We used a large model to "predict" the future of tokenization/Web3 in 2026.

Written by: Frank, MSX Researcher at Maiton

Given the same data, do you trust human interpretation or AI "fortune telling"?

It sounds like a cyberpunk joke, but in today's world where opinions are as numerous as hairs and highly emotional, what the market lacks is precisely that kind of real sample information that doesn't have time to be embellished.

After all, position size really can make people lie.

At the turn of the year, in order to get as close as possible to the truth about tokenization/Web3 at the turning point of 2025-2026, we conducted an anonymous sampling of front-line Builders at Maiton MSX, mainly focusing on three questions:

• What are your keywords for US stocks/tokenization in 2025?
• Your actual holdings in 2025 (excluding US stocks/Web3);
• What sectors are you most willing to allocate your portfolio to in 2026 (not limited to US stocks/Web3)?

After some "coercion and enticement," a reasonably satisfactory sample size was collected and compressed into three word clouds that were almost entirely free of subjective judgment. Considering that human interpretation inevitably carries bias, we decided to invite the "pure" ChatGPT, Gemini, and DeepSeek to "read" these three word clouds under completely consistent prompt constraints.

Perhaps AI really does understand this generation of Builders better than we do, and can better piece together the fate of the US stock market, Web3, and tokenization in 2026 from these fragments.

1. What are the key words for US stock market/tokenization in 2025 in your eyes?

Prompt: You will see a sample of keywords from frontline practitioners in the US stock market tokenization/Web3 sector. These keywords are used to answer the same question: "What will the US stock market/tokenization sector look like in 2025?" Do not judge their accuracy or introduce any external market information; simply interpret the keywords themselves from the perspective of data structure and language features, focusing on the following answers:

Which keywords have the highest/lowest/most abrupt frequency, and what structural state might they reflect? Which keywords or semantic directions exhibit clear consensus characteristics? Are there keyword combinations with similar meanings, opposite directions, or strong tensions? What stage characteristics of the market are these keywords collectively suggesting? If you only consider these keywords, without introducing any external information, how would you interpret the market outlook for 2025 and 2026?

Please maintain a descriptive and restrained analytical tone, avoid giving investment advice or making definitive judgments, and keep your reply to no more than 1,000 words.

Editor's Note: The reply has been slightly edited and polished to enhance readability, but the overall direction and ideas have not been significantly adjusted (the same applies below).

ChatGPT: The most frequent keywords are concentrated on "acceleration" and "institutionalization," suggesting that the track is undergoing a process of significant change in pace but has not yet stabilized; the mid-frequency keywords include "AI," "compliance," and "differentiation," reflecting a backgrounded and normalized existence; low-frequency or isolated words such as "weightlessness," "unknown," "battleground," and "start" carry strong subjective feelings or emotional metaphors, which may reflect individuals' discomfort, sense of competition, or uncertainty about structural changes.

The overall corpus exhibits linguistic characteristics of the "early-middle stage - structure building stage," with more high-frequency verbs than nouns, indicating that the focus of the discussion is still on process and change rather than stable form; multiple keywords point to rules, verification, and processes rather than specific products or models; emotional words are present but have not become dominant, indicating that uncertainty is perceived but has not yet spiraled out of control.

Based solely on these keywords, 2025 can be understood as a transitional year of acceleration but not yet finalized, where consensus is gathering but also diverging.

Gemini: The high-frequency words "acceleration" and "institutionalization" indicate a strong consensus on growth after the intervention of professional forces; the mid-frequency words "AI", "compliance", and "differentiation" outline the uneven state of technological means, regulatory frameworks and market performance; the more abrupt low-frequency words "weightlessness", "process-oriented", and "asset verification" reflect the underlying anxiety under the prosperity and the micro-perspective of shifting from narrative to implementation.

The overall data presents a state of "high momentum and low center of gravity". The coexistence of "institutionalization" and "compliance" means that this acceleration is not blind expansion, but a structural displacement constrained by rules, which also constitutes the certainty of the industry.

