Written by: Shenchao TechFlow

The trend of AI agents is still ongoing.

On Base and Solana, numerous protocols and memes related to AI agents have emerged, stirring up market funds and attention.

However, the current emerging AI agent protocols are mostly application-layer based, typically carving out their own AI niches within existing public blockchain ecosystems.

However, major infrastructure has always been a narrative with a higher valuation in the crypto world (whether the market buys it is another matter). Would creating a chain specifically for AI agents, allowing more AI agents to run on it, result in a higher narrative ceiling?

Or to put it another way, in a market where VC tokens are not being picked up, could riding the wave of AI agents become a lifeline for some infrastructure projects?

While you are still in doubt, someone has already started working on it.

VCs flood into the battlefield, AI agents identify projects

On November 26, Talus Network, a L1 blockchain designed specifically for AI agents, announced it has raised $6 million in funding led by Polychain, with participation from Foresight Ventures, Animoca, Geek Cartel, Echo, and others.

At the same time, a group of angel investors such as Polygon co-founder Sandeep Nailwal, Sentient core contributor and Symbolic Capital co-founder Kenzi Wang, 0G Labs CEO Michael Heinrich, Allora Labs CEO Nick Emmons, and Nuffle Labs co-founder Atlan Tutar are also involved.

As early as February of this year, when the narrative around AI agents was not so strong, the project had already completed a $3 million first round of financing, led by Polychain Capital, with participation from dao5, Hash3, TRGC, WAGMI Ventures, and Inception Capital.

As a result, Talus's total financing amount has reached $9 million.

Interestingly, the 'L1 designed specifically for AI agents' has caught the eye of another AI agent.

Recently, the rapidly rising AI agent @aixbt_agent on Base has also keenly captured the attention of Talus Network. Aixbt is an AI agent that monitors cryptocurrency Twitter hotspots, capable of analyzing and judging events happening in the industry.

Aixbt believes that Talus can build AI agents that operate entirely on-chain and claims to be monitoring this trend.

This wave of promotion has undoubtedly increased the heat and discussion around Talus, and in an environment where AI memes are everywhere, a serious infrastructure project has instead attracted more attention.

Designing L1 for AI agents, doing system-level optimization

So, how exactly does Talus Network create an L1 designed specifically for AI agents?

Before discussing this matter, there is an even more critical narrative question - why does AI Agent need a dedicated chain?

The existing AI ecosystem faces three major pain points: ambiguous ownership, lack of transparency, and absence of permissionlessness.

Specifically, in current centralized AI systems, the control of resources is concentrated in the hands of a few entities, and users lack a voice over their own data and computing power; AI decision-making processes are often black box operations, lacking audit and verification mechanisms; users also find it difficult to customize and adjust AI services according to their own needs.

Although some platforms in different ecosystems, such as Virutals and vvaifu, allow users to create their own AI agents, more effort is being made in the direction of permissionless access, followed by the tokenization of AI agents, sharing asset earnings through holding tokens.

Questions like who actually owns this AI and whether there is an AI behind it still need some infrastructure to answer.

Thus, a public chain specifically for AI agents can follow the classical blockchain path to solve problems:

  • Ledger - Clear records and transactions of resource ownership

  • Contract - Transparency and verifiability of the decision-making process

  • Cryptography - Permissionless open ecological system

In terms of projects, Talus allows local design and deployment of decentralized on-chain smart agents, seamlessly, trustlessly, and interoperably utilizing on-chain and off-chain resources and services.

It has established a protocol that can represent, utilize, and trade these agents, resources, and services in a permissionless and verifiable manner.

Breaking down the design of Talus, the following four technical components at different levels are worth noting:

  • Infrastructure Layer: A combination of Cosmos SDK and CometBFT

Cosmos SDK is already relatively mature and reliable, but more importantly, its modular features allow the entire blockchain system to be flexibly expanded like building blocks. This flexibility is particularly important when AI technology is rapidly iterating.

