Author: a16z

Translation: Deep Tide TechFlow

a16z (Andreessen Horowitz) recently released its list of "big ideas" that may emerge in the technology sector by 2026, proposed by partners from its Apps, American Dynamism, Biotechnology, Cryptocurrency, Growth, Infrastructure, and Speedrun teams.

Below are some selected big ideas and insights from special contributors in the cryptocurrency field, covering a wide range of topics from smart agents and artificial intelligence (AI), stablecoins, tokenization and finance, privacy and security, to prediction markets and other applications. For more insights into the technological landscape of 2026, please read the full article.

Building the Future

Trading platforms are just the starting point, not the endpoint

Today, apart from stablecoins and some core infrastructure, almost all well-performing cryptocurrency companies have transformed or are transitioning into trading platforms. However, if "every crypto company becomes a trading platform," what will be the ultimate outcome? A plethora of homogeneous competition could not only distract users but may also leave only a few winners. Companies that pivot too early to trading may miss the opportunity to build more competitive and sustainable business models.

I fully understand the difficult position founders face in striving to keep their companies financially healthy, but the relentless pursuit of short-term product-market fit can come at a cost. This issue is particularly pronounced in the crypto industry, where the unique dynamics surrounding tokens and speculation often lead founders down the path of "instant gratification," akin to a "marshmallow test."

There is nothing wrong with trading—it is indeed one of the important functions of market operation—but it is not necessarily the ultimate goal. Founders who focus on the product itself and seek product-market fit from a long-term perspective may ultimately become the bigger winners.

– Arianna Simpson, General Partner of the a16z Crypto Team

New Thoughts on Stablecoins, RWA Tokenization, Payments, and Finance

Thinking about Real World Asset (RWA) tokenization and stablecoins in a more crypto-native way

We have seen banks, fintech companies, and asset management institutions show strong interest in bringing U.S. stocks, commodities, indices, and other traditional assets on-chain. However, as more traditional assets are introduced to the blockchain, their tokenization methods are often "reified"—based on existing concepts of real-world assets without fully leveraging crypto-native characteristics.

In contrast, synthetic asset forms like perpetual futures (perps) can provide deeper liquidity and are simpler to implement. Perpetual contracts also offer an easily understandable leverage mechanism, making them perhaps the most crypto-market-demand-aligned native derivatives. Emerging market stocks may be one of the most interesting asset classes to experiment with "perpifying." For example, for certain stocks, the liquidity of their "zero days to expiration" (0DTE) options market is often deeper than that of the spot market, making "perpifying" a worthwhile experiment.

Ultimately, it all comes down to the choice between "perpifying vs. tokenizing"; in any case, we have reason to expect to see more crypto-native tokenization of real-world assets in the coming year.

Similarly, in 2026, the stablecoin space will see more "issuance innovations, not just tokenization." Stablecoins have already become mainstream in 2025, and their issuance continues to grow.

However, stablecoins lacking a strong credit infrastructure resemble "narrow banks," holding specific high-liquidity assets deemed extremely safe. While narrow banks are an effective product, I do not believe they will become a long-term pillar of the on-chain economy.

We have seen many emerging asset managers, curators, and protocols begin to push for on-chain asset-backed loans, secured by off-chain collateral. Typically, these loans are generated off-chain first and then tokenized. However, I believe the advantages of this tokenization method are limited, possibly only in distributing it to users already on-chain. Therefore, debt assets should be generated directly on-chain rather than being generated off-chain and then tokenized. Generating debt assets on-chain can reduce loan servicing costs, backend structural costs, and improve accessibility. The challenge lies in compliance and standardization, but developers are already working to address these issues.

– Guy Wuollet, General Partner of the a16z Crypto Team

Stablecoins drive upgrades to bank core ledgers, opening up new payment scenarios

Today, most banks still operate outdated software systems that modern developers find hard to recognize: as early adopters of large software systems, banks were using them as far back as the 1960s and 70s. By the 80s and 90s, second-generation core banking software began to emerge (e.g., Temenos's GLOBUS and InfoSys's Finacle). However, these systems have gradually aged, and the pace of upgrades has been too slow. As a result, many key core ledgers in the banking industry—critical databases that record deposits, collateral, and other obligations—still run on mainframe computers using COBOL programming language, relying on batch file interfaces rather than modern APIs.

The majority of global assets are still stored in these decades-old core ledgers. Although these systems have been proven over time, gained the trust of regulators, and are deeply integrated into complex banking scenarios, they have also become a barrier to innovation. For instance, adding key features like real-time payments can take months or even years, facing a mountain of technical debt and complex regulatory requirements.

