Recently, I spent a few days in Discord and found that the community vibe of the KITE project is quite different from typical cryptocurrency projects. While others discuss price charts and market movements, people here are talking about how to set up multi-signatures for AI proxies, how to optimize payment protocol delays, and even researching how zero-knowledge proofs can be integrated with proxy authentication. This tech-savvy atmosphere has made me curious about what @GoKiteAI actually wants to achieve.
The data from the Ozone test network is the most straightforward answer. From the beginning of 2025 to mid-December, it processed a total of 1.7 billion proxy interactions, issued 17.8 million proxy passports, and the highest single-day interaction volume reached 1.01 million times. At first glance, these numbers are quite impressive, but the key question is whether these interactions reflect real demand or just volume manipulation. The test network tokens are not worth much, so why are so many people using it?
I reviewed the task design of the testnet and found they were quite clever. It is not a simple faucet token distribution model, but rather a complete design of agent interaction scenarios. Users need to deploy their own AI agents to invoke various services, such as the Professor AI ecological exploration tool, CryptoBuddy price inquiry, and Sherlock fraud detection. Each completed task earns experience points and badges. This gamification design turns the testing process into a learning experience.
More importantly, the tasks behind these correspond to real technical scenarios. When you set an AI agent to query prices from multiple exchanges and then execute arbitrage, you need to understand how session keys are generated, how spending limits are set, and how payment protocols are invoked. This knowledge is not just for testing but is foundational skills for developing applications on the mainnet in the future. From this perspective, the Ozone testnet is essentially a large developer training camp.
I noticed a detail: the monthly active users of the testnet reached 9.7 million in November, but the actual number of active wallets was only 7.8 million. This gap indicates that many users are using multiple agents; it is common for one person to control three to five different-function AI agents. This aligns perfectly with KITE's design philosophy. In the AI agent economy, you don't just have one assistant; you have an entire agent team, with each agent responsible for different fields and capable of collaborating with each other.
The number of over 100 Kite Modules is also worth pondering. These are not developed by the team itself but are contributed by third-party developers, indicating that the ecosystem is beginning to grow organically. I looked at the top-ranked modules, and Agentic Payment accounts for 40% of the interaction volume, which is the core payment function and meets expectations. The Identity & Auth module has issued 17.8 million passports, indicating a strong demand for identity authentication. The staking activity of the Governance module is also not low, meaning a significant portion of users are participating in governance experiments.
There is an interesting game here. Although testnet tokens are not valuable, early participants often receive airdrop rewards when the mainnet launches. Therefore, many are using real money as a time investment to exchange for potential future gains. This mechanism filters out users who are genuinely willing to learn and build, rather than just clicking to claim airdrops.
From a technical perspective, the testnet can stably run at 450 reasoning calls per second and 10,000+ TPS. What level of performance is this for dedicated chains? For horizontal comparison, Solana's theoretical TPS is 65,000, but in actual operation, it often drops to several thousand due to network congestion. Avalanche's subnet can reach 4,500 TPS, but that is under almost no load. KITE's 10,000+ TPS is measured under actual business load, and there has never been a downtime. Stability is key.
A block time of 1 second and near-zero gas fees are essential for AI agent scenarios. Imagine an automated trading bot; it needs to respond to market changes in milliseconds. If each transaction takes 15 seconds to confirm, the arbitrage opportunity will be gone. If each transaction incurs a gas fee of several dollars, high-frequency strategies cannot profit at all. KITE has optimized both parameters to the extreme, showcasing the benefits of vertical optimization.
But performance is just the foundation. What is truly interesting is KITE's innovations in identity and access management. The three-layer key system is not their invention; BIP-32 hierarchical deterministic wallets have long been the industry standard. However, applying this system to AI agent scenarios and making targeted optimizations is their unique contribution.
