As we approach 2026, the Archetype team is focusing on future technology trends.
Application Chains Are Viable
—Aadharsh Pannirselvam
The reasoning is simple: chains that are carefully designed, built, and optimized for applications will inevitably deliver amazing experiences. And the best application chains next year will innovate from fundamental modules and first principles.
The recent influx of developers, users, institutions, and capital is significantly different from previous groups entering the on-chain ecosystem: they are more focused on practical experience rather than abstract concepts like decentralization and censorship resistance. In practice, this cultural demand sometimes aligns with existing infrastructure and sometimes conflicts with it.
For applications like Blackbird or Farcaster that target general users and hide the technical details of cryptography, certain aspects of the user experience are particularly important. Even some centralized design decisions that were considered heretical three years ago, such as node co-location, a single sequencer, and customized databases, are now seen as reasonable choices. The same applies to projects like Hyperliquid and GTE, whose success or failure often depends on millisecond-level speed, minimum price increments, and optimal pricing.
But this doesn't apply to all new applications.
For example, although people feel comfortable with centralization, there is also a counterbalancing force: a growing number of institutions and individuals are paying attention to privacy. The needs and usage experiences of crypto applications can be vastly different, so the required infrastructure should also vary.
Fortunately, creating a specific chain from scratch to meet user needs is far less complex than it was two years ago. In fact, today the process is similar to assembling a custom computer.
Of course, you can personally select every hard drive, fan, and cable. But if that level of customization isn't necessary (which is often the case), you can also choose services like Digital Storm or Framework, which offer various pre-configured custom computer solutions for different needs. If you want a middle ground, you can add components yourself based on the merchant's pre-selected parts, all of which have undergone compatibility testing to ensure the final device runs at high performance. This increases modularity and flexibility while eliminating components you don't actually need.
When integrating fundamental components like consensus mechanisms, execution layers, data storage, and liquidity, applications will build solutions that reflect different cultural characteristics. These solutions always cater to differentiated demands (i.e., different definitions of user experience), serve their respective audience groups, and ultimately achieve value retention. The degree of differentiation can be compared to the differences between rugged laptops, business laptops, desktops, and MacBooks, but they also blend and coexist to some extent, as these computers don't run entirely independent operating systems. More importantly, each necessary component becomes a knob that applications can freely adjust, allowing developers to iterate and tweak at will without worrying about making destructive changes to the underlying protocol.
Circle's acquisition of the Malachite team from Informal Systems indicates that controlling customized block space is clearly a broader strategic priority. In the coming year, I look forward to seeing various applications and development teams define and own their on-chain components based on the foundational building blocks and default configurations provided by companies like Commonware and Delta. This is like a HashiCorp or Stripe Atlas for the blockchain and block space领域.
Ultimately, this will enable applications to directly control their cash flow, leverage the unique advantages of the models they build, provide the best user experience in their own way, and thus create a lasting competitive moat.
Prediction Markets Will Continue to Innovate (But Only Some Will Succeed)
—Tommy Hang
One of the most high-profile applications in this cycle is prediction markets. With weekly trading volumes on major platforms soaring to a historic high of $20 billion, prediction markets have clearly taken substantial steps toward mainstream adoption.
This momentum has spurred numerous projects in related areas, aiming either to complement the shortcomings of current market leaders like Polymarket and Kalshi or to challenge their leading positions. But amidst the market hype, only by distinguishing real innovation from market noise can we truly identify the trends worth watching in 2026.
From a market structure perspective, I am particularly interested in solutions that can reduce spreads and increase open interest. Although market creation is still permissioned and selective, liquidity in prediction markets remains relatively thin for both market makers and traders. There are tangible opportunities for improvement through products like lending to optimize routing systems, innovate liquidity models, and enhance collateral efficiency.
Trading volume across different sectors is also a key factor determining the competitiveness of various platforms. For example, over 90% of Kalshi's trading volume in November came from sports prediction markets, highlighting the competitive positioning some platforms naturally have in capturing advantageous liquidity. In contrast, Polymarket's trading volume in crypto-related markets and political markets is 5-10 times that of Kalshi.
However, on-chain prediction markets still have a long way to go before achieving true mass adoption. A highly illustrative example is the 2025 Super Bowl: this single event alone generated $23 billion in daily trading volume in off-chain betting markets, which is more than ten times the current daily trading volume of all on-chain markets combined.