This is where stablecoins come into play. Over the past few years, stablecoins have found product-market fit and successfully entered the mainstream financial sector. This year, traditional financial (TradFi) institutions have embraced stablecoins with a new level of enthusiasm. Financial tools like stablecoins, tokenized deposits, tokenized government bonds, and on-chain bonds enable banks, fintech companies, and financial institutions to develop new products and serve more customers. More importantly, these innovations do not require institutions to rewrite their legacy systems—despite their aging, these systems have been running stably for decades. Thus, stablecoins provide institutions with a new avenue for innovation.

– Sam Broner

The Future of Smart Agents and AI

Using AI to perform substantive research tasks

As a mathematical economist, I found it very difficult to get consumer-grade AI models to understand my workflow earlier this year; however, by November, I was able to give the model abstract instructions as if it were a PhD student… and sometimes it would return entirely new and correctly executed answers. Not only that, we are beginning to see AI being used in broader research fields—especially in reasoning, where AI models can now not only assist directly in discovery but also autonomously solve the Putnam problem (perhaps the hardest university math exam in the world).

It remains unclear in which fields this research assistance will be most beneficial and how it will provide help. However, I anticipate that AI's research capabilities will inspire and motivate a new "polymath" research style: one that is more inclined to speculate relationships between various ideas and quickly deduce from more hypothetical answers. These answers may not be entirely accurate, but at least within certain logical frameworks, they can guide the right direction. Ironically, this approach is somewhat akin to leveraging the power of model "hallucinations": when these models become "smart" enough, allowing them to explore freely in abstract spaces, they may produce some nonsense, but sometimes they also lead to breakthrough discoveries, just as humans are most creative when they break free from linear thinking and clear directions.

Thinking about problems in this way requires a new AI workflow—not just a "agent-to-agent" model, but a more complex "agent-wrapped agent" model—where different levels of models assist researchers in evaluating the proposals of earlier models and gradually distilling valuable content from them. I have used this approach to write papers, while others have used it for patent searches, inventing new forms of art, and even (unfortunately) discovering new smart contract attack vectors.

However, running this "wrapped reasoning agent" research model requires better interoperability between models and finding a way to identify and reasonably compensate each model's contributions—issues that cryptographic technology can help solve.

– Scott Kominers, Member of the a16z Crypto Research Team, Professor at Harvard Business School

The invisible tax imposed by AI agents on open networks

With the rise of AI agents, an "invisible tax" is pressuring open networks and fundamentally disrupting their economic foundations. This disruption stems from the increasing asymmetry between the context layer and execution layer of the internet: currently, AI agents extract data from ad-supported content websites (context layer) to provide convenience to users while systematically bypassing the revenue sources that support content creation (such as advertising and subscriptions).

To prevent further decline of open networks (and to protect the diverse content that fuels AI), we need to deploy technological and economic solutions on a large scale. This may include next-generation sponsored content, micro-attribution systems, or other innovative funding models. Existing AI licensing agreements have proven to be only temporary stopgaps, often compensating content providers for only a small fraction of the revenue lost due to AI traffic encroachment.

The network needs a brand new techno-economic model that allows value to flow automatically. The most critical shift next year will be transitioning from static licensing models to compensation models based on real-time usage. This means testing and expanding systems—potentially utilizing blockchain-supported micropayments and complex attribution standards—to automatically reward every entity contributing information for tasks successfully completed by AI agents.

– Liz Harkavy, a16z Crypto Investment Team

Privacy is the Moat

Privacy will become the most important moat in the cryptocurrency field

Privacy is one of the key characteristics driving the on-chain transition of global finance. However, it is also an important element that is almost universally lacking in today's blockchains. For most blockchains, privacy issues are often considered an afterthought.

Today, privacy itself has become a key differentiating feature of blockchains. More importantly, privacy can create a "chain lock-in" effect, or a privacy network effect. Especially in an era where performance competition is no longer a sufficient advantage, privacy becomes particularly crucial.

With cross-chain bridge protocols, as long as all information is public, it becomes very easy for users to migrate between different chains. However, once privacy is introduced, this convenience disappears: transferring tokens across chains is easy, but transmitting privacy across chains is extremely difficult. When users enter or exit a privacy chain, whether switching to a public chain or another privacy chain, they face risks because those observing on-chain data, mempool, or network traffic may infer the user's identity. Crossing the boundary between privacy chains and public chains, or even between two privacy chains, can leak various metadata, such as the correlation of transaction times and amounts, making it easier to track users.