For a concrete example, suppose you are an e-commerce company with a procurement AI agent responsible for automatic replenishment. You set a daily procurement budget of $10,000, but you don't want it to spend all $10,000 at once, so you set a single transaction limit of $3,000. You also stipulate that purchases can only be made from whitelisted suppliers and cannot exceed 30 days of inventory. These rules on KITE are just a few smart contract function calls. Every time the agent wants to spend money, the contract automatically checks whether all constraints are met.
What is more aggressive is that these constraints can be dynamic. For example, if you integrate an oracle from an inventory management system, when the inventory of a certain product falls below a safety line, the procurement agent's budget automatically increases. When the inventory exceeds the limit, the budget automatically decreases. Such dynamic adjustments require complex workflow engines in traditional enterprise IT systems, but on KITE, it is simply on-chain contract logic.
I initially thought the Proof of AI mechanism was a marketing gimmick, but upon closer examination, I found it indeed solves real problems. A significant issue for AI agents is the black box problem; you don't know why a certain decision was made, and there is no accountability when things go wrong. PoAI requires that every agent operation leaves a proof on the chain, including what data was used, which model was invoked, and what rules were followed to make the decision.
This proof system can not only be audited afterward but, more importantly, can be used to establish a reputation system. An AI agent with 100,000 successful transaction records, zero disputes, and an average response time of 0.5 seconds will have a high on-chain reputation score. When other users choose service providers, they can directly see this reputation score, just like viewing shop ratings on Taobao. The difference is that Taobao ratings can be gamed, but on-chain transaction records cannot be faked.
When I saw the integration of Brevis, my first reaction was: what does zero-knowledge proof have to do with AI agent payments? Later, I realized that in enterprise scenarios, you often need to prove something without exposing details. For instance, your procurement agent needs to prove it has received a certain discount but cannot reveal the specific purchasing price, as it is a trade secret. Through zero-knowledge proof, the agent can generate a proof that convinces the other party that you indeed received the discount, but without showing the specific numbers.
Irys's data storage solution is also crucial. The volume of interaction data generated by AI agents is staggering. If it were all stored on-chain, the cost would be exorbitant. Irys provides a programmable data layer, with important metadata and hashes stored on-chain to ensure immutability. The complete data exists on Irys, significantly reducing costs, but it is also decentralized storage, meaning it won't be lost if a single server fails.
I think Pieverse's cross-chain solution is underestimated. The AI agent economy cannot happen on just one chain. Users are on Ethereum, services are on BSC, and payments are on KITE. This kind of cross-chain collaboration is the norm. Pieverse has achieved seamless cross-chain payments through the x402b protocol and pieUSD, allowing users to be unaware of the underlying chain switches, just like using WeChat Pay without knowing which bank is settling it.
The global roadshow that started in November has also revealed some information. They went to Dubai, Chiang Mai, Seoul, Tokyo, Hong Kong, Singapore, San Francisco, and Denver. This route covers the major crypto and AI communities worldwide, but it is not for marketing presentations. Instead, they are organizing developer meetups and workshops. The event in Chiang Mai was co-hosted with OpenBuild and ETHChiangMai, and the event in Seoul featured the CEO of Perplexity giving a talk.
This activity strategy is very clear; they are not seeking retail investor attention but developer recognition. The ultimate success of a public chain project depends on how many developers build applications on it and how many users utilize those applications. Therefore, they are investing resources in building the developer community. This strategy may not yield immediate returns in the short term, but in the long run, it is the correct path.
In early December, CEO Chi Zhang shared the stage with executives from BlackRock and Solana at the AD Finance Week, discussing the trillion-dollar scale of the autonomous agent economy. This high positioning indicates that KITE is not aiming for the retail market but rather the institutional market. PayPal Ventures' lead investment also confirms this judgment. Traditional payment giants invest in a crypto project not for short-term price fluctuations but for long-term technology and market positioning.
The timing of the MiCAR white paper's release is also noteworthy. The EU's regulatory framework for crypto assets, MiCAR, will officially take effect in the second half of 2025, requiring all crypto projects operating in the EU to meet a series of compliance requirements, including anti-money laundering, user identity verification, transaction reporting, etc. KITE has prepared compliance documents in advance, indicating their intention to apply for an EU operating license.
