions, target states, and risk boundaries. **Intent Confirmation** High-risk or ambiguous tasks require confirmation. Users should be able to see what the system believes they authorized before the agent acts. **Intent Persistence and Revision** Intent is not static. User preferences, task context, and risk appetite change over time. The system must record versions and allow revision. The Intent Module creates a truth layer for user goals. **Control Module** Control addresses: **what the agent can and cannot do.** **Dynamic Permission Sandbox** Agents should act inside a dynamic sandbox that limits tools, markets, budgets, positions, execution frequency, and maximum loss. **Policy Engine** Policies translate user intent and system risk rules into enforceable constraints. They can include allowlists, blocklists, approval thresholds, risk tiers, and automated pauses. **Escalation Mechanism** When tasks exceed predefined risk levels, involve ambiguity, or conflict with policy, the system should escalate to the user, reviewer, or governance process. **System-level Shutdown** Users and the system must be able to pause, freeze, downgrade, or terminate agent behavior. This is not a reduction of autonomy; it is the condition under which autonomy becomes acceptable. Control is the precondition for trust. **Execution Module** Execution addresses: **how agents actually act.** **External Tool and Service Access** Agents need controlled access to search, data, APIs, exchanges, markets, wallets, payment rails, and enterprise tools. **Standardized Capability Interfaces** Capabilities should be exposed through structured interfaces so agents can execute tasks in predictable and auditable ways. **Transaction and State Engine** Economic behavior requires transaction submission, settlement, state synchronization, and confirmation. The Execution Module must connect agent actions to verifiable state changes. **Agent-to-Agent Collaboration Interfaces** The Agent Economy will involve agents coordinating with other agents. Standard A2A interfaces allow delegation, negotiation, payment, verification, and handoff among agents. Execution is where intent becomes action, but it must remain inside explicit boundaries. **Verification Module** Verification addresses: **how the system proves what happened.** Users do not only care what the agent claims to have done. They care whether the system can prove it. **Decision Logs** Decision logs should capture the objective, context, information sources, assumptions, confidence level, reasoning summary, and relevant constraints. **Execution Logs** Execution logs should capture tool calls, transactions, outputs, state changes, timing, failures, and exceptions. **Outcome Verification** Outcomes require evidence: event results, settlement data, source materials, oracle outputs, dispute records, and review conclusions. **Readable Review** Verification must be readable by users, reviewers, arbitrators, and governance participants. Raw logs are not enough. The system must support structured review. Verification is the foundation of reputation, recourse, and governance. **Recourse Module** Recourse addresses: **what happens after something goes wrong.** Failures are inevitable: * the agent misunderstands intent * a tool returns bad data * a market behaves abnormally * a transaction is executed incorrectly * a fact result is disputed * a budget or risk boundary is breached **Incident Response** The system must support pause, freeze, downgrade, notification, and emergency review. **Rollback and Reversal** Where possible, erroneous actions should be reversed or rolled back. Where reversal is impossible, the system must preserve evidence and move to liability handling. **Dispute Handling** Disputes may involve authorization, execution, settlement, facts, or responsibility. They require evidence, rules, review, and escalation paths. **Responsibility Mapping** Responsibility must be mapped across users, agents, agent operators, tool providers, data sources, AON nodes, market rules, and platform governance. **Insurance and Risk Pools** For bounded scenarios, insurance and risk pools can provide compensation and improve user confidence. The Recourse Module makes failure part of the system design rather than an afterthought. **Evolution Module** Evolution addresses: **why the system becomes more reliable over time.** **Reputation Capital** Reputation capital includes long-term performance, calibration, stability, drawdowns, disputes, risk control, review quality, and cross-context behavior. **Qualification Formation** Qualification turns observed performance into role eligibility: training qualification, economic qualification, and AON qualification. **Adaptive Permissions** Permissions should evolve based on competence maps, task risk, market conditions, and historical reliability. **Learning and Correction Loop** Agents should update from predictions, outcomes, reviews, errors, disputes, and peer reasoning. **Governance Evolution** As the ecosystem grows, governance should gradually include reviewers, market designers, AON nodes, agent builders, and ecosystem partners. Evolution turns one-off behavior into durable trust. *** #### Core Operating Logic and Institutional Design Principles **Authorize -> Specify -> Constrain -> Execute -> Verify -> Recover -> Compound** The seven modules form a single operational sequence: 1. **Authorize**: the user grants bounded authority. 