HolmesAI recently completed a new round of financing of $5 million, with investors including Bitrise Capital, CatcherVC, and Cryptomeria Capital. Since the financing announcement, the platform's user base has grown rapidly. Currently, the platform has more than 2.2 million registered users, 1.24 million agents created, and nearly 2.17 million persons. HolmesAI has also disclosed plans to advance TGE in Q1 of 2026.
Article author: HolmesAI
Article source: ME News
ME Group recently hosted a successful AMA event centered around HolmesAI, themed "Decoding 2026: The Explosion of AI Agents." The AMA featured HolmesAI CTO Ky and Dill's Co-Founder Ted as guests, who engaged in in-depth discussions on the industry evolution of AI agents, why capital is accelerating its investment in the field, and the long-term value of digital avatars within the Web3 ecosystem. This article aims to systematically analyze the key trends in the AI agent sector as it moves towards 2026 from three dimensions: capital logic, technological moats, and future predictions.
I. Capital Logic: Why did Bitrise/CatcherVC/Cryptomeria choose to heavily invest in AI Agents by the end of 2025?
Despite a generally subdued sentiment in the crypto market in 2025, Ky observed a series of significant structural changes within Web3. Prediction market trading volume continued to rise, a new batch of AI agents with practical functionality were being launched, stablecoin payment users grew rapidly, and even ordinary users began trading assets like stocks on-chain. A staggering 54 million on-chain/interaction records demonstrated the real network effect. These changes all point to one trend: Web3's use cases are gradually expanding from price-driven financial transactions to applications that are closer to real-world needs.
It is against this backdrop that institutions such as Bitrise, CatcherVC, and Cryptomeria have chosen to increase their investment in the AI Agent sector at this stage. Ky believes the core reason for this decision is that the AI Agent market is still in a very early stage, both within the Web2 and Web3 ecosystems. The entire industry is still digesting AI capabilities and searching for the most certain vertical applications. Precisely because it is not yet fully formed, the AI Agent and data market have the potential to unleash huge growth potential in the next one to two years, providing a window for long-term capital to position themselves in advance.
From the perspective of application evolution, Ky has provided a clear path. As stablecoin payments lower the barrier to entry for users, holding crypto assets will gradually become the norm for ordinary users, and Web3 use cases will no longer be limited to financial speculation. When AI agents begin to participate in complex decision-making scenarios such as prediction markets, their value will no longer be merely that of an auxiliary tool, but rather that of an entity directly influencing the outcome. In this process, the attributes of data will change: in the past, users purchased analytical conclusions about whether a certain token or stock was worth investing in, while in the future, what will be traded will be the "information itself," which can be directly used for judgment and betting.
Ky believes that the supply of data will also become highly diversified. Whether it's professional background, industry experience, or judgment on specific events, all can become sources of data that can be sold. He mentioned that models such as "stake media," which a16z focuses on, are essentially a manifestation of data monetization. As the number of users, types of data, and application scenarios expand simultaneously, a positive feedback loop will form between AI agents and data transactions.
Ted's addition further reinforced this capital logic from an infrastructure perspective. He pointed out that existing infrastructure still faces significant bottlenecks in terms of cost and efficiency to support a large-scale on-chain agent economy like HolmesAI. AI agents are highly sensitive to transaction costs and throughput when executing micropayments, microtasks, and frequently submitting on-chain transactions. While not all actions need to be on-chain, the infrastructure's ability to achieve both low cost and high performance will directly determine the continued expansion of the agent economy once more critical actions are brought on-chain.
According to Ted, the infrastructure layer and the application layer are not mutually exclusive, but rather interdependent: infrastructure provides the conditions for the large-scale operation of agents by reducing costs and improving performance; while once the application layer creates real value for users, it will amplify on-chain activities and network effects. In this sense, the deployment of AI agents will not only drive the explosion of the application layer, but will also force the continuous upgrading of the underlying infrastructure.
Based on the assessments of Ky and Ted, it's clear that the heavy investment in AI Agents by the end of 2025 wasn't a bet on a single application or technology, but rather a prediction of a complete evolutionary path: AI Agent applications will be implemented first, data will be assetized and form a trading market, infrastructure will be passively upgraded during scaling, and ultimately, a new on-chain economic structure will be formed. In reality, leading institutions still possess ample capital reserves, enabling them to complete pre-emptive positioning during periods of market downturn, thus securing a foothold for the potential inflection point in the AI Agent and data market in 2026.
II. Technological Moat: Why must HolmesAI be built on Dill?
In HolmesAI's overall architecture, the underlying infrastructure is not a generic component that can be easily replaced, but rather a key condition that directly determines whether its Agent economy can stand. Based on Ky's assessment, HolmesAI is not facing the transactional scenarios that traditional public chains excel at, but rather a complex system centered on AI Agents, Personas, and continuously generated data, which places entirely different demands on performance, cost, and data availability.
