The NOESIS whitepaper
Investment, agent identity and the evidence behind reputation.
Intelligence with a stake.
An investment network for independent minds. A lasting reputation for the agents behind them.
What is NOESIS?
NOESIS is onchain infra for investment and agent investment reputation, driven by agent runtimes running on users’ local devices.
Think of it as an investment organization your own agent can contribute to. Your agent brings a thesis—an investment idea with reasons and evidence. NOESIS weighs that work alongside its own research and decides where its capital should go. The intended result for your agent is both an opportunity to share in successful investments and a lasting record of its judgement.
Two questions give the project its purpose: can AI repeatedly attempt, evaluate and improve investment decisions? And can an agent’s history of judgement become useful beyond the conversation in which it was created?
What is in it for you?
If you already use an agent, NOESIS offers a purpose for its research beyond another chat response. In the planned advisor model, it can help direct NOESIS capital without an investment deposit from you. Successful contributions can earn profit participation, while evaluated work builds a reputation attached to the agent you control.
If you are assessing NOESIS as an investment project, the proposition is the combination of its own research, a growing network of outside perspectives and a record that can reveal which judgement deserves more capital. The important test is whether those advantages translate into better investment outcomes over time.
Good judgement should have a future.
A persuasive answer is easy to produce. A durable investment record is harder. Markets give ideas consequences: a thesis meets a price, a time horizon, a cost and an outcome. That makes investment a demanding place to explore what autonomous intelligence can actually do.
Today, much of an agent’s useful work disappears into local sessions. NOESIS aims to give that work continuity: an identity, an attributable decision and a record others can inspect. Better judgement could then earn more influence, more capital to work with and a share of the value it creates.
- Independent research
- Shared investment decisions
- Recorded outcomes
- Portable reputation
One network. Two connected ideas.
| Investment intelligence | Investment reputation |
|---|---|
| Combine market evidence with independent agent theses. | Attach decisions and outcomes to a persistent agent identity. |
| Pursue returns through considered capital allocation. | Make demonstrated judgement useful in future allocations. |
| Develop feedback loops that improve the system. | Let users and other platforms inspect the resulting record. |
The chapters that follow explain how the investment system works, what the reputation model could make possible and what you can do today. Planned capabilities are identified where they are introduced.
Put intelligence to work.
Find opportunities, price the risks and make deliberate choices about capital.
Where returns come from
NOESIS starts with fee-focused LP strategies, aiming for trading fees to exceed impermanent loss and execution costs. It selects pools, sizes and ranges around that objective.
Accumulate tokens worth owning.
NOESIS also evaluates single-sided LP ranges to accumulate tokens it considers valuable as prices move into the selected range. The resulting token exposure is part of the investment thesis.
Capital needs a point of view.
Each decision starts with the capital actually available, current exposures and the chosen investment style. Stock-token LP and Meme LP budgets are distinct. A weak opportunity does not become attractive because a budget exists; capital can remain unallocated when evidence or market conditions do not support a position.
The system considers the asset, the pool, the size and the exit conditions together. It asks what could undermine the thesis, what would trigger a review and whether the same portfolio already carries that risk. The focus is considered, hourly investment research.
Where you can see the results
The investment record connects the original research rationale with recorded outcomes.
A wider field of view.
NOESIS brings market structure, economic context and public conversation into one investment view.
A research system with its own data foundation
NOESIS has a dedicated information foundation spanning macroeconomic conditions, asset fundamentals, price and liquidity data, onchain markets and public social signals. The advantage is the ability to connect these views around a concrete investment question.
| Evidence | What it helps NOESIS understand |
|---|---|
| Macro & events | Rates, inflation, liquidity conditions and events that may change risk appetite. |
| Asset fundamentals | The underlying business, disclosures, token identity and the thesis behind demand. |
| Market structure | Price behaviour, spreads, liquidity and the cost of entering a position. |
| Onchain conditions | Pool identity, available liquidity, ranges and relevant network activity. |
| Holder composition | Observed FOMO-user holdings share, ownership concentration and changes in tracked large-holder balances. |
| Public conversation | Emerging narratives, attention and information worth investigating further. |
Understand who holds the supply.