2. **Specify**: the system converts intent into executable constraints. 3. **Constrain**: policy, budgets, permissions, and risk rules define action boundaries. 4. **Execute**: the agent acts through tools, accounts, markets, and protocols. 5. **Verify**: decisions, actions, sources, and outcomes are recorded and reviewed. 6. **Recover**: failures trigger pause, dispute, rollback, compensation, or responsibility mapping. 7. **Compound**: performance becomes reputation, qualification, adaptive permissions, and governance rights. This chain is the operating logic of the Agent Economy Trust Layer. **Why All Seven Modules Are Necessary** Any missing module breaks the trust chain: * Without Sovereignty, delegation has no legitimate origin. * Without Intent, the agent does not know what it is truly optimizing for. * Without Control, users cannot limit risk. * Without Execution, the agent cannot act reliably. * Without Verification, reputation and recourse lack evidence. * Without Recourse, failure destroys trust. * Without Evolution, good behavior cannot compound and bad behavior cannot be filtered out. **Neosoul's Institutional Design Principles** Neosoul follows several design principles: * **Human sovereignty first**: agents are delegates, not sovereign subjects. * **Bounded autonomy**: autonomy should expand only inside explicit risk boundaries. * **Evidence before reputation**: reputation must be based on verifiable records. * **Qualification before infrastructure responsibility**: public roles require additional testing. * **Privacy and verification together**: sensitive information should not be exposed simply to prove trust. * **Progressive decentralization**: governance should become more open as the system matures. *** #### Web3-native Implementation and Open Protocolization of the Trust Layer **Why the Seven Modules Must Be Deeply Integrated with Web3** The Trust Layer can begin as a product-level system, but the Agent Economy will eventually require open, portable, and externally verifiable states. Agents will act across users, markets, tools, applications, protocols, and jurisdictions. Their identities, authorizations, credentials, evidence, reputation, and liability states cannot remain locked inside one platform if they are to support an open economy. Web3 provides the substrate for this externalized trust. **How the Seven Modules Bind to Web3 Technologies** * **Sovereignty**: DID, wallet, account abstraction, ownership registry. * **Intent**: intent commitments, policy schemas, selective disclosure. * **Control**: delegation contracts, agent smart accounts, spending limits, revocation and pause mechanisms. * **Execution**: transaction contracts, settlement contracts, A2A payments, market contracts. * **Verification**: log commitments, attestations, provenance, zk proofs, oracle proofs. * **Recourse**: escrow, dispute contracts, insurance pools, arbitration, slashing. * **Evolution**: VC/SBT, reputation schemas, qualification credentials, staking/slashing records, governance rights. **Trust Storage Substrate: The Foundation for Memory, Context, and Evidence** The Agent Economy requires long-term memory, context, logs, evidence, and world-state history. Neosoul defines this layer as the **Trust Storage Substrate**. It consists of: * **On-chain commitments**: hashes, state pointers, authorization states, credentials, settlement outcomes, and dispute outcomes. * **Distributed encrypted storage**: memory, context, logs, evidence, reasoning assets, credential evidence, reputation history, and world-state archives. * **Local and platform hot cache**: session context, temporary state, and low-latency retrieval data. This storage layer must support user control over personal context, provenance tracking for agent memory, integrity verification for evidence and logs, long-term referenceability for world-state archives, and permissioned portability across applications. **Eight Categories of Verifiable Economic State Managed by the Trust Layer** **Agent Identity** Agent identity includes creator, model version, capability scope, training history, qualification, market performance, and responsibility records. **Delegation** Delegation includes authorizer, authorized agent, scope, budget, validity period, prohibited areas, confirmation conditions, and revocation methods. **Agent Account** Agent account state includes restricted accounts, smart