Based on the overall characteristics of the keywords, this sector is currently in a "transitional explosive period." It combines the chaos (weightlessness, uncertainty) of the initial stage with the rigor (compliance, process-oriented) of the eve of maturity. The market is no longer monolithic but shows obvious "differentiation," indicating that tokenization is no longer an isolated asset experiment but has begun to be deeply coupled with mainstream finance and cutting-edge technology.

Overall, from 2025 to 2026, it will be a time of frantic activity amidst anxiety, reshaping of rules, and a period of intense reliance on technological validation for professional entry into the market.

DeepSeek: "Acceleration" occupies the absolute core, while "institutionalization", "differentiation", "compliance" and "AI" form the second tier; most words appear only once, showing discrete long-tail characteristics. For example, "weightlessness" seems rather abrupt in the overall dynamic and positive vocabulary context (such as acceleration, opening, breaking the circle), which may metaphorically imply a state of temporarily losing the traditional frame of reference in a rapidly changing environment.

The terms "institutionalization" and "compliance" are closely related, both pointing to the trend of deep involvement of traditional financial system rules and participants; "differentiation" suggests that there may be separation of levels or paths within the market or sector.

The dynamic nature of track development (acceleration, variability, and volatility) and the institutionalization process (institutionalization, compliance, process-orientation, and structuring) constitute the two strongest sets of consensus semantic directions.

Both, operating in parallel, depict a field seeking to establish rules and structures amidst rapid change.

II. Your actual holdings in 2025 (excluding US stocks/Web3)

Prompt: You will see a sample of keywords from frontline practitioners in the US stock tokenization/Web3 industry. These keywords are used to answer the same question: "What are your actual holdings in 2025 (not limited to US stocks/Web3)?" Do not judge the correctness of the statements, nor introduce any external market information. The analysis will be based solely on the composition of the samples and interpreted from a data structure perspective.

Categorize the assets/sectors involved; are there any core assets that are repeatedly allocated? Is the overall portfolio more concentrated or diversified, revealing what kind of risk tolerance? What behavioral characteristics can be observed from these actual holdings?

Please maintain a descriptive and restrained analytical tone, avoid giving investment advice or making definitive judgments, and keep your reply to no more than 1,000 words.

ChatGPT: From the sample structure, the holdings can be clearly divided into several asset clusters.

• Crypto native assets: High-frequency terms include BTC, ETH, SOL, BNB, USDT, and USDC; secondary/edge assets include XRP, DOGE, ASTER, SUPRA, and XTTA. This category has a clear advantage in terms of quantity and repetition, and includes both mainstream assets and exploratory allocations.
• US stock technology and finance related stocks include both technology/platform stocks—GOOGL, AAPL, AMAZN, NVDA, TSLA, PLTR, INTC, TSM, MU, WDC—and finance/trading related stocks—COIN, HOOD; showing a preference for infrastructure, platform, and technology capabilities.
• US consumer and traditional defensive stocks, such as MCD, KO, LLY, and JD, are few in number but clearly play a role in structural hedging or stabilization.
• Indices and cash-like/safe havens: VOO, gold, USDT, USDC, etc., provide liquidity and volatility buffers in the overall portfolio;

Based on the frequency of recurrence, a very clear core asset layer exists:

• BTC is the single asset that appears most frequently in the sample, and it appears multiple times and in a dense manner.
• GOOGL has the highest degree of overlap in the US stock market, significantly higher than other tech stocks;
• ETH, SOL, and BNB constitute a layer-two, repeatedly configured portfolio of crypto assets;
• USDT has been repeatedly mentioned as a non-risky asset, indicating that it is not a "transitional asset";

This structure suggests that the respondents did not distribute their holdings evenly, but rather repeatedly layered them on a few high-confidence assets while maintaining diversification in the periphery.

Based solely on this set of real holding keywords, without introducing any external judgment, it can be understood as a combination structure that uses a few core assets as anchors, allocates assets in parallel across multiple asset systems, and retains liquidity and room for exploration .

Gemini: Based on a preliminary analysis of the sample data, these holdings can be divided into the following five categories.

The data structure reveals a highly centralized consensus phenomenon within this group. BTC is the most frequently mentioned asset, followed by GOOGL, ETH, and SOL, which together form the "four corner anchors" of practitioners' holdings in 2025.

