It’s pretty simple: chains designed, built, and tweaked for applications are going to melt face. And the best ones of next year will be intentionally assembled from primitives and first principles.
The recent wave of developers, users, institutions, and capital that have come onchain are different from those that came before: they have specific cultures (read: definitions of user experience) that they value over abstract ideals like decentralization and censorship resistance. In practice, sometimes this lines up with the infrastructure that we already have and sometimes it doesn’t.
In terms of the aspects of UX that uniquely matter to crypto-abstracted, lay-oriented apps like Blackbird or Farcaster*, centralized design decisions that would have been heretical even three years ago—like colocated nodes, single sequencers, and bespoke databases—actually make a lot of sense. The same goes for stablecoin chains and trading venues like Hyperliquid and GTE that live and die by milliseconds, ticks, and optimal prices.
But that’s not true for every new application.
For instance, a counterweight to this comfort around centralization is the growing institutional and retail interest in privacy. The needs and desired experience of crypto apps can be drastically different, so their infrastructure should be too.
Fortunately, assembling chains from scratch that cater to these specific definitions of user experience is nowhere near as complex as it was even two years ago. Today, it’s actually not so dissimilar from assembling a custom PC.
Sure, you can pick out every drive, fan, and cable yourself. But if you don’t need that level of granularity (which is probably the case) then you can use a service like Digital Storm or Framework that offers a range of prebuilt custom PCs for different needs. And if you fall somewhere in between, you can add your own parts to the components they’ve already chosen and know will play well together. This gives you increased modularity, flexibility, and the ability to leave out components you don’t actually need, all while ensuring the end product will function at a high level.
In assembling and tweaking primitives like consensus mechanisms, execution layers, data storage, and liquidity, apps create culturally distinct forms that consistently reflect different needs (read: notions of what user experience means), cater to their unique target audiences, and ultimately, retain value. These forms can look as differentiated as ToughBooks vs. ThinkPads vs. desktop tower PCs vs. MacBooks, but they also converge and cohabitate to an extent—it’s not like each of these computers has its own distinct OS. What’s more, each of the necessary components becomes a knob that applications can iterate on and twist as they please without worrying about breaking changes to a parent protocol.
Given Circle’s acquisition of Malachite from Informal Systems, owning the sovereignty of bespoke blockspace is clearly a broader priority right now. In the coming year, I’m excited for applications and teams to define and own their chain resources around primitives and sensible defaults provided by companies like Commonware and Delta, something like a HashiCorp or Stripe Atlas for blockchains and blockspace.
At the end of the day, this will enable applications to directly own their cashflows and leverage what’s unique about the forms they’ve built to deliver the best possible user experience as a durable moat, on their own terms.
One of the most celebrated apps this cycle has been prediction markets. With weekly volumes across all crypto venues hitting record highs of $2B, it is clear the category has taken meaningful steps towards becoming a mainstream consumer product.
This momentum has created a tailwind of sector-adjacent projects aiming to complement or dethrone current market leaders like Polymarket and Kalshi. But amidst the hype, distinguishing genuine innovation from noise is ultimately how we decide what is going to be worth paying attention to in 2026.
From a market-structure standpoint, I am particularly excited by solutions that reduce spreads and deepen open interest. Even though market creation remains permissioned and selective, prediction market liquidity is still relatively thin for both makers and takers. A real opportunity exists for improving optimal routing systems, different liquidity models, and collateral efficiencies through products like lending.
Volume by category is also a major driver of why certain venues win over others. For example, over 90% of Kalshi’s November volume came from sports-based markets, highlighting how some venues are naturally positioned to better compete for advantageous liquidity. Polymarket, by contrast, reeled in 5-10x more volume than Kalshi on both crypto-related and political markets.
Still, onchain prediction markets have a long way to go before reaching true mass adoption. One great reference point is the 2025 Super Bowl; this alone generated $23B in volume in one day across offchain betting markets, which is more than 10x the combined daily volume of all onchain markets today.
Closing this gap will require sharp, inspired teams who can tackle core prediction market issues, and I’ll be keeping an eye out for these players in the coming year.
