Author: Gandalf, Techub News
Introduction
AI can gather information, read financial reports, organize data, and even generate a seemingly complete research report in just a few minutes. However, for investors, a faster pace of information acquisition does not automatically lead to better judgment. The truly difficult questions are often: why was I optimistic from the beginning? Which judgments were validated by facts, and which just happened to make money? How can I avoid repeating the same mistakes next time?
Crazyox has a long background in law and has also been an author, later entering Web3 and continuously investing in AI products and automation practices. Legal training has conditioned her to habitually reverse-engineer conclusions: what is the evidence? Is this fact or inference? Is there contrary evidence? What new facts may overturn existing judgments?
This evidence-based thinking became the starting point for her initiative "Building Block Investment Research." What she aims to create is not an AI tool that predicts the market and provides buy and sell answers for users, but rather a system that helps individual investors build a research framework, preserve decision points, track changes in evidence, accept counter-evidence, and continuously review their investment decisions. It is more like an "investment research incorrect question collection": not only recording buys and sells and profits and losses but how a person thinks and where they repeatedly make mistakes.
Core Viewpoint
"Information will become cheaper and cheaper, but judgment will not automatically improve."
In Crazyox's view, the best-fit role for AI is to enhance the efficiency of research and cognition: organizing materials, tracking new facts, finding supporting and opposing evidence, and highlighting evidence conflicts; however, the final judgment, evidence weighting, and investment decision must be retained and borne by the user.
The goal of "Building Block Investment Research" is also not to provide a set of unquestionable standard answers but to let users start from a basic viewpoint and break it down, verify it, overturn it, or rebuild it. The longer it is used, the more the system should not just embody the methodology of the product initiator but gradually solidify into the user's own research framework, investment incorrect question collection, and cognitive history.
From Lawyer Thinking to an Investment Research Incorrect Question Collection
TECHUB NEWS: Please introduce yourself and why you initiated "Building Block Investment Research"?
Crazyox: My background is quite cross-disciplinary; I have worked in law for many years, also written books, then entered Web3, and am now mainly engaged in AI products and automation, while maintaining a long-term focus on investment research.
The earliest origin of Building Block Investment Research came from a pain point of mine. AI is already able to quickly search for information, read financial reports, and write research reports; in the future, information production will become increasingly cheaper. However, the real difficulty is not "can I find information," but how to judge: why was I optimistic initially? What was right, and what was wrong later? Can I make fewer of the same mistakes next time?
So initially, I just wanted to create an "investment incorrect question collection," which gradually evolved into Building Block Investment Research. It does not predict the market for people but helps users build, verify, review, and iterate their investment judgments.
TECHUB NEWS: How does the habit of legal work influence your investment research methods?
Crazyox: Lawyers have a professional habit: when someone presents a conclusion, my first reaction is usually, "where is the evidence?" Then I will ask, is this fact or inference? Is the source of evidence reliable? Is there contrary evidence? What new facts, once presented, could overturn the original conclusion?
Later, I realized that law and investment are very similar, as both require making judgments in situations of incomplete information. What legal training gave me is not to keep finding evidence to prove "I am right," rather it taught me to habitually question: if I am wrong, where would I be wrong?
This is also why Building Block Investment Research incorporates evidence, counter-evidence, and falsification mechanisms into the research process. Research is not about seeking more support for an existing position but allowing one’s views to be challenged by facts.
Permanently Retaining Decision Points
TECHUB NEWS: What is an "investment incorrect question collection"? How is it different from typical trade records?
Crazyox: Traditional trade records usually note when to buy, when to sell, how much was made or lost, more like a report card. However, when students compile incorrect question collections, they do not just record how many points they scored, but clarify why a certain question was wrong.
Building Block Investment Research aims to accomplish two things. First, to leave a decision point for every significant investment decision. Whether buying, increasing positions, decreasing positions, or changing judgments, users can mark this decision on the K-line and timeline while preserving the investment logic, core drivers, and evidence from that time. After six months, when looking back, users can return to that information environment and ask themselves: if I only knew that information at the time, how good was this decision?
This is a bit like the "black box" of investment decisions. It preserves not the explanations rearranged by outcomes today, but the genuine opinions, grounds, and unknowns present at the time of decision-making.
Second, every round of research and review will generate a report documenting how opinions changed, whether evidence strengthened or weakened, how performance fared, and how it can ultimately be attributed. In the future, these reports can also be analyzed by AI.
The incorrect question collection actually accumulates not "which stocks I have bought," but "how I have thought in the past and where I repeatedly made mistakes."
TECHUB NEWS: Why is recording the original judgment a prerequisite for effective review?
Crazyox: Because it is easy for people to re-interpret their past selves using results they already know today. Therefore, the system will preserve a T0, which is the decision point. The T0 records not the subsequent interpretations but the actual opinions, grounds, and unknowns at that time; later, new facts can be added, but the initial judgment cannot be rewritten.