Closing this gap requires sharp and insightful teams to solve the core problems of prediction markets. In the coming year, I will be closely watching the development of these teams.
Autonomous Curators Will Expand the DeFi Market Size
—Eskender Abebe
The curation layer in DeFi is currently at two extremes: purely algorithmic (hard-coded interest rate curves, fixed rebalancing rules) or purely manual (risk committees, active managers). Autonomous curators represent a third model: AI agents (large language models + toolchains + decision loops) curating and managing risk strategies for vaults, lending markets, and structured products. They not only execute fixed rules but also reason about risk, yield, and strategy.
Take the curators in the Morpho market as an example: they need to define collateral policies, loan-to-value ratio caps, and risk parameters to design yield products. This is still a human-dependent bottleneck, and AI agents can achieve scale. At that point, autonomous curators will directly compete with algorithmic models and human managers.
When will we see the "God's move" in DeFi?
When I talk to crypto fund managers about AI, the answers usually fall into two categories: either they believe large language models are about to take over all trading desks, or they think they are hallucinating toys that simply cannot survive in real markets. Both views miss the architectural transformation. Agents, through emotionless execution, systematic strategy adherence, and flexible reasoning, are entering areas where humans introduce interference and pure algorithms are too brittle. They will likely supervise or integrate underlying algorithms rather than simply replace them. Large language models play the role of chief architect designing safety barriers, while deterministic code still resides in the core areas requiring low-latency responses.
When the cost of deep reasoning drops to a few cents, the most profitable crypto vaults will no longer depend on the smartest people, but on who has the strongest computing power.
Short Video Becomes the New Shopping Mall
—Katie Chiou
Short video is rapidly becoming the main channel for people to discover (and ultimately purchase) content they love. TikTok Shop created over $20 billion in Gross Merchandise Volume (GMV) in the first half of 2025, nearly doubling year-over-year, subtly cultivating a global user habit of viewing entertainment content as a new shopping mall.
In response, Instagram has transformed its Reels short video feature from a defensive product into a revenue engine. This format not only brings higher exposure but also contributes a significant share to Meta's advertising revenue expectations for 2025. And the live-streaming e-commerce platform Whatnot has long proven that the conversion rate of a personality-driven live sales model is something traditional e-commerce can't hope to match.
The core logic of this phenomenon is simple: when people watch content in real-time, their decision-making speed increases significantly. Every swipe of the screen constitutes a decision point. Major platforms know this well, so the boundary between the recommendation feed and the shopping checkout process is quickly blurring. The feed is the new shelf, and every creator is a sales channel.
Artificial intelligence is pushing this trend even further. It reduces video production costs, increases content output, and allows creators and brands to test ideas more easily in real-time. More content means more possibilities for conversion, and platforms are optimizing every second of video to maximize users' purchase intent.
Cryptography is born for this transformation. As the pace of content accelerates, faster and more economical payment channels become necessary. When the shopping process becomes seamless and is directly embedded into the content itself, a system is needed that can settle micropayments, programmatically distribute revenue, and track contributions across a complex web of collaborations. Cryptography is designed for this flow of funds. It's hard to imagine achieving a truly scalable, deeply integrated live-streaming e-commerce era without it.
Blockchain Will Drive a New AI Arms Race
—Danny Sursock
Over the past few years, the spotlight in the AI field has always been on the multi-armed race between hyperscalers and startup giants, while DeAI entrepreneurs could only摸索 in the shadows.
However, while the outside world was looking elsewhere, several crypto-native teams have made significant progress in decentralized training and inference, gradually moving from theoretical design phases to testing and production environments.
Today, teams like Ritual, Pluralis, Exo, Odyn, Ambient, Bagel have entered their golden age of development. A new generation of competitors is poised to unleash an explosive multi-dimensional impact on the fundamental development trajectory of artificial intelligence.
Models trained in globally distributed environments can break scalability bottlenecks. These models employ innovative asynchronous communication and parallel processing methods, whose effectiveness is being proven in production-scale operational tests.
The combination of emerging consensus mechanisms and privacy computing components is making verifiable confidential inference a realistic choice in the on-chain developer toolkit.
Revolutionary blockchain architectures, combining smart contracts with flexible computational structures, provide efficient operating environments for autonomous AI agents, using crypto assets as a medium of exchange.