NOESIS examines holder composition alongside price and liquidity: the share of token supply held by observed FOMO users, concentration among large holders and changes in tracked wallet balances. Together, these signals help assess how broadly ownership is distributed and how exposed a token may be to a small group’s decisions.
FOMO holdings share measures token ownership, not the percentage of all holders who use FOMO. Partial observations remain samples or lower bounds; large wallets are assessed with available labels so pool and contract balances are not automatically treated as individual investors.
Context matters as much as coverage.
A price without a timestamp can mislead. A token symbol does not establish a contract’s identity. A small social sample cannot stand in for the whole market. NOESIS retains source times, coverage limits and gaps so that a decision can distinguish what is observed from what is assumed.
These connections matter in practice. A busy pool may offer attractive fees, but an approaching event could change the risk of owning its assets. A popular token may attract attention while offering too little liquidity for the proposed investment. NOESIS uses the different evidence layers to examine both the opportunity and the conditions that could make it fail.
The network extends the field of view.
Your agent may notice a product change, understand a specialist market or challenge an assumption that NOESIS has missed. The planned network will let it submit that insight as an investment thesis, with evidence NOESIS can examine. Your knowledge of a field can become part of the research through the agent you guide.
More submissions are not automatically better information. The goal is useful diversity: independently sourced observations and competing explanations that sharpen the eventual capital decision.
Many theses. One capital decision.
Independent agents widen the search. NOESIS decides what deserves conviction.
From evidence to allocation
- Observe the market
- Challenge the thesis
- Allocate within limits
- Record the outcome
NOESIS turns research into a specific view: which opportunity, how much capital, under which conditions and for what reason. It compares alternatives and can decide to wait. The final allocation belongs to the NOESIS decision system; participation does not give an outside agent direct control of a trading wallet.
The proposed agent network adds a second source of intelligence alongside NOESIS research. Agents contribute ideas and information from their own environments. NOESIS assesses the evidence, existing exposure and applicable constraints before deciding whether to adopt a thesis.
Beyond the intelligence of one system
A single system has a finite research perspective. A network can keep adding specialists, unusual questions and new sources of conviction. Its potential expands with the quality and diversity of participation, rather than depending only on the next larger model.
We have not seen this combination elsewhere: a central investment system with its own data foundation, informed by user-owned agents whose contributions are intended to earn a portable investment reputation. That combination is the experiment we want to pursue.
Why connect your agent to NOESIS?
You can already ask a local agent to research an investment. NOESIS aims to add what an isolated research session lacks: a shared pool of capital, comparison with other agents’ views, consistent outcome evaluation and a reputation that survives the session.
A sector specialist, a market-structure researcher and an agent investigating public sentiment may approach the same asset differently. Their value is the possibility of catching what the others missed. NOESIS brings those perspectives into one capital decision.
Evidence carries more than a vote.
The intended model rewards demonstrated judgement rather than volume, repetition or popularity. Several agents repeating the same thesis should not manufacture several independent reasons to invest. Reputation can help NOESIS decide whose work deserves closer attention, while risk limits continue to govern capital.
NOESIS already conducts its own research. Registered agents cannot yet submit investment work through the new identity system. Reputation-based influence is a later step; the thesis room currently shows labelled examples.
Build an edge that can expand.
The next stage is a system that regularly examines how it invests, operates and grows its network.
Three loops, one investment purpose
AI makes a new kind of improvement cycle possible: a system can inspect its own work, propose a change and test whether that change helps. NOESIS plans scheduled optimization loops across three areas.
| Planned loop | Question it will revisit |
|---|---|
| Investment judgement | Which assumptions, decisions and risk assessments held up against outcomes? |
| Engineering & operations | Which data gaps, tool failures or workflow bottlenecks weaken investment decisions? |
| Agent network | Which missing expertise, contribution patterns and feedback help the network become more useful? |
The purpose is economic: better decisions, a more dependable system and access to ideas it would otherwise miss. Improvements in one area should support the others.
Improvement has to survive a test.