Compared to many homogeneous new chains, the transaction fees of these chains may be driven down to nearly zero due to competition, while blockchains with privacy features can form stronger network effects. The reality is that if a "general-purpose" blockchain does not have a mature ecosystem, killer applications, or unfair distribution advantages, there is almost no reason for users to choose to use or build on it, let alone develop loyalty towards it.

On public blockchains, users can easily transact with others on different chains—the chain they join does not matter. However, on private blockchains, the chain that users choose to join becomes particularly important, as once they join, they are less likely to switch to another chain to avoid the risk of privacy exposure. This phenomenon creates a "winner-takes-all" dynamic. And since privacy is crucial for most real-world application scenarios, a few privacy chains may ultimately dominate the cryptocurrency space.

– Ali Yahya, General Partner of the a16z Crypto Team

Other Industries and Applications

Prediction markets will become larger, broader, and smarter

Prediction markets have gradually entered the mainstream, and in the coming year, as they intersect with cryptocurrency technology and artificial intelligence (AI), they will become larger in scale, more widely applied, and smarter, while also presenting new significant challenges for developers.

First, there will be more contracts listed in prediction markets. This means we can not only obtain real-time odds for major elections or geopolitical events but also predict various nuanced outcomes and complex cross-events. As these new contracts continuously extract more information and gradually integrate into the news ecosystem (a trend that has already begun), they will raise important social issues, such as how to balance the value of information and how to better design these markets to make them more transparent and auditable—issues that can be addressed through cryptographic technology.

To cope with the influx of new contracts, we need new ways to reach consensus on real events to resolve these contracts. Centralized platform solutions (such as confirming whether a certain event has actually occurred) are certainly important, but controversial cases like the Zelensky lawsuit market and the Venezuelan election market have also exposed their limitations. To address these edge cases and help prediction markets expand into more practical application scenarios, new decentralized governance mechanisms and large language model (LLM) oracles can assist in determining the truth of disputed outcomes.

The potential of AI is not limited to LLM-driven oracles. For example, AI agents active on these platforms can gather signals globally, gaining short-term trading advantages. This not only helps us view the world from a new perspective but also allows for more accurate predictions of future trends. (Projects like Prophet Arena have already generated excitement in this field.) Besides serving as complex political analysts providing insights, these AI agents may also reveal fundamental predictive factors of complex social events as we study the strategies they emerge.

Will prediction markets replace opinion polls? No. Instead, they will make opinion polls better (and the information from opinion polls can also be input into prediction markets). As a political economy professor, I am most excited about the potential for prediction markets to work in synergy with the rich and diverse ecosystem of opinion polls—but we need to rely on new technologies, such as AI, which can improve the experience of surveys; and cryptographic technology, which can provide new ways to verify that survey and questionnaire participants are human rather than bots.

– Andy Hall, a16z Crypto Research Advisor, Professor of Political Economy at Stanford University

Cryptographic technology will expand to new applications beyond blockchain

For years, SNARKs (Zero-Knowledge Succinct Non-Interactive Arguments of Knowledge, a cryptographic proof that verifies correctness without re-executing computations) have primarily been used in the blockchain space. This is due to their enormous computational overhead: proving the workload of a computation can be up to a million times more than directly running that computation. In scenarios where the cost can be distributed among thousands of validators, this overhead is worthwhile, but in other scenarios, it becomes impractical.

This situation is about to change. By 2026, the computational overhead of zkVM (Zero-Knowledge Virtual Machine) provers will be reduced to about 10,000 times, while their memory usage will only require a few hundred megabytes—fast enough to run on mobile devices and cheap enough to be widely applied in various scenarios. Here’s a reason why "10,000 times" may be a critical threshold: the parallel throughput capacity of high-end GPUs is about 10,000 times that of laptop CPUs. By the end of 2026, a single GPU will be able to generate real-time proofs of computations executed by CPUs.

This will unlock the visions proposed in some early research papers: verifiable cloud computing. If you are already running CPU workloads in the cloud (because your computational tasks are not sufficient to be accelerated by GPUs, or you lack the relevant expertise, or due to historical reasons), you will be able to obtain cryptographic proofs of computational correctness at reasonable cost. Moreover, the prover has been optimized for GPUs, so your code does not require additional adjustments.

– Justin Thaler, Member of the a16z Crypto Research Team, Associate Professor of Computer Science at Georgetown University

—— a16z Crypto Editorial Team

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。