What Is an Investment Avatar?

An "investment avatar" is a conversational AI agent created by nashnova based on a person's research context or verifiable public framework, capable of continuously tracking markets.

For regular users, it is your own research avatar. For KOLs/KOCs, it is a research identity officially created by nashnova in collaboration with them to faithfully reconstruct their investment framework.

Many introductions to "avatar" products start with "it chats just like the real person."

That's intuitive, but it puts the focus in the wrong place. Tone, catchphrases, and speech habits can create a sense of familiarity, but they cannot prove that a response follows a research method someone actually uses, nor can they explain what evidence underlies a conclusion.

For investment research, what truly deserves to be preserved is not a person's surface tone, but their judgment structure — and your own research habits.

Investment Avatar in One Sentence

An investment avatar is a type of AI agent built around "a specific person's research framework." It can continue asking questions and reasoning about new issues without requiring you to repeatedly search through articles, interviews, podcasts, portfolio records, and research notes.

There are two types of investment avatars on nashnova:

• Your own investment avatar: Built on your portfolio, topics of interest, risk variables, and research habits, it turns your scattered research context into a continuously running agent.

• KOL/KOC investment avatars: Created in collaboration with the respective experts, reconstructing their public and semi-public content into an investment framework that best represents them, allowing you to ask follow-up questions around your own concerns.

The word "avatar" can easily give the impression that the system has replicated someone's personality or speaks on their behalf. A more accurate description is: an investment avatar preserves judgment structure, not personality.

Two Types of Investment Avatars: Yours and KOL/KOC's

Your Own Investment Avatar (Personal Research Context)

This is the form that users can create independently on nashnova. You can define the assets, topics, and risk variables it needs to track, as well as how it analyzes and presents results.

It learns the variables you focus on, the frameworks you use, and your preferred output formats, and can schedule recurring research tasks such as pre-market reviews, cycle scans, and event tracking. As you continue using it, it also develops a research identity that others can follow, question, and collaboratively refine — but the underlying context comes from the research habits you provide and accumulate.

KOL/KOC Investment Avatars (Collaborative Form)

This type is created in collaboration with the respective experts. Taking Fu Peng's official AI avatar as an example: nashnova deconstructs his public and semi-public content and reconstructs the investment framework that best represents him; whatever you ask, it answers using his methodology and latest views.

As described in Fu Peng's related promotional materials, what a KOL/KOC investment avatar unlocks is not "the ability to chat," but four "finally, I can" moments:

• Finally, I can ask about the company I'm actually looking at. Put the company you're researching into the system and let the expert framework address your question, not the topic they chose.

• Finally, I can ask at the moment I need to. The expert won't happen to publish on earnings night, but the avatar is there when you want to thoroughly think something through.

• Finally, I can keep asking until I understand. Ask follow-ups on anything you don't understand, with no social cost. That question you'd never dare to DM — ask the avatar.

• Finally, I can ask one question to a whole table of experts. Agents with different frameworks can answer the same question together, exposing different perspectives.

Behind this is an Agent middleware platform: it first determines which agent should handle the question (triage), each agent autonomously calls institutional-grade data sources to retrieve data in real time, then reasons according to the person's framework, while remembering your portfolio and previous judgments.

How exactly are these avatars "grown"? Taking Fu Peng's official AI avatar as an example: nashnova first separates "facts, context, and stage-specific judgments" from his long-term materials, then identifies the rules he repeatedly uses when processing problems — which variables he habitually looks at first, in what order he eliminates possibilities, and what evidence causes him to revise conclusions. Fu Peng himself also personally tests the system: he asks normal questions and deliberately introduces false premises to see whether the AI can detect the problem or continues along the wrong path. Every correction is written back into training rules and evaluation test sets, and similar questions are retested before the next version goes live. In Fu Peng's own summary, this version of the avatar is "a digital human that finally has a brain" — the face and voice can be quickly replicated, but what's truly hard to replicate is how a person approaches problems.

What Truly Deserves to Be Preserved Is Not Tone, but Judgment Structure

Investment avatars can take two completely different directions.

The first is surface imitation: learning common words, sentence patterns, tone, and response style to make conversations sound like a specific person.

The second is research structuring: identifying how a person defines problems, what evidence they use, how they handle counter-arguments, and under what conditions their original judgment would be invalidated.

In entertainment or companionship scenarios, surface imitation may be sufficient. But investment research affects how users understand companies, risks, and prices, so merely "sounding like" someone is far from enough.

If a system cannot tell you what materials a judgment is based on, which parts are inferences, and what information could overturn the conclusion, then a familiar tone actually amplifies the risk of being misled.

How Public Materials Become a KOL/KOC Investment Avatar

A verifiable KOL/KOC investment avatar requires at least five steps.