The high frequency of BNB, TSLA, and USDT reflects a dual consideration of industry ecosystem position and liquidity while pursuing high beta returns.

Despite the diverse range of targets involved, approximately 50% of the sample size was occupied by the top 5-6 core targets. This indicates that practitioners' strategy in 2025 was not to allocate indiscriminately, but rather to heavily invest in blue-chip assets with strong narrative support.

Specifically, the risk-taking approach can be summarized as a barbell strategy, where one end is highly volatile technology growth and crypto assets, such as RKLB (aerospace), SOL, and various semiconductor assets; the other end is highly liquid or certain, such as USDT, gold, and VOO (S&P 500 ETF).

Although the sample includes traditional value stocks such as McDonald's and Coca-Cola, they are only used as embellishments. This indicates that in the holding logic of 2025, the status of traditional dividend-paying assets has been marginalized, and practitioners tend to define "value return" as "technology growth supported by cash flow".

DeepSeek: The holdings in the sample can be roughly divided into three categories:

• Cryptocurrencies: including BTC, ETH, SOL, BNB, XRP, DOGE, etc. These assets appear frequently and cover mainstream and some altcoins;
• Stablecoins and commodities: Stablecoins such as USDT and USDC, and commodities such as gold, are typically associated with hedging or liquidity management due to their low price volatility;
• The US stock market covers multiple sectors and market capitalizations, including: technology stocks such as GOOGL (Google), AAPL (Apple), TSLA (Tesla), NVDA (Nvidia), COIN (Coinbase), PLTR (Palantir), INTC (Intel), MU (Micron Technology), and TSM (TSMC); consumer and traditional industry stocks such as MCD (McDonald's), KO (Coca-Cola), and LLY (Eli Lilly); financial and platform stocks such as HOOD (Robinhood) and FIG (possibly a financial group); and ETFs such as VOO (S&P 500 ETF).

Core assets (BTC, GOOGL, ETH, SOL) accounted for approximately 37% of the total occurrences, resulting in a portfolio structure characterized by "concentrated core assets and diversified peripheral assets," meaning that there is a heavy allocation to core assets while investments are diversified across numerous other assets.

Overall, the portfolio data structure shows an allocation strategy that seeks a balance between risk and opportunity, anchoring core sectors while building buffers through diversification and hedging tools.

3. What sectors would you most like to allocate your portfolio to in 2026 (excluding US stocks/Web3)?

Prompt: You will see a sample of keywords from frontline practitioners in the US stock tokenization/Web3 industry. These keywords are used to answer the same question: "What sectors are you most willing to allocate to in 2026 (not limited to US stocks/Web3)?" Do not judge whether they are right or wrong, and do not introduce any external market information. Just perform logical analysis based on the composition of the sample and interpret it from the perspective of data structure.

The relevant targets/sectors are categorized by asset classification. In 2026, which sectors/targets did these investors prefer? Were there any core assets repeatedly allocated to them? Was the overall portfolio more concentrated or diversified, revealing their risk tolerance? What behavioral characteristics can be observed from these actual holdings? What assets/sectors had the highest consensus in 2025 and 2026, and how did they change?

Please maintain a descriptive and restrained analytical tone, avoid giving investment advice or making definitive judgments, and keep your reply to no more than 1,000 words.

ChatGPT: Compared to 2025, the keywords for 2026 show a significant change: specific targets and abstract sectors began to appear side by side, indicating that respondents were emphasizing directional and structural judgments rather than single asset selections when expressing their allocation intentions.

The samples can be roughly divided into the following categories:

• Crypto native assets: BTC, ETH, and SOL are the core high-frequency terms, while stablecoins and MEME crypto assets still occupy the center of the sample, but they show a clear stratification.
• In the US stock market, the focus is on technology and high-tech sectors: individual stocks include TSLA, GOOGL, NVDA, AMZN, PLTR, and HOOD; sectors include technology stocks, chips, memory, and aerospace. The shift from "stock selection" to "industry focus" is very obvious.
• Commodities and physical assets: Precious metals (gold, silver, copper) and oil were not prominent in the 2025 sample, but were clearly and systematically mentioned in the 2026 intentions;
• Indices and hedging tools: S&P 500 and SQQQ represent attention to overall market or systemic volatility;

In terms of repetition, the core themes of 2026 are very concentrated: BTC remains the most frequently appearing keyword, far exceeding other assets; TSLA and NVDA appear repeatedly in the US stock market, significantly more than other individual stocks.