Such compliance preparations may be a burden for C-end projects, but they are a necessity for B-end projects. Enterprise clients will not use a platform with an unclear regulatory identity, especially in scenarios involving funds. If one day a regulatory ban is issued, all operations would have to cease, which is too risky. KITE's compliance approach, although costly in the short term, opens up a completely different market space.
Recruiting for the Head of Product and Blockchain Infrastructure Engineer positions in mid-December also reveals information. The product manager indicates a shift from technical demos to commercial products, while the infrastructure engineer signifies that mainnet development has entered the sprint phase. The requirements for these positions are quite high; it's not just about hiring anyone, indicating they are preparing talent for the mainnet launch.
Speaking of the mainnet, the timeline of Q1 2026 now seems quite tight. Transitioning from testnet to mainnet is not as simple as code deployment. Security audits, economic model validation, ecological application readiness, and exchange liquidity—all of these require time. I estimate that the actual mainnet might be delayed to Q2, but this is not necessarily a bad thing. For infrastructure projects, stability is more important than speed.
After the mainnet launch, the key will be to watch a few data points. First, whether TVL can accumulate quickly. There were no real funds during the testnet phase. Once the mainnet goes live, how much money users are willing to stake or provide liquidity on KITE’s chain will be the most direct vote of confidence in the project. Second, the quantity and quality of ecological applications. Out of 100 testnet modules, how many can transition to the mainnet and attract new applications? Third, daily active addresses and transaction volume reflect actual usage frequency.
I am personally more concerned about KITE's progress in enterprise applications. AI agent applications for consumers, such as personal assistants and chatbots, have low payment volumes and don't require high infrastructure. The real imagination lies in B2B scenarios, like supply chain automation, financial robots, and customer service systems. The payment frequency and amounts in these scenarios are on a different scale.
If KITE can secure several benchmark enterprise clients, such as a multinational e-commerce procurement system or a financial institution's intelligent clearing system, that would have a different significance. This would not only mean revenue and traffic but also validate the entire technical solution. The fault tolerance of enterprise clients is much lower than that of retail investors; passing their tests indicates that the system is indeed reliable.
From the competitive landscape perspective, the AI agent payment sector is still in its very early stages. Bittensor focuses on decentralized AI training, Fetch is about IoT and multi-agent collaboration, Autonolas emphasizes DAO governance, and KITE focuses on payment infrastructure. These companies are not direct competitors; instead, they occupy different segments of the AI agent value chain. Future collaboration may outweigh competition.
However, true competition may come from traditional payment giants. If Visa or Mastercard decides to implement AI agent payment solutions, their brand, compliance, and channel advantages are unmatched by crypto-native projects. PayPal's investment in KITE is, on one hand, a bet on this direction, and on the other hand, it may also be a hedge against risk. If blockchain truly wins in the future, at least they have established a presence.
What KITE is doing now is essentially betting on a trend: the AI agent economy will explode in the coming years, and blockchain is the best payment and settlement layer for this economy. Whether this bet pays off depends on two premises: first, whether AI can truly develop to a level of large-scale autonomous decision-making, and second, whether the advantages of decentralization can offset the efficiency disadvantages of centralization.
I tend to believe that this trend will happen, but the pace may be slower than optimists think. The technological preparation may only take 1 to 2 years, but legal, regulatory, and organizational adaptation may take 5 to 10 years. This presents both challenges and opportunities for KITE. The challenge is to endure a long waiting period, and the opportunity is that if they can persist, the first-mover advantage will be quite evident. @GoKiteAI's performance during the testnet phase has been decent; 1.7 billion interactions and 17.8 million passports are not fabricated false prosperity. There are real developers testing real scenarios behind this. Over 100 modules indicate that the ecosystem is beginning to grow organically. Compliance preparations and enterprise-level positioning show a long-term strategy. The next 6 to 12 months will be critical nodes for whether the mainnet can be successfully launched and whether the ecosystem can continue to grow. #KITE, how far can this path go? We let data speak.
