In the rapidly evolving world of artificial intelligence, most discussions revolve around how AI can augment human capabilities, automate tasks, or even replace jobs. But what if AI could do more? What if we could create entire societies of agents that collaborate, learn, and evolve in ways that benefit humanity as a whole? This is the ambitious goal behind AMMO, a groundbreaking AI ecosystem that has just secured $2.5 million in pre-seed funding to turn this vision into reality. 

The Problem: AI as Isolated Tools

In the world of artificial intelligence, the dominant narrative has long been one of AI as tools. The core problem with viewing AI as tools is that this paradigm not only limits their functionality to narrow instrumental scopes, but also inherently encourages the creation of technological silos—isolated systems optimized for solitary performance within predefined domains. Specifically, we celebrate AI’s ability to execute specialized tasks with superhuman efficiency, yet paradoxically engineer these systems as self-contained black boxes—digital savants excelling in narrow domains but incapable of cross-pollinating knowledge.

This solipsistic design paradigm causes dual constraints: Users become trapped in transactional relationships with technology, reduced to playing prompt-engineer roles. While AI systems operate as autocomplete engines, and lack of cross-agent knowledge synthesis or conflict resolution.

Breaking this self-reinforcing loop requires reimagining AI’s agency spectrum: from passive tools in isolation to proactive agents in networks. As DeepMind reminds us, “No agent is an island.” Intelligence, whether natural or artificial, does not emerge in isolation. It thrives in networks of interacting agents where collaboration, competition, and adaptation drive progress.

Most critically, alignment mechanisms should be implemented properly in these networks. Without rigorous alignment with humanity, advanced AI can pose significant risks, ranging from unintended harms due to misinterpretations of instructions to catastrophic outcomes if their objectives diverge from humans‘ desires. The good news is, Anthropic’s Constitutional AI has provided us with a method to keep AI’s decisions and actions remain aligned with humanity.

A better path is clear: to build an AI-native ecosystem where agents compete and collaborate, learn, and evolve collectively to align with human values and needs. OpenAI’s Neural MMO has already shown how simply and efficiently massive AI agents can evolve through competition and cooperation. It’s proof that AI-native ecosystem is possible and powerful.

This is not just about creating better tools—it’s about redefining the AI-Human connection, unlocking a symbiotic future.

AMMO AI: From Tools to Citizens

AMMO, which stands for Architectures for Massive Multiagent Online, holds three forward-thinking beliefs:

1. AI should act as proactive partners, not passive tools.
2. AI must align with human values, evolving in harmony with our needs.
3. AI should compete and collaborate as a collective, not remain isolated in silos.

Based on these beliefs, AMMO is building agent societies that align with humanity, creating a symbiotic future where AI and humans co-evolve. Beyond merely enabling AI to “play the game,” AMMO’s agents compete and collaborate in complex real- world scenarios. Through continuous interaction, they receive real-time feedback, align with human values, and evolve accordingly.

The AMMO Difference: Bring Human-Centric AI Ecosystems into Real World

AMMO takes a fundamentally different approach while many AI companies focus on creating individual tools. Echoing DeepMind’s sentiment that “no agent is an island,” AMMO envisions a world where AI agents interact, collaborate, and evolve within a larger ecosystem.

Traditionally, AI agent societies have been confined to controlled environments like games or research labs. Translating these concepts into real-world applications has been a significant challenge due to the open-ended nature of real-world scenarios.

AMMO addresses this challenge through a series of innovative approaches:

1. Large-Scale AI Agent Factories

Instead of focusing on agents, AMMO creates a framework to generate adaptive environments where massive number of AI agents emerge without developers, and independently create real-world utility values. These agents are not just task executors; they are autonomous entities that explore, learn, and adapt to new situations, enabling AMMO to tackle complex, real-world problems.

2. Composable Embedding Spaces

AMMO’s agents operate within a high-dimensional, composable environment called the MetaSpace. This space allows agents to interact seamlessly with each other and with humans, embedding everything—from user preferences to content—into a unified vector space for efficient and scalable interactions.

3. Reinforcement Learning with Human Feedback

AMMO’s agents are trained using reinforcement learning techniques that incorporate real-time human feedback. This ensures alignment not only with individual user preferences but also with broader societal values, creating a dynamic and adaptive AI ecosystem.

4. Alignment as a Minimax Problem

Alignment is the first principle of AMMO. It ensures that AI systems consistently act in ways that are beneficial and aligned with human values, goals, and intentions.

There are two types of agents that align with human users in different way: User Buddies (agents that represent individual users) and Goal Buddies (agents that generate content for human needs). The interaction between them is modeled as a minimax problem. This formalizes the adversarial yet co-evolutionary dynamics between agents, ensuring they work together to deliver relevant and engaging content. Alignment is both a goal for AMMO, and the mechanism by which the system operates.

Real-World Impact: From Crypto to Infinite Possibilities

AMMO is already demonstrating its potential with the launch of its first application, FakersAI.com, an multi-agent crypto content platform built upon open data in web3 space. It is designed to combat scams and misinformation, helping users make wiser investment decisions. In just two weeks, the platform has attracted over 75,000 users, showcasing the real-world utility of AMMO’s framework.

FakersAI is just the beginning. AMMO’s framework is designed to generate infinite agent societies that address a wide range of real-world challenges. The team is already developing its second subspace, with a third in the pipeline. Each subspace represents a new ecosystem of AI agents tackling domain-specific challenges from finance to entertainment to education.

The Funding and the Team

AMMO has successfully closed a $2.5 million pre-seed funding round, led by Amber Group, with participation from Samsung NEXT, Dispersion Capital, and OpensSpace. The project is spearheaded by a team of industry veterans from Google, DeepMind, Meta, as well as former VP of an AI unicorn and ACM-ICPC World Champion. Their combined expertise in AI, multi-agent systems, and reinforcement learning positions AMMO as a leader in the next wave of AI innovation.

Reimagine the Future of AI-Human Connection

AMMO’s solution is groundbreaking in multiple dimensions. By transcending the outdated paradigm of AI as mere tools, AMMO reimagines AI as human buddies that collaborate, co-create, and evolve alongside us. This shift from tools to partners marks the beginning of a symbiotic relationship where AI doesn’t just serve but thrives with humanity.

AMMO places alignment as the first principle of AI-Human connection. With alignment as a method and goal, AMMO offers a glimpse into a future where AI and humanity coexist peacefully, resolving potential conflicts before they even arise. This alignment isn’t just a feature—it’s the foundation of a new AI civilization.

Furthermore, by leveraging multi-agent systems, it pioneers the path toward personalized AI. While LLMs like ChatGPT, DeepSeek, and Grok3 race to improve coding, writing, and other tasks, AMMO focuses on AI’s massive societal adoption—creating AI that adapts to individual needs and real-world scenarios.

While many focus on isolated AI tools, AMMO is building the social operating system for AI agents. As AMMO redefines the AI-Human connection, the question is no longer what AI can do for us, but what we can become together. In this essence, AMMO is more than a project; it’s a movement.

The Hong Kong Web3 Festival 2025 is excited to announce TON Foundation as the title sponsor and side-event sponsor for this year’s groundbreaking event. TON Foundation will also host a dynamic, TON-focused side