Thus, reviewing after some time creates an opportunity to distinguish whether the initial judgment was correct or if money was obtained merely by chance. Without the original judgment, there is no real review.
Constructing a Research Framework with Building Blocks
TECHUB NEWS: Why is it called "Building Block Investment Research"? What do these building blocks specifically refer to?
Crazyox: Many ordinary investors do not lack viewpoints, but they do not know how to research their views. For instance, someone thinks "AI computing power demand will continue to grow," or "the storage industry's upward cycle is beginning," what should be researched next? Traditional investment research often provides a whole set of complex templates, which can easily lead ordinary users to give up.
Building Block Investment Research seeks to start from the simplest question: "Why do you think this is promising?" Subsequently, one viewpoint can be broken down into several core driving factors, such as demand, capital expenditure, financial reports, cash flow, expansion, pricing, inventory cycles, or market share.
Furthermore, each core driver can link to different Skills, which are the building blocks of research methods, for example, valuation analysis, AI value migration, dilution analysis, leverage risk, etc. The entire logic can be summarized as:
Viewpoint → Core Driver → Skill Research Method → Evidence → Forming Personal Research Framework
It is called building blocks because it is not a fixed template. A specific Skill that is not applicable can be removed; discovering a missing key driver can be added; if the entire framework is proven problematic by new facts, it can be dismantled and rebuilt. The goal is to enable someone who is initially unfamiliar with systematic investment research to start from their viewpoint, piece by piece, building up a research framework, and continually verifying, deconstructing, and reconstructing it in the actual market.
TECHUB NEWS: From the preliminary idea to verifiable investment arguments, how does the system help users verify?
Crazyox: One of the difficulties of subjective investment is how one's viewpoint can be "backtested." Suppose a tech stock is picked and rises by 30% in a year; does this prove strong research ability? Not necessarily. If during the same period, QQQ rises by 40%, then the meaning of this 30% is different.
Therefore, Building Block Investment Research will introduce external benchmarks. Researching a tech stock can be compared with QQQ, broader US stocks can be compared with S&P 500 or SPY, and specific industries can be compared with corresponding industry ETFs. It does not only look at absolute returns but also assesses relative performance.
Yet returns are just one dimension. Another critical dimension is: whether the evidence supporting the viewpoint subsequently strengthened or weakened? That is, the system simultaneously observes whether the research logic has been validated by facts and whether the results have outperformed the corresponding benchmarks. Only when these two factors are combined can there be an opportunity to gradually differentiate research ability, market Beta, and luck.
Allowing AI to Search for Counter-evidence
TECHUB NEWS: How do you handle supporting evidence, counter-evidence, and the weight of Skills and evidence?
Crazyox: Here, three levels need to be differentiated: core drivers answer "what determines my judgment"; Skills answer "what method do I use to study it"; Evidence answers "does the fact actually support it or not."
Currently, we do not pre-set fixed weights for Skills. For example, regulating valuation analysis to account for 30%, inventory cycles for 20% seems scientific, but the significance of the same research method does not vary under different companies, stages, and investment logics. Skills are the ruler, not the facts themselves.
The true distinction in weight needs to be made on evidence. The original company financial reports, realized income, should not have the same probative power as market rumors or management's predictions about the future. AI can assist in searching for evidence, highlighting sources, finding counter-evidence, and discovering conflicts, but ultimately, it is up to the user to determine the importance of a piece of evidence to their judgment.
As for which Skills are genuinely effective in the long term, I would rather let the review results tell the users, rather than artificially stipulate it at the beginning. This is also an important future layer in Building Block Investment Research: allowing research methods themselves to undergo long-term testing.
TECHUB NEWS: When new financial reports, industry data, or company events occur, how does the tool help users return to the research logic rather than follow the stock price?
Crazyox: The core idea is simple: do not let price do the thinking for you.
When new financial reports emerge, the system should first return to the original core drivers. For example, if the initial optimism about a company was based on believing that capital expenditure would ultimately translate into revenue and cash flow, then the next financial report should check: has investment increased? Is income realized? Is cash flow keeping up? Is the original supporting evidence strengthening or weakening? Has it triggered the previously set reevaluation conditions?
Rather than determining that investment logic has been validated simply because the stock price rose by 8% that day. Price changes do not equate to changes in investment arguments. What the system should genuinely continue to track is what facts have occurred and whether those facts have altered the original research logic.
TECHUB NEWS: Do you hope AI to become a "counter-researcher" that actively challenges users' viewpoints?
Crazyox: Absolutely, and I believe this may be one of the most valuable roles for AI in investment research. Investors can easily fall prey to confirmation bias: if they have already bought into a company, AI might just find ten favorable pieces of information according to the existing viewpoint every day, effectively turning it into a high-level information echo chamber.
Therefore, the more I believe in something, the more AI should attack it. If you think demand will grow, AI should seek evidence of weakening demand; if you believe the company has a moat, AI should look for competitors that are circumventing that moat; if you think management is excellent, AI should check whether their past commitments have truly been fulfilled.