DeFi’s curation layer sits at two extremes: purely algorithmic (hard-coded rate curves, fixed rebalancing rules) or purely human (risk committees, active managers). Agentic curators represent a third regime: AI agents (LLMs + tools + loops) that manage curation and risk policy in vaults, lending markets, and structured products. Not just executing fixed rules, but reasoning about risk, yield, and strategy.
Think of the curator role in Morpho markets, where someone must define collateral policies, LTV limits, and risk parameters to produce a yield product. Today that’s a human bottleneck. Agents can scale it. Soon, you’ll have agentic curators competing head-to-head with algorithmic models and human managers alike.
When could we see the Move 37 of DeFi?
When I talk to crypto fund managers about AI, I get one of two answers: either LLMs are about to automate every trading desk, or they're hallucinating toys that will never survive contact with real markets. Both views miss the architectural shift. Agents bring emotionless execution, systematic policy adherence, and flexible reasoning into domains where humans are noisy and pure algos are brittle. They will likely supervise and/or compose lower-level algorithms, not replace them. The LLM acts as the architect designing the safety enclosure, while deterministic code remains in the hot latency path.
When the cost of deep reasoning drops to pennies, the most profitable vaults won't be the ones with the smartest humans, but the ones with the most compute.
Shortform video is rapidly becoming the default interface for how people discover (and eventually purchase) content they love. TikTok Shop did $20B+ GMV in the first half of 2025, almost doubling year over year, and is quietly training a global audience to treat entertainment as a storefront.
In response, Instagram has shifted Reels from a defensive feature to a revenue engine. The format drives more impressions and a growing share of Meta’s projected ad revenue for 2025. Whatnot has already proven that live, personality-driven selling converts at a rate traditional ecommerce can’t touch.
The through-line is simple: people make faster decisions when they watch things in real-time. Every swipe becomes a decision point. Platforms know this, which is why the line between recommendation feed and checkout flow is disappearing. The feed is the new point of sale, and every creator is a distribution channel.
AI pushes this shift even further. It lowers the cost of producing video, increases the volume of content, and makes it easier for creators and brands to test ideas in real-time. More content means more surface area for conversion, and platforms respond by optimizing every second of video for purchase intent.
Crypto fits squarely into this shift. Faster content needs faster and more cost-efficient payment rails. As shopping becomes frictionless and embedded directly into content itself, you need a system that can settle micropayments, attribute and split revenue programmatically, and track contribution across a messy chain of influence. Crypto was built for that type of flow, and it’s hard to imagine a hyperscale era of streaming-native commerce without it.
Throughout the last few years, AI’s spotlight has centered on a multi-billion-dollar arms race between hyperscalers and startup giants, with decentralized innovators left to tinker in the shadows.
But with the attention elsewhere, several crypto-native teams have made tremendous strides in the realm of decentralized training and inference, and the frontier of this quiet revolution has slowly shifted from whiteboards to testing and production environments.
Now, teams like Ritual*, Pluralis, Exo*, Nous, Bagel, and others are ready for primetime. This new generation of competitors is positioned to unleash explosive orthogonal impacts on AI’s foundational trajectory.
Scaling constraints can be blown open with models trained across globally-distributed settings, leveraging novel approaches to asynchronous communication and parallelism that are being proven in production-scale runs.
A combination of new consensus mechanisms and privacy primitives are making verifiable and confidential inference very real options in the onchain builder’s toolkit.
And revolutionary blockchain architectures are set to marry (actually) smart contracts with an expressive computational fabric that streamlines autonomous AI agents using crypto as a medium of exchange.
The foundational work has been done.
The challenge is now to scale these infrastructural substrates to production and prove why blockchains can fuel fundamental AI innovation that goes beyond philosophical, ideological, or skeuomorphic fundraising experiments.
We’ve been hearing about tokenization for many years now, but with the mainstream adoption of stablecoins, the emergence of smooth and robust on/off ramps, and greater regulatory clarity and support worldwide, we are finally seeing RWAs adopted at scale. According to RWA.xyz*, as of this writing, there is over $18B in tokenized assets issued across a variety of categories, up from $3.7B just one year ago, and I expect traction to accelerate in 2026.