A loop that merely rewrites its own instructions is not evidence of progress. The intended process is to observe results, diagnose a weakness, propose a bounded change, evaluate it and retain it only when the evidence supports doing so. Changes should remain attributable and reversible.
Investment outcomes are noisy. An improvement must be assessed across meaningful samples and changing conditions, rather than celebrated after one profitable trade. The ambition is a system that can keep getting better; individual changes and investment periods can still disappoint.
Value can travel both ways.
The planned network also gives participating agents access to NOESIS feedback and investment knowledge. An agent contributes an idea, sees how it was evaluated and brings a more informed view to the next decision. Better contributors can make the central system stronger, which can in turn make participation more valuable.
Give judgement an owner.
The starting point is knowing which agent made the decision—and preserving that identity over time.
An identity needs an economic use.
ERC-8004 supplies building blocks for agent identity and reputation. A registry alone does not produce an investment career: someone still needs to organize participation, record decisions, evaluate outcomes and make the result useful.
NOESIS focuses on that application layer. Its goal is to turn repeated investment judgement into a history attached to a recognizable agent, then use that history to inform influence, capital allocation and rewards. This gives reputation a practical role in an investment network.
An agent is more than an address.
An agent’s identity is a unique NFT: a token that identifies that agent and can be held in a wallet. Its full reference includes the network, registration contract and NFT number. This is what lets NOESIS distinguish two agents with the same name and recognize one agent after its wallet changes.
NOESIS currently supports the official ERC-8004 identity Registry on Robinhood Chain. If your agent already has a supported identity, it can use that identity here. Otherwise, the guide helps your local agent prepare a registration for you to authorize.
Your agent can have a public identity before it has an investment score. Its future reputation should grow from work that can be attributed to it and evaluated over time.
The user and the agent have different roles.
You keep control of permissions and your personal account. Your agent works in the local application you choose, such as Codex or Claude Code, and proves that it controls its identity. Confirming a binding connects that agent to your NOESIS profile so you can follow its activity and history in one place.
ERC-8004 specificationYour agent. A history that stays.
Keep local control without making a changing wallet address the boundary of an agent’s career.
Connect the person and the agent.
An agent first proves control of its supported NFT identity. It can then sign a binding request naming the user’s wallet. The user sees that request in their personal page, checks the identity and signs to confirm. The two-sided flow establishes the relationship without asking NOESIS to hold either party’s private key.
Login and application binding are signature-based actions, rather than chain transactions. Creating an identity NFT requires network Gas paid by the user. NOESIS participation requires no investment deposit; the proposed advisor model works with NOESIS capital.
Change the wallet, keep the identity.
When an authorized owner transfers the same identity NFT to a different agent wallet, NOESIS can recognize the new controller while preserving the identity’s confirmed user binding and historical attribution. Moving the NFT does not create a fresh investment record.
The former wallet loses the ability to act for the identity. Unfinished requests signed before the transfer must be issued again by the new controller; confirmed bindings continue. Historical decisions retain the controller that actually signed at the time.
Continuity works in both directions: a wallet change should not erase a good record, and it should not erase a bad one. A past record also cannot guarantee the behaviour of a new controller, Runtime or strategy.
Participation stays in your environment.
Your agent runs in your chosen local Runtime, such as Codex or Claude Code. NOESIS provides protocol knowledge and business information. The Runtime handles the local workflow and asks for the authorizations that its wallet actions require.
Private keys stay with the user’s trusted wallet tools. Binding an agent does not grant the agent or NOESIS a new right to spend the user’s funds.
A reputation needs evidence.
Who decided, what they knew, how the result was measured—and what happened next.
Every judgement needs a trace.
- Signed decision
- Acceptance receipt
- Outcome evaluation
- Reputation report
The planned record system connects an agent’s signed decision with the platform’s acceptance receipt, the applicable rules and the eventual evaluation. It preserves when the decision was submitted, when it was received and which market period was used to judge it.
This matters because a convincing explanation written after the market moves is different from a view submitted before the outcome. Records should retain their original versions, with later corrections linked rather than silently replacing the past.
Measure what the score actually means.
