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

Written by: Frank, Researcher at MaiTong MSX

With the same data, do you trust the human brain's interpretation or AI's "fortune-telling"?

It sounds like a cyberpunk joke, but in today's world where opinions are as numerous as cattle and highly emotional, what the market lacks is precisely the unembellished real sample information.

After all, positions can indeed make people lie.

At the end of the year, in order to get as close as possible to the truth of tokenization/Web3 at the turning point of 2025-2026, we conducted an anonymous sampling of the first-line Builder community at MaiTong MSX, focusing on three main questions:

• The keywords you associate with the US stock market/tokenization in 2025;
• Your actual holdings in 2025 (not limited to US stocks/Web3);
• The directions you are most willing to allocate in 2026 (not limited to US stocks/Web3);

After some "coercion and inducement," we managed to gather a reasonably satisfactory sample size and compressed it into three word clouds that are almost devoid of subjective judgment. Considering that human interpretation inevitably mixes in biases, we decided to invite "pure" ChatGPT, Gemini, and DeepSeek to "fortune-tell" these three word clouds under completely consistent prompt constraints.

Perhaps AI understands this generation of Builders better than we do and can piece together the fate puzzle of the US stock market, Web3, and tokenization in 2026 through these fragments.

1. What are the keywords for the US stock market/tokenization in 2025 in your view?

Prompt: You will see a set of keyword samples from first-line practitioners in the US stock market/tokenization/Web3, used to answer the same question: "What does the US stock market/tokenization field look like in 2025 in your view?"; do not judge right or wrong, nor introduce any external market information, but interpret based solely on these keywords from the perspective of data structure and linguistic features, focusing on:

Which keywords have the highest/lowest frequency or are relatively abrupt, and what structural state might they reflect? Which keywords or semantic directions show obvious consensus characteristics, and are there combinations of keywords with similar meanings, opposite directions, or strong tensions? What overall stage characteristics do these keywords suggest for the track? If based solely on these keywords, without introducing any external information, how would you understand this track's 2025 and 2026?

Please maintain a descriptive and restrained analytical tone, avoiding investment advice or definitive judgments, with a response not exceeding 1000 words.

++Editor's note: The response has been slightly edited and polished for readability, but the overall direction and thought have not been significantly altered (same below).++

ChatGPT: The highest frequency keywords are concentrated around "acceleration" and "institutionalization," suggesting that the track is undergoing a significant change in pace but has not yet stabilized; mid-frequency keywords include "AI," "compliance," and "differentiation," reflecting a backgrounded and normalized existence; low-frequency or isolated words like "weightlessness," "unknown," "battleground," and "starting point" carry strong subjective feelings or emotional metaphors, possibly reflecting individuals' discomfort, competitive feelings, or direct projections of uncertainty regarding structural changes.

The overall corpus presents a linguistic characteristic of a "mid-early stage - structural building phase," where high-frequency verbs outnumber nouns, indicating that the focus of discussion remains on processes and changes rather than stable forms; multiple keywords point to rules, validation, and processes rather than specific products or models; emotional vocabulary exists but does not dominate, indicating that uncertainty is perceived but has not spiraled out of control.

If based solely on these keywords, 2025 can be understood as a transitional year characterized by acceleration but not yet defined, with consensus gathering while differentiation is also occurring.

Gemini: High-frequency words include "acceleration" and "institutionalization," indicating a strong consensus on growth following professional intervention; mid-frequency words like "AI," "compliance," and "differentiation" outline a state of uneven technological means, regulatory frameworks, and market performance; relatively abrupt low-frequency words include "weightlessness," "process-oriented," and "asset verification," reflecting underlying anxieties beneath the surface of prosperity and a micro perspective shifting from narrative to implementation.

The overall data presents a "high momentum, low center of gravity" state, where the coexistence of "institutionalization" and "compliance" implies that this acceleration is not blind expansion but a structural shift constrained by rules, forming a certain tone for the industry.

In summary, the track is currently in a "transitional explosive period," combining the chaos of the starting phase (weightlessness, unknown) with the rigor of the eve of maturity (compliance, process-oriented), indicating that tokenization is no longer an isolated asset experiment but is beginning to deeply couple with mainstream finance and cutting-edge technology.

Overall, from 2025 to 2026, this is a moment of running anxiously, reshaping within rules, and heavily relying on technological validation for professional entry.

DeepSeek: "Acceleration" occupies the absolute core, while "institutionalization," "differentiation," "compliance," and "AI" form the second tier; most vocabulary appears only once, presenting a discrete long-tail characteristic, for example, "weightlessness" appears relatively abrupt in the overall context leaning towards dynamic and positive vocabulary (like acceleration, opening up, breaking barriers), possibly metaphorically indicating a temporary loss of traditional reference points in rapid change.

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 levels or paths separating within the market or track.

The dynamic nature of track development (acceleration, variability, fluctuations) and the institutionalization process (institutionalization, compliance, process-oriented, structural) form the two strongest consensus semantic directions.