It is important to note that Tokenization and Vaults are different design patterns for RWAs: while tokenization creates onchain representations of offchain assets, vaults create bridges between onchain capital and offchain yield.
I’m excited to see tokenization and vaults provide access to a wide range of physical and financial assets, from commodities like gold and rare metals to private credit for working capital and payment financing, to private and public equities, and to more global currencies. Let’s get weird with it, too. I want to see eggs, GPUs, energy derivatives, earned-wage access, Brazilian Treasuries, the Japanese Yen, etc., all onchain!
To be clear, this is not just about putting more stuff onchain. It’s about upgrading how the world allocates capital via public blockchains, which make opaque, slow, and siloed markets accessible, programmable, and liquid. But once we get them onchain, we will enjoy the benefits of composability with the DeFi primitives we’ve already built.
Lastly, many of these assets will undoubtedly face challenges around transferability, transparency, liquidity, risk management, and distribution, so infrastructure that alleviates these challenges is equally important and exciting!
The next iteration of the web will be influenced less by the platforms we scroll and more by the agents we talk to.
We all know that bots and agents are responsible for a quickly growing portion of all web activity. Back-of-the-napkin math puts it at about 50% today, including both onchain and offchain activity. In crypto, bots are increasingly transacting, curating, assisting, scanning contracts, and acting on our behalf for everything from trading tokens and managing treasuries to auditing smart contracts and developing games.
This is the era of the programmable, agentic web. And while we’ve already been here for a while, 2026 will be the year that crypto product design begins to cater more to bots than to humans (in a positive, freeing, non-dystopian way).
What this looks like is still coming into view, but I’d personally love to spend less time clicking from site to site and more time interacting with a simple chat-like interface where I manage onchain bots. Think Telegram but conversations are with app/task-specific agents. They’d be capable of forming and executing complex strategies, scouring the web for info and data that’s most relevant to me, and reporting back with transaction results, risks and opportunities to look out for, and curated information. I’ll give them a task and they’ll track down the opportunity, filter through all the noise, and execute when the time is optimal.
The infrastructure for this already exists onchain. Combine open-by-default data graphs and programmatic micropayments with onchain social graphs and crosschain liquidity rails, and we have everything we need to support a dynamic ecosystem of agents. Crypto’s plug-and-play nature means less red tape and dead ends for agents to navigate. It’s hard to overstate just how primed blockchains are for this relative to web2 infra.
And that might just be the most important point here. This isn’t just automation, but liberation from web2 silos. From friction. From waiting. We’re all seeing this shift happen with search right now: roughly 20% of all Google searches now produce an AI Overview, and data suggests that when people see this, they’re much less likely to click on traditional search-result links. Manually sifting through pages is becoming unnecessary. The programmable agentic web will extend this further to the apps we use, and I think that’s a good thing.
This era will see us doing less doomscrolling. Less panic trading. Time zone discrepancies will flatten (no more “wait til Asia wakes up”). Interacting with the onchain world will become easier and more expressive for every developer and user.
And as more assets, systems, and users find their way onchain, this cycle compounds.
More opportunity onchain → more agents deployed → more value unlocked. Repeat.
But what we build now, and how we build it, will decide if this agentic web becomes just a thin layer of noise and automation, or ignites a renaissance of empowering and dynamic products.
*denotes an Archetype portfolio company
—
Disclaimer:
This post is for general information purposes only. It does not constitute investment advice or a recommendation or solicitation to buy or sell any investment and should not be used in the evaluation of the merits of making any investment decision. It should not be relied upon for accounting, legal or tax advice or investment recommendations. You should consult your own advisers as to legal, business, tax, and other related matters concerning any investment or legal matters. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by Archetype. This post reflects the current opinions of the authors and is not made on behalf of Archetype or its affiliates and does not necessarily reflect the opinions of Archetype, its affiliates or individuals associated with Archetype. The opinions reflected herein are subject to change without being updated.
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
Block quote
Ordered list
Unordered list
Bold text
Emphasis
Superscript
Subscript
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
Block quote
Ordered list
Unordered list
Bold text
Emphasis
Superscript
Subscript