Based solely on this set of keywords, the allocation intentions in 2026 can be understood as a structural allocation tendency centered around a few highly consensus-driven core assets and industry directions, while being expressed through track-based, resource-based, and tool-based approaches to cope with uncertainty.

Gemini: The sample data for 2026 spans a wider range than that for 2025, and can be mainly summarized into the following four categories:

BTC was mentioned more than half of the samples in 2026, making it the only asset with a consensus among all investors; TSLA and RKLB/Aerospace followed closely behind. Compared to 2025, practitioners' willingness to allocate to hardcore technology and outer space exploration has significantly increased, indicating that the investment focus is shifting from digital to the extreme frontiers of the physical world.

The holding structure presents a dumbbell-shaped structure of "one pole and many strong ones". One pole is the extremely concentrated BTC, and the other pole is the extremely dispersed hard technology targets (involving multiple sub-sectors such as aerospace, storage, and semiconductors).

The risk-taking approach is characterized by a combination of "extreme aggressiveness and extreme defensiveness." On the one hand, high-volatility assets such as MEME and RKLB are allocated, while on the other hand, hedging tools such as SQQQ, gold, silver, copper, and oil are introduced. This indicates that practitioners do not expect a one-sided rise in 2026, but rather structural opportunities amidst violent fluctuations. Their risk-taking approach has shifted from simply pursuing growth to managing uncertainty.

Overall, the keywords for 2026 show a significant increase in the density of gold, silver, copper, oil, storage, and aerospace. This reflects that after deep cultivation in the digital track (Web3), practitioners are beginning to focus their logic on tangible assets and underlying energy/resources.

A key characteristic is the in-depth exploration of the industry chain. Respondents are no longer satisfied with just allocating to leading chip stocks such as NVDA, but extend downwards to storage and specific hardware links such as MU, WDC, and SNDK, demonstrating a professional and in-depth exploration behavior based on industry knowledge.

If 2025 was a year of "road building" for industry players (focusing on compliance, platforms, and infrastructure), then 2026 will be a year of "expedition" (focusing on aerospace, deep hardware, and global bulk resources).

DeepSeek: According to word frequency statistics, BTC was mentioned most frequently, accounting for 20.4% of the total mentions, followed closely by TSLA (9.3%). Together, they accounted for about 30%, indicating a relatively concentrated preference among top-ranked assets. NVDA, RKLB, GOOGL, HOOD, stablecoins, and precious metals were also mentioned multiple times, but overall, they were still scattered across numerous assets and themes, presenting a "core-periphery" structure—the core was highly concentrated in a few assets, while the peripheral was widely dispersed.

Cryptocurrencies (especially BTC) and technology stocks (TSLA, NVDA, etc.) are the main focus, while thematic sectors such as aerospace (represented by RKLB), chips, storage, and safe-haven assets such as precious metals and stablecoins are also attracting attention.

It is worth noting that the allocation to stablecoins, precious metals (gold, silver, copper), oil, and SQQQ is more explicit than in the 2025 holdings sample, indicating a defensive mentality against macroeconomic uncertainties or market corrections.

At the individual stock/target level, compared to 2025, the consensus of BTC has been further strengthened, maintaining its absolute core position; while the consensus of GOOGL, ETH, and SOL has decreased significantly, while the consensus of TSLA and NVDA has increased. This may reflect a shift in allocation focus from dispersed cryptocurrencies and large-cap technology stocks to more specific leading targets.

At the sector level, the allocation direction has expanded to thematic sectors (aerospace, chips, storage) and commodities (precious metals, oil), and the consensus has spread from specific targets to diversified themes.

Overall, the data structure shows that from 2025 to 2026, the consensus on asset allocation has evolved from broadly investing in cryptocurrencies and technology stocks to strengthening Bitcoin's core, focusing on specific technology targets, and increasing thematic investment and hedging allocations. This reflects the adjustment of investors' strategies in the evolution of the sector: seeking a balance between growth and defense amidst uncertainty.