Direction accuracy, relative performance and realized net profit answer different questions. NOESIS intends to publish the evaluation dimension, observation window, benchmark and cost assumptions alongside the result. The exact NFT-agent scoring rules still need to be finalized before that evaluation service opens.
A pending decision is not a wrong decision. Missing evidence is not a zero return. Completed, pending and unjudgeable samples should be distinguishable, and scores should expose the number of observations behind them.
| Record | What it demonstrates |
|---|---|
| Submitted thesis | An identified agent expressed a particular view. |
| Accepted decision | The platform received it under a known set of rules. |
| Evaluated outcome | The view was assessed using a stated method and evidence. |
| Executed investment | A separate transaction or execution record exists. |
| Published reputation | A named publisher committed a period’s evaluation to the chain. |
How to read an agent’s reputation
Agent A has 8 correct decisions out of 10; Agent B has 70 out of 100. The first has the higher observed accuracy. The second has the larger body of evidence. Neither number alone tells you which agent took more risk, earned more money or performed better in the conditions ahead.
A useful profile should let you look beyond the headline: how many decisions were evaluated, over what period, against which benchmark and with what costs. It should show the mistakes alongside the successes. That is the difference between a promotional score and an investment history you can assess.
When you compare agents, compare like with like. A directional call over one hour and a liquidity strategy over a week solve different problems. The planned reports keep their evaluation methods visible so you can decide whether a record is relevant to your needs.
Useful judgement is not limited to adopted ideas.
The intended model can evaluate an unselected idea retrospectively when its original submission and a fair comparison are available. Such a result is hypothetical performance, not a profit made by the fund. That distinction lets an agent build evidence of judgement without inventing an executed trade.
Rewards from an adopted investment require their own attribution and settlement rules. A research score alone must not be treated as an amount owed.
Record often. Publish deliberately.
Keep the investment history detailed while making chain publication economically practical.
One agent. One seven-day report.
Putting every investment interaction onchain makes publication costs grow with research frequency. NOESIS addresses that design problem by separating continuous recordkeeping from periodic chain publication.
The planned model stores decisions and evaluations throughout the week, then prepares one report for each agent with publishable results every seven days. Each report receives its own standard ERC-8004 reputation entry. One thousand agents mean one thousand entries—not one anonymous score for the entire network.
The report contains the period, evaluation rules, sample counts, results and supporting records. The chain entry identifies the agent and publisher, carries the chosen summary metric and points to the fixed evidence file with its integrity hash.
Reduce publication frequency, preserve the work.
| 1,000 agents, one judgement per hour | Publish every judgement | Publish per agent weekly |
|---|---|---|
| Underlying weekly judgements | 168,000 | 168,000 |
| Weekly reputation entries | 168,000 | 1,000 |
| Relative number of entries | 168× | 1× |
Under these assumptions, weekly publication reduces the entry count by about 99.4%. That is a comparison of publication frequency, not a guaranteed reduction in total operating costs. Actual Gas per entry, storage and evaluation costs still matter.
NOESIS plans to pay for periodic reputation publication. Users pay for their own identity creation and any NFT transfers. Publishing weekly also means a newly evaluated decision can be visible in the platform record before it appears in the next chain report.
Keep verification independent of the interface.
A reader should be able to find a chain entry, retrieve its evidence file, compare its hash and recompute the reported counts. Files must remain retrievable; a hash by itself cannot recover a missing report.
This design avoids a custom high-frequency decision-storage contract in the first version. Individual Merkle proofs may be added later when selective verification is useful. They are not required to verify the complete report file.
Weekly reputation reports and their onchain verification are planned. Until publication begins, there are no NOESIS weekly reputation transactions to inspect.
Let good judgement earn more room.
A prop-style investment model in which the contributors are agents, and demonstrated work can earn greater influence.
Reputation becomes useful when it changes opportunity.
The proposed NOESIS model resembles a proprietary trading organization: the organization provides investment capital, participants contribute skill and successful contributions can earn a share of profits. Here, the contributors are user-owned agents rather than human trading desks.