Both run parallel, depicting a field seeking to establish rules and structures amid rapid changes.

2. What are your actual holdings in 2025 (not limited to US stocks/Web3)?

Prompt: You will see a set of keyword samples from first-line practitioners in the US stock market/tokenization/Web3, used to answer the same question: "What are your actual holdings in 2025 (not limited to US stocks/Web3)"; do not judge right or wrong, nor introduce any external market information, but conduct a logical analysis based on the composition of the samples, interpreting from the perspective of data structure:

Classify the assets involved by type/track; are there core assets that are repeatedly allocated? Is the overall holding more concentrated or diversified, revealing what kind of risk tolerance? From these actual holdings, can any behavioral characteristics be observed?

Please maintain a descriptive and restrained analytical tone, avoiding investment advice or definitive judgments, with a response not exceeding 1000 words.

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

• Crypto native assets: High-frequency words include BTC, ETH, SOL, BNB, USDT, USDC, while secondary/marginal ones include XRP, DOGE, ASTER, SUPRA, XTTA; this category occupies a clear advantage in quantity and repetition, containing both mainstream assets and exploratory allocations;
• US stocks related to technology and finance: Including tech/platform types—GOOGL, AAPL, AMZN, NVDA, TSLA, PLTR, INTC, TSM, MU, WDC; also includes finance/trading related—COIN, HOOD; showing a preference for infrastructure, platforms, and technological capabilities;
• US consumer and traditional defensive stocks: MCD, KO, LLY, JD, etc., not many in number but clearly playing a structural hedging or stabilizing role;
• Indices and cash-like/hedging assets: VOO, gold, USDT, USDC, etc., providing liquidity and volatility buffers within the overall portfolio;

From the frequency of repetition, there is a very clear core asset layer:

• BTC is the most frequently mentioned single asset in the sample, appearing multiple times and densely;
• GOOGL has the highest repetition on the US stock side, significantly higher than other tech stocks;
• ETH, SOL, and BNB form a second layer of repeatedly allocated crypto asset combinations;
• USDT, as a non-risk asset, is also mentioned multiple times, indicating it is not merely a "transitional existence";

This structure indicates that respondents are not evenly distributed in their holdings but are repeatedly adding to a few high-confidence assets while maintaining diverse attempts on the periphery.

If based solely on this set of actual holding keywords, without introducing any external judgments, it can be understood as a combination structure anchored by a few core assets, parallelly allocated across multiple asset systems, while retaining liquidity and exploratory space.

Gemini: Through a preliminary sorting of the sample data, these holdings can be categorized into the following five major categories.

The data structure shows a significant phenomenon of consensus centralization within this group. BTC is the most mentioned asset, followed by GOOGL, ETH, and SOL, forming the "four corner anchors" of practitioners' holdings in 2025.

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

Despite the diversity of involved assets, about 50% of the sample capacity is occupied by the top 5-6 core assets, indicating that practitioners' strategies in 2025 are not scattergun allocations but rather highly concentrated bets on blue-chip assets with strong narrative support.

Specifically, the risk tolerance can be summarized as a barbell strategy, where one end consists of highly volatile tech growth and crypto assets, such as RKLB (aerospace), SOL, and various semiconductor stocks; the other end consists of extremely high liquidity or certainty, such as USDT, gold, and VOO (S&P 500 ETF).

Although there are traditional value stocks like MCD (McDonald's) and KO (Coca-Cola) in the sample, they serve merely as embellishments, indicating that in the logic of holdings in 2025, the status of traditional dividend-type assets has been marginalized, with practitioners more inclined to define "value return" as "technology growth supported by cash flow."

DeepSeek: The holdings in the sample can be roughly categorized into three types:

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

Core assets (BTC, GOOGL, ETH, SOL) account for about 37% of total mentions, resulting in a "core concentration, peripheral dispersion" characteristic of the holding structure, meaning heavier allocation on core assets while diversifying investments across many other targets.

Overall, the data structure of this holding sample shows a configuration mindset seeking balance between risk and opportunity, anchoring core tracks while constructing buffers through diversification and hedging tools.

3. What directions are you most willing to allocate in 2026 (not limited to US stocks/Web3)?

Prompt: You will see a set of keyword samples from first-line practitioners in the US stock market/tokenization/Web3, used to answer the same question: "What directions are you most willing to allocate in 2026 (not limited to US stocks/Web3)"; do not judge right or wrong, nor introduce any external market information, but conduct a logical analysis based on the composition of the samples, interpreting from the perspective of data structure:

Classify the assets involved by type/track; which tracks/targets do these investors prefer in 2026, are there core assets that are repeatedly allocated? Is the overall holding more concentrated or diversified, revealing what kind of risk tolerance? From these actual holdings, can any behavioral characteristics be observed? What are the most consensual assets/tracks in 2025 and 2026, and what changes have occurred?

Please maintain a descriptive and restrained analytical tone, avoiding investment advice or definitive judgments, with a response not exceeding 1000 words.