I hope Building Block Investment Research can possess the capability of "Attack My Thesis"—to attack my investment logic. It is not for the sake of opposition but to force users to confront the evidence they are most unwilling to see.
Review Ability, Not Standard Answers
TECHUB NEWS: How should a conclusion about an investment be set with conditions for invalidation?
Crazyox: The invalidation conditions essentially answer: after what facts occur, am I willing to admit that the original judgment needs to be re-evaluated, or has even ceased to be valid?
For example, if one is optimistic about a company expecting its revenue and profits to continuously improve, then one should define observing indicators, verification periods, what counts as constitutive, and under what conditions must it be reevaluated. If revenue grows while gross margin or operating profit worsens, the original core drivers may have encountered issues.
The verification building blocks in the system will preserve the research process around "observation indicators, constitutive conditions, reevaluation conditions, verification periods." The reason to write this down before making decisions is that once a position is established, it is easy for people to justify their decisions. First, define when to acknowledge a mistake, and then allow yourself to stake a bet.
TECHUB NEWS: If an investment is profitable but the research logic was actually wrong, or if there is a short-term loss but the research process is rigorous, how should they be evaluated?
Crazyox: One cannot simply evaluate by saying "making money means the judgment was right, losing means it was wrong." Therefore, Building Block Investment Research attempts to adopt a four-quadrant attribution framework: one axis is whether the original evidence has strengthened or weakened within subsequent facts; the other axis is whether the final outcome has outperformed the corresponding market benchmark.
When evidence strengthens but does not outperform the benchmark, it does not necessarily mean the research is completely wrong; it could mean the industry was correctly identified but the specific target was mischosen, or it might just be an issue of price or timing; when evidence strengthens and also outperforms the benchmark, the emphasis should be on continuing to identify which core drivers and research methods are worth reusing; when evidence weakens and results are poor, it should enter the incorrect question collection for reflecting on where the research logic went wrong; the most cautionary scenario is when evidence weakens yet one still makes a profit because this could simply be the overall market rise or sentiment rewarding a flawed method.
Making money does not equate to clear-sightedness. A very dangerous scenario in investing is making money using the wrong method. The four quadrants hope to separate ability, luck, and market Beta as much as possible, making the review extend beyond just "did I make money" to further question: was this time genuinely effective research, or was the market simply rewarding me by chance?
TECHUB NEWS: After participating in the "AI Investment Research Elite Incubation Program," what do you most hope to verify?
Crazyox: What I most want to verify now is not whether AI can write research reports—this is not much suspense. What I genuinely care about is whether AI can improve the quality of human judgment, not just increase the speed of information production.
Secondly, I want to see whether this process of "viewpoint → core driver → Skill → evidence → verification → review" can genuinely be utilized by individuals who are initially unfamiliar with systematic investment research, allowing them to form their own research framework step by step, starting from their own viewpoint.
Lastly, about the product itself: which building blocks are useful, which Skills are effective, and which designs should be dismantled and rebuilt. This is also consistent with the methodology of Building Block Investment Research—propose a hypothesis, accept counter-evidence; if wrong, amend, and if invalid, dismantle.
Conclusion
Crazyox hopes that what users retain in the long term is not a pile of research reports or a set of standard templates dictated by the product but three more enduring assets: their own research framework, their own investment incorrect question collection, and a personal cognitive evolution history written by the actual market and continuous review.
When users have used it long enough, the system may discover some rules from their historical judgments: when researching a certain type of company, which core drivers or Skills are more effective; which methods have repeatedly caused misjudgments; in what market environments, which pieces of evidence deserve more attention. At that point, the value of AI will not just be in researching a company or organizing a piece of material but in helping users to reflect and improve themselves.
"The longer I use this system, the less it seems like my Building Block Investment Research, and the more it resembles your Building Block Investment Research," Crazyox said.
In the future, information, research reports, and many standardized analysis capabilities may become increasingly cheaper. However, individual judgment abilities will not be generated automatically. True investment capability may not be about always being right but about discovering one’s errors more quickly and accurately understanding why one went wrong, leading to better judgments next time.
Editing Note: This article is based on the final draft and interview outline of TECHUB NEWS with Crazyox. The product positioning, functional mechanisms, research methods, and subsequent ideas about "Building Block Investment Research" are all based on the interviewee's statements and provided materials. The project is still in the development and verification stage, and the content herein does not constitute recommendations, forecasts, or guarantees regarding any securities, funds, asset classes, or market trends.
Disclaimer: This article is for informational exchange and discussions on investment research methods and does not constitute any investment advice. The prices of securities, funds, futures contracts, and virtual assets can rise or fall, and past performance does not represent future results. Companies, indices, ETFs, research frameworks, or cases mentioned in the text are for illustrative purposes only and do not constitute trading advice. Readers should not rely solely on the content of this article to make investment decisions and should prudently evaluate based on their own investment goals, financial situations, and risk tolerance, consulting independent professional opinions if necessary.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