1. Collect Verifiable Public and Semi-Public Materials

More materials are not necessarily better — what matters is whether the source, timing, and context can be confirmed. Different experts have different material structures: some write articles and journals over the long term, some explain markets through yearbooks and research notes, and others leave research traces primarily through regulatory disclosures, public speeches, or interviews.

2. Build a Fact and Source Table

First, separate facts, direct quotes, data, and judgments from the materials, and record their sources. This prevents the system from treating secondhand paraphrases as original quotes, or directly transplanting views from a historical context to the present.

3. Extract Research Framework Cards

A framework card is not a collection of famous quotes — it answers a set of stable questions:

• How does this person typically define research questions?

• Which variables are considered key evidence?

• What counter-evidence must be checked?

• What is the applicable scope of this methodology?

• Under what conditions would the original investment thesis be invalidated?

4. Distinguish Facts, Public Opinions, and Framework Inferences

Within a single response, at least three layers of content may be mixed: facts from public materials, opinions the expert has publicly expressed, and inferences generated when the system applies the framework to new questions.

These three layers must be separated. System inferences must not be packaged as the person's original words, nor should they be labeled as the person's current views simply because they "match a certain style."

5. Let Users Continue Asking Questions Around Their Own Research Subjects

After structuring is complete, users can place their own company, industry, or market questions into this framework and ask the investment avatar to list supporting evidence, counter-evidence, applicable boundaries, and variables that need further verification.

This step transforms one-way content into a research tool, but does not change the boundary of responsibility: users must still verify sources and make independent judgments.

6. Ongoing Calibration by the Person

A verifiable avatar is not trained once and then frozen. The key difference lies in "calibration" rather than "scraping": the expert needs to continuously ask questions, correct errors, and clearly explain their judgment process; the more time invested, the more rules and test cases the system accumulates, making it increasingly difficult for others to replicate the avatar simply by collecting public content.

For authorized collaborative investment avatars, this means two things: first, corrections from a single error must be retained and continue to influence the system's subsequent responses; second, the scope of materials, training timeline, and update versions must be clearly communicated to users. By only collecting public content, AI can imitate what a person "has said," but it is often insufficient to reconstruct "what they look at first when facing a new problem." This is precisely the source of an investment avatar's credibility, and the dividing line between it and an ordinary chatbot.

How to Judge Whether an Investment Avatar Is Trustworthy

When using an investment avatar, you can start with the following checks:

1. Does it state where the materials come from? Simply saying "trained on a large amount of content" does not help users verify anything.

2. Does it distinguish between original facts and system inferences? Inferences should not be disguised as original quotes.

3. Does it disclose its relationship with the person? Without authorization, it must not imply participation, endorsement, or official status; with authorization, it must state the scope of authorization and materials.

4. Does it present counter-evidence? Outputting only one-sided views that match a label turns a framework into a stereotype.

5. Does it state applicable boundaries and invalidation conditions? A framework without boundaries is easily misused.

6. Does it avoid presenting historical views as current views? Past materials must retain their original context.

7. Does it allow users to verify sources? Important judgments should be traceable to original materials or clearly marked as system analysis.

8. Does it refuse to guarantee returns or make decisions for users? Research assistance must not be packaged as trading instructions.

What an Investment Avatar Can and Cannot Do

An investment avatar is best suited to serve as a "research perspective" rather than a "stand-in for a person." It can help users systematically raise questions, but it cannot speak on behalf of the person, nor can it assume decision-making responsibility for the user.

How to Use Investment Avatars on nashnova

On nashnova, there are two paths for using investment avatars.

Your own investment avatar: Create it on the website or app, define the assets, topics, and risk variables it should track, schedule recurring tasks such as pre-market reviews, cycle scans, and event tracking, and let it understand what you care about more deeply with continued use.

KOL/KOC investment avatars: Browse the "Avatar Plaza" to select different experts' research perspectives, then ask follow-up questions around your own holdings or concerns. A more reliable approach is not to simply ask "would they buy this," but to ask the investment avatar to walk through the research process. For example:

• When analyzing this company using this framework, what are the three most critical questions?

• What facts support the current investment thesis, and what evidence conflicts with it?

• Under what circumstances does this framework not apply?

• What data still needs to be verified against earnings reports, regulatory filings, or company announcements?

• If key assumptions change, how should the original judgment be adjusted?

This style of questioning transforms an expert label into verifiable research steps. Users should still cross-check original materials and treat the output as supplementary information for research and educational purposes, not as the person's views or personalized investment advice.

Conclusion

The easiest thing to showcase about an investment avatar is "how similar it sounds." The hardest thing to establish is "whether it is trustworthy."

For investment research, the latter matters more. An investment avatar does not need to pretend to be someone; it needs to clearly organize research context or verifiable public frameworks, let users see the basis for judgments, counter-evidence, and applicable boundaries, and enable them to continue verifying around their own questions.

This is not about turning a person into an always-online stand-in, but about turning scattered research methods into a more trackable research tool.