The foundational work is done.
The current challenge is scaling these infrastructure layers to production size and demonstrating why blockchain technology can drive fundamental AI innovation, rather than remaining merely philosophical, ideological, or metaphysical fundraising experiments.
RWA Will See Real Adoption
—Dmitriy Berenzon
Today, RWA tokenization is reaching scale. Although the concept of tokenization has been discussed for years, with the widespread mainstream adoption of stablecoins, the increasing availability of convenient and stable fiat on/off ramps, and global regulatory frameworks gradually clarifying and offering support, this field has finally achieved breakthrough development. According to the latest data from the RWA.xyz platform, the total issuance of tokenized assets across various categories now exceeds $18 billion, compared to just $3.7 billion a year ago. Growth momentum is expected to accelerate further by 2026.
It's important to note that tokenization and the vault model are two different design patterns for real-world asset tokenization: tokenization creates an on-chain representation of off-chain assets, while the vault model builds a bridge between on-chain capital and off-chain yield.
I'm excited to see that tokenization and vault technology allow us to access various physical and financial assets, from commodities like gold and rare earth metals, to private credit for working capital and payment financing, to private and public equity, and more global currencies. Let's think bigger and go further. I want to see eggs, GPUs, energy derivatives, salary advances, Brazilian government bonds, Japanese yen, and more, all on-chain.
To be clear, this is not simply about putting more assets on-chain. The core is upgrading the global capital allocation model through public blockchain technology, transforming opaque, inefficient, and fragmented markets into a new paradigm that is open, transparent, programmable, and highly liquid. Once these assets are successfully on-chain, we will benefit from the composability advantages they offer when combined with existing DeFi.
Finally, these assets will undoubtedly face challenges in terms of transferability, transparency, liquidity, risk management, and distribution, so infrastructure that can mitigate these challenges is equally important and exciting.
A Product Renaissance Driven by Agents Is Coming
—Ash Egan
The influence of the next-generation web will be determined less by the platforms we swipe with our fingers and more by the intelligent agents we converse with.
We all know that the share of bots and agents in all network activity is growing rapidly. Rough estimates suggest it's now around 50%, including both on-chain and off-chain activity. In crypto, bots are increasingly trading, curating, assisting, scanning contracts, and handling everything from token swaps and treasury management to auditing smart contracts and developing games on our behalf.
This is the era of programmable, agentified networks. Although we are already in it, 2026 will mark a turning point where crypto product design (in a positive, open, non-dystopian way) will cater more to the needs of bots than humans.
Although this vision is still emerging, personally, I look forward to spending less time clicking around different websites and more time interacting with a simple, chat-like interface to manage on-chain bots. Imagine an experience like Telegram, but the conversation is with intelligent agents specialized for applications or tasks. These agents can formulate and execute complex strategies, gather the most relevant information and data on the network for me, and feed back trading results, risks and opportunities to watch, and filtered information. I just give the command, and they lock onto opportunities, eliminate all noise, and execute precisely at the optimal moment.
The infrastructure supporting this vision already exists on the blockchain. Combining the default open data graph, programmatic micropayments, on-chain social graphs, and cross-chain liquidity channels, we have everything needed to support a dynamic ecosystem of intelligent agents. The plug-and-play nature of cryptocurrency means less friction, and agents will encounter fewer dead ends during operation. The readiness of blockchain for this, compared to Web2 infrastructure, cannot be overstated.
This might be the most crucial point here. This is not just automation; it's a liberation from Web2's closed ecosystems, from friction, from waiting. We are witnessing this shift happening in search: about 20% of Google searches now generate AI overviews, and data shows that when people see this overview, their willingness to click on traditional search result links drops significantly. The process of manually sifting through pages is becoming unnecessary. A programmable, autonomously executing network ecosystem will extend this transformation further into the applications we use, and I believe this is a good thing.
In this era, we will have less anxiety, less frantic trading. Time zone differences will flatten (no more "waiting for the Asian market to open"). Interaction with the on-chain world will become easier and more expressive for every developer and user.
As more assets, systems, and users move on-chain, this process will create a snowball effect.
More on-chain opportunities → Increased deployment of agents → Increased value release, and so on.
But what we build now and how we build it will determine whether this intelligent network becomes a superficial noise of automation or fosters a renaissance of empowering, dynamic products.