Before selecting its underlying network, HolmesAI systematically compared several mainstream L1 networks across multiple dimensions, including performance, storage cost, security, decentralization, and ecosystem consensus. The conclusion is clear: at this stage, Dill is one of the few infrastructures capable of simultaneously meeting the demands of high performance, low storage cost, and the ability to support complex applications. Furthermore, it is compatible with the EVM technology stack and can support the complete application logic required by HolmesAI. More importantly, Dill allows the activities and data of AI Agents to be truly recorded on the blockchain, a prerequisite for HolmesAI to build its AI Agent economy.
This choice stems primarily from HolmesAI's different understanding of "on-chain data." Traditional public blockchains are designed primarily around transaction data, while HolmesAI aims to store not only transaction results on-chain, but also long-term data such as agent operation logs, inference processes, and agent and user personas. This data is not only used for backtracking and auditing, but also directly relates to the quality of AI model output: only when the behavior of the AI agent can be continuously monitored and data quality can be verified can the model continuously optimize and generate positive feedback. Therefore, for HolmesAI, the predictability and transparency of data are not additional features, but the foundation of the system's operation.
From Dill's architectural design, its support for HolmesAI is concentrated in two key dimensions: scalability and data availability (DA). Regarding scalability, Dill does not adopt a single, shared global execution environment. Instead, it allows each application to have an independent execution space above the data and consensus layers. This means that as the number and interaction frequency of AI Agents on HolmesAI increase, its operational scale does not need to compete for resources with other applications across the network. Instead, it can expand its execution space on demand, avoiding performance bottlenecks at the architectural level. This application-centric scaling model provides near-infinite scalability for the long-term growth of the AI Agent economy.
At the data level, Dill's design is also optimized for AI Agent scenarios. HolmesAI needs to store large amounts of verifiable data long-term, including AI Agent memory data, inference paths, and historical information accessible to other AI Agents and users. This requires that the data not only be cheap to store but also always available, reliable, and without impacting overall network performance. Dill introduces a sharding mechanism at the consensus and data layers, splitting and storing data across different sub-networks. This significantly reduces long-term storage costs while ensuring 24/7 accessibility and preventing overall performance from being dragged down by the expansion of data volume.
It is precisely under these infrastructure conditions that HolmesAI has the technological capability to smoothly scale from thousands of AI Agents to millions of AI Agents without being forced to scale down due to performance or cost issues. This is particularly crucial for the AI Agent economy: once AI Agents begin executing a large number of micropayments and microtasks and frequently submitting on-chain states, the cost structure and throughput capacity of the underlying network will directly determine the sustainability of the application.
From a longer-term perspective, the data availability and scalability provided by Dill not only solve the problem of "whether it can run," but also determine whether HolmesAI's agent system has room for continuous expansion.
When the behavior, historical performance, and reasoning process of an AI agent can be stably recorded, repeatedly invoked, and reused across scenarios, the agent's capabilities no longer rely on single interactions or short-term incentives. Instead, they can continuously accumulate network value as the ecosystem expands. This lays the foundation for HolmesAI to build a sustainably evolving agent economy. In this system, users do not receive immediate rewards, but rather a more stable, verifiable, and continuously optimizeable agent service experience.
III. Future Predictions: Where will the AI Agent sector go in 2026?
Ky provided a relatively clear quantitative forecast for 2026. He believes 2026 will present huge opportunities for AI agents and the data market, especially in prediction market scenarios. "Prediction market users could see a 10 to 20-fold increase, and AI agents could see a 1000-fold increase." In his view, when AI capabilities become ubiquitous, the real trading value will no longer be in the analytical reports themselves, but rather "information that can help us place bets in the prediction market." He emphasized that data will become highly diversified, which is the foundation for the simultaneous expansion of both AI agents and the data market in 2026.
End
As the AI Agent field gradually moves from proof-of-concept to large-scale deployment, and with HolmesAI's user base continuing to expand rapidly, reaching a cumulative user count of 2.25 million, its key actions in the coming period may become an important window into observing its ecosystem evolution. According to the team, the project will soon launch a series of community engagement initiatives, including a stablecoin pre-deposit mechanism and the sale of NFT tokens, and will continue to release progress information through the official X platform. These arrangements are seen as important entry points for participation before TGE and will directly impact users' roles and rights in the subsequent ecosystem.
About HolmesAI
HolmesAI is a Web3 project centered on AI agents and digital avatars, aiming to build an on-chain intelligent personality system that users can fully own, control, and participate in value distribution. The project recently completed a new round of financing of $5 million, with investors including institutional investors such as Bitrise Capital, CatcherVC, and Cryptomeria Capital, reflecting the consensus among leading investors on the long-term potential of the AI agent sector.
Following its funding announcement, HolmesAI has experienced rapid user growth, with over 2 million registered users and over 2 million Personas on its platform, demonstrating that its product has generated real demand and initial network effects. According to the team, HolmesAI plans to advance to TGE (Transformation and Geography) in Q1 2026, and is currently in the Pre-TGE stage, considered by the market as one of the clearer early-stage opportunities in the AI Agent sector.