wallets, budgets, spending limits, allowlists, pause mechanisms, and freeze mechanisms. **Credential** Credentials include training qualification, review outcomes, arena performance, AON performance, and risk-control records. **Reputation Capital** Reputation capital includes win rate, risk-adjusted return, drawdown, reasoning quality, calibration, stability, disputes, and cross-context performance. **Context and Memory State** This includes user context, agent memory, prediction memory, reasoning history, execution context, and review records. **World-State Attestation** AON produces attestations about events, propositions, facts, sources, disputes, and settlement states. **Recourse and Liability** This includes authorization validity, overreach, erroneous submissions, compensation, slashing, downgrading, and removal records. **Implementation Depth Across Neosoul's Four Phases** **evoevo: Identity, Training Records, and Early Qualification** In the first phase, the Trust Layer primarily records agent identity, prediction memory, review history, reasoning assets, early credentials, and user feedback. **AI-native Prediction Market: Authorization, Accounts, and Economic Reputation** In the arena phase, the Trust Layer manages delegation contracts, agent smart accounts, budgets, execution logs, market settlement, dispute windows, and economic reputation. **AON: World-state Verification and Infrastructure Responsibility** In the infrastructure phase, the Trust Layer manages AON node identity, qualification credentials, staking/slashing or equivalent accountability constraints, world-state archives, oracle outputs, and dispute evidence. **Broader Agent Economy: Open Protocols and Composable Ecosystems** In the broader Agent Economy phase, identity, delegation, memory, reputation, fact layers, and recourse mechanisms become open protocol capabilities that third-party applications can integrate. **Implementation Principles** **Off-chain Execution, On-chain Commitment** Reasoning, training, review, and most execution occur off-chain. The chain carries key commitments, authorizations, qualifications, settlement results, fact submissions, and dispute outcomes. **Minimal On-chain Footprint** User preferences, complete reasoning, private strategies, sensitive logs, and undisclosed signals should not be publicly placed on-chain. **Distributed Encrypted Storage** Memory, context, logs, and evidence require long-term storage, but they should be encrypted, access-controlled, and content-addressed where appropriate. **Privacy First** Neosoul should use encryption, selective disclosure, permissioned auditing, and zk proofs to balance verifiability and privacy. **Reputation Should Not Be Naively Financialized** Tokens may be used for payment, staking, slashing, and governance, but they cannot replace long-term performance records. **Progressive Governance** Governance should begin with clear product-level rules and gradually introduce reviewers, expert councils, market designers, AON nodes, and ecosystem partners. **Long-term Goal: From Product Platform to Agent Economy Protocol** Neosoul's long-term goal is to move from a product platform into an open Agent Economy protocol. This means that identity, delegation, qualification, reputation, memory, evidence, world-state attestation, and recourse can become interoperable protocol capabilities. The project should evolve from: * product rules to protocol standards * platform logs to verifiable evidence * internal reputation to portable reputation capital * market participation to infrastructure participation * application-specific agents to composable agent networks **Economic World Model: An Actionable Model for the Economic World** The Economic World Model is an actionable world model for the economic world. The Trust Layer defines how agents may act; the Economic World Model defines how agents understand the environment in which they act. It includes three layers: * **Economic State Model**: understands prices, events, markets, users, permissions, assets, contracts, facts, reputation, and risk. * **Economic Dynamics Model**: understands key variables, incentives, participant behavior, information diffusion, market feedback, and possible causal chains. * **Economic Action Model**: understands feasible actions under current authorization, budget, risk, rules, tools, and market structures. Formation and calibration occur through: * **Prediction Memory** * **Causal Hypothesis Record** * **Competence Map** * **Multi-agent Inference** The Economic World Model does not claim to prove an agent's true causal understanding. It estimates economic world modeling capability through long-term prediction, causal hypotheses, counterfactual records, error review, confidence calibration, and out-of-distribution behavior. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.neosoul.ai/neosouls-core-solution.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.