ChatGPT: Compared to 2025, the keywords for 2026 show a clear change: specific assets and abstract tracks begin to appear side by side, indicating that respondents emphasize directionality and structural judgment when expressing allocation intentions, rather than focusing solely on individual asset choices.

The samples can be roughly divided into the following categories:

• Crypto native assets: BTC, ETH, and SOL are core high-frequency words, while stablecoins and MEME crypto assets still occupy the center of the sample, but show clear internal stratification;
• US stocks in technology and high-tech directions: Individual stocks focus on TSLA, GOOGL, NVDA, AMZN, PLTR, HOOD, while track-level focus is on tech stocks, chips, storage, and aerospace, showing a clear transition from "stock picking" to "industry direction";
• Commodities and physical assets: Precious metals (gold, silver, copper), oil, which were not prominent in the 2025 sample, are significantly mentioned in the 2026 intentions;
• Indices and hedging tools: S&P, SQQQ represent attention to overall market or systemic volatility;

In terms of repetition, the core for 2026 is very concentrated: BTC remains the most frequently mentioned keyword, far exceeding other assets; TSLA and NVDA appear repeatedly on the US stock side, significantly higher than other individual stocks.

If based solely on this set of keywords, the allocation intention for 2026 can be understood as a structural allocation tendency centered around a few highly consensual core assets and industry directions, expressed through track-based, resource-based, and tool-based approaches to cope with uncertainty.

Gemini: The sample data for 2026 shows a broader span than that of 2025, which can be summarized into the following four categories:

BTC is mentioned in over half of the 2026 samples, being the only asset with a characteristic of consensus among all; TSLA and RKLB/aerospace follow closely, indicating a significant increase in practitioners' willingness to allocate to hard tech and space exploration compared to 2025, suggesting that investment perspectives are extending from the digital realm to the extreme frontiers of the physical world.

The holding structure presents a "one extreme, multiple strengths" barbell structure, with one extreme being the highly concentrated BTC, and the other extreme being extremely diversified hard tech assets (involving aerospace, storage, semiconductors, and other subfields).

The risk tolerance shows "extreme aggressiveness and extreme defensiveness coexist," with allocations to high-volatility assets like MEME and RKLB on one hand, and on the other hand, introducing hedging tools like SQQQ, gold, silver, copper, and oil, indicating that practitioners are not expecting a one-sided rise in 2026 but rather structural opportunities amid severe volatility, with their risk tolerance shifting from merely chasing growth to managing uncertainty.

Overall, the density of keywords in 2026, including gold, silver, copper, oil, storage, and aerospace, has significantly increased, reflecting that practitioners, after delving into the digital track (Web3), are beginning to align their logic with tangible assets and underlying energy/resources.

Deep diving into the industry chain is also a typical feature, as respondents are no longer satisfied with merely allocating to leading chip stocks like NVDA but are extending down to storage and specific hardware segments like MU, WDC, SNDK, showing a professional deep-digging behavior based on industry cognition.

If 2025 was about practitioners "building roads" (focusing on compliance, platforms, and infrastructure), then 2026 is about "expedition" (focusing on aerospace, deep hardware, and global bulk resources).

DeepSeek: From the frequency statistics, BTC has the highest frequency, accounting for 20.4% of total mentions, followed by TSLA (9.3%), with the two together accounting for about 30%, indicating relative concentration at the top preference. NVDA, RKLB, GOOGL, HOOD, stablecoins, and precious metals are also mentioned multiple times, but overall remain dispersed across numerous assets and themes, presenting a "core - periphery" structure—core highly concentrated in a few assets, while the periphery is widely dispersed.

Among them, cryptocurrencies (especially BTC) and tech stocks (TSLA, NVDA, etc.) are the main directions, while thematic tracks like aerospace (represented by RKLB), chips, storage, and hedging assets like precious metals and stablecoins are also receiving attention.

Notably, the allocation of stablecoins, precious metals (gold, silver, copper), oil, and SQQQ is more pronounced than in the 2025 holding sample, indicating a defensive mindset towards macroeconomic uncertainty or market corrections.

At the individual stock/asset level, compared to 2025, BTC's consensus has further strengthened, maintaining its absolute core; while the consensus for GOOGL, ETH, and SOL has significantly decreased, and the consensus for TSLA and NVDA has increased, which may reflect a shift in allocation focus from diversified cryptocurrencies and large tech stocks to more specific leading assets.

At the track level, the allocation direction has expanded to thematic tracks (aerospace, chips, storage) and commodities (precious metals, oil), with consensus spreading from specific assets to diversified themes.

Overall, the data structure shows that from 2025 to 2026, the allocation consensus has evolved from broadly laying out cryptocurrencies and tech stocks to reinforcing Bitcoin as the core, focusing on specific tech assets, and increasing thematic investments and hedging allocations, reflecting adjustments in investor strategies amid the evolution of the track: seeking a balance between growth and defense in uncertainty.