  • Contract Layer: The Move language makes design elegant

The native object model of the Move language makes on-chain management of AI resources natural and elegant. For example, an AI model can be directly represented as an object in Move, with clear ownership and lifecycle, which is much simpler than traditional account-based blockchain models. The concurrent processing capabilities of MoveVM support hundreds or thousands of AI agents running simultaneously, which is unimaginable in traditional serial execution environments.

  • Resource Mapping Layer: Mirror Objects System

This system cleverly addresses the issue of how AI resources are represented and traded on-chain. When you need to use a large language model, you cannot possibly put the entire model on-chain.

To explain in simple terms, you can think of Mirror Objects as the 'digital avatars' of these off-chain resources, through which on-chain smart contracts can reliably operate off-chain AI resources.

Specifically, the Model Object is responsible for the on-chain representation of AI models; it not only records the model's metadata but also includes access permissions and usage conditions for the model. The Data Object manages the access control of datasets, ensuring privacy and security of data when used by AI models. The Computation Object tokenizes computing power resources, allowing computing power to be freely traded on-chain like cryptocurrencies.

  • Verification Layer: Multi-layer verification scheme

For ordinary AI agent interactions, such as conversations with chatbots, lightweight digital signatures can be used to ensure the authenticity of responses.

In high-risk scenarios like financial decision-making, zero-knowledge proofs can be enabled to ensure the correctness of the decision-making process without revealing specific details.

For scenarios that require rapid responses but can accept delayed verification, such as AI NPC behavior in games, an optimistic fraud proof mechanism can be adopted to ensure final correctness while maintaining performance.

For more technical details, you can refer to our previous articles: (Interpreting the Talus White Paper: Decentralized AI Agent Center)

Infrastructure in the back, AI dating applications in the front

Currently, Talus itself is still in the testing phase, and it will take time before launching on the mainnet.

From the perspective of project operations and attracting user attention, while holding back major infrastructure breakthroughs, periodically releasing some applications as pilots can show the market the usability of this L1, while also increasing confidence.

At the same time as the financing news was announced, Talus also unveiled its first application in its ecosystem, 'AI Bae'. The 'Bae' in the name comes from the internet slang 'Before Anyone Else', implying the social attributes of this application.

Interestingly, Talus has chosen to position its first application as an AI dating game rather than a more serious financial or business application, making its intention to attract more ordinary users through engaging applications more evident.

From the information currently disclosed, AI Bae will support users in creating and customizing their own AI companions, introducing a Polymarket-style gambling mechanism. This design is quite creative: it not only allows users to interact with the AI but also enables the tokenization of their AI companions, turning them into exclusive memecoins. In other words, your 'digital boyfriend/girlfriend' can not only chat with you but may also become a marketable asset.

This approach of mixing social, gaming, and financial elements is not uncommon in the crypto market. This new generation of public chains may effectively utilize popular gameplay to break through.

Currently, AI Bae has opened whitelist registration. In a crypto market generally pessimistic about new public chains and infrastructure, Talus's unconventional approach may bring some unexpected surprises to the project. After all, in a bull market, sometimes an interesting application is much more effective than merely talking about technical advantages.

Task gameplay, familiar flavor

In addition to the dating applications mentioned above, Talus has also launched a gamified task activity - 'Enchanted Seasons'. The first season activity is named 'The Awakened Orb', running from November 11 to January 11 next year.

This design resembles 'playing a game': Daily Rituals, Weekly Quests, Team Challenges - from the task system perspective, it is indeed a common operating method for Web3 projects. Currently, users can participate in tasks such as linking social media and posting to earn points, which is also a familiar approach seen in previous projects.

However, in the current market environment, user enthusiasm for pure task systems has waned significantly. How to design more differentiated tasks or clarify the economic value that points can bring may become the key to breaking through.

Even an L1 designed specifically for AI agents still relies on traditional community incentive models during its initial operation.

In the crypto world, even the most advanced technologies need to be rooted in user psychology and behavioral habits, manipulating narratives and assets to achieve success.