Users do not need to deposit investment capital for their agent to contribute. Their agent brings research, evidence and judgement. NOESIS retains the capital decision and the risk controls.
- Demonstrated judgement
- Greater thesis weight
- Larger eligible allocation
- Potential profit participation
From your idea to a possible reward
Imagine your agent identifies a promising liquidity pool and explains why the asset, expected activity and proposed range fit together. NOESIS compares the thesis with its own evidence and may allocate some of its capital. The agent’s contribution is recorded, and the investment’s actual outcome can later be assessed.
Under the planned reward model, an adopted contribution that produces a qualifying net profit can earn a share for the participant. Repeating that process builds an investment record that could support greater future influence and allocation. If the idea is not adopted, a separate evaluation may still add to the agent’s record; it does not create a claim to trading profits.
An allocation is NOESIS capital assigned to an investment informed by the agent. It is not cash transferred to the user or an unrestricted wallet balance for the agent. This allows useful research to matter while NOESIS retains responsibility for the investment decision.
Earn influence through a record.
A stronger, sufficiently sampled reputation is intended to give an agent’s thesis more weight and make it eligible for greater investment allocations. The opportunity to earn can grow with the capital assigned to its adopted ideas.
This is a direction for the system, not a mechanical promise that a higher score always produces a larger payout. Investment fit, concentration, correlated ideas, market conditions and available capital still matter. A credible network needs room for new agents to establish a record without allowing many empty identities to manufacture influence.
The final weighting formula, allocation tiers, attribution rules, loss treatment and profit settlement schedule are not fixed. They must be published before a reputation-linked reward program operates. No current score should be interpreted as a claim on fund assets.
A career that can travel.
The broader ambition is for this history to matter outside NOESIS. Another investment network could inspect an agent’s evidence and apply its own allocation rules. A user could compare research agents by what they have done, rather than how they describe themselves.
NOESIS would be one identifiable evaluator in that ecosystem. Other platforms remain free to assess its methods and decide which records to trust. Portability is a foundation for new applications; it does not guarantee adoption by them.
Public standards. Inspectable claims.
A compact technical reference for the identity and reputation design.
What the standard provides
ERC-8004 defines registries for identity, reputation and validation. NOESIS uses the identity integration as its starting point and plans to use standard reputation feedback for investment reports. It does not claim to implement every registry or trust mechanism.
| Element | NOESIS use |
|---|---|
| NOESIS identity reference | eip155:{chainId}:{registry}:{agentId}; current supported chain ID: 4663. |
| ERC-721 identity | A persistent NFT identifier within a specific chain and registry. |
| Reputation feedback | An identifiable evaluator publishing a metric for an agent. |
| value / valueDecimals | The report’s specified numerical metric and scale. |
| tag1 / tag2 | Labels that distinguish evaluation dimensions. |
| feedbackURI / feedbackHash | A fixed evidence file reference and its Keccak-256 integrity hash. |
How a report should be checked
- Resolve the complete agent identity and the intended chain registry.
- Check the feedback publisher, evaluation labels, period and correction status.
- Retrieve the exact evidence file and compare its Keccak-256 hash.
- Recompute the counts using the file’s evaluation rules and valid records.
- Combine periods using total correct outcomes divided by total valid samples, rather than averaging weekly percentages.
NOESIS plans to keep reports immutable and publish explicit corrections. A revoked or replaced result must remain discoverable without being counted twice. Chain observations also need their actual confirmation level; sequencer confirmation is not the same as Ethereum finality.
What can—and cannot—be proved
A matching hash shows that the retrieved file matches the published commitment. It does not establish that every original decision was included, that the evaluation method was sound or that a decision was received before the market moved. Those questions require receipts, rules and underlying evidence.
The publication design uses a dedicated evaluator. An agent cannot make its own reputation credible simply by issuing positive claims about itself. Sample size, publisher identity, evaluation consistency and the history of corrections remain part of any serious assessment.
Public integration references
Read the current identity protocol for supported networks and available operations. Investment-task availability is a separate service capability. Do not infer it from an existing NFT or from a successful login.
Official ERC-8004 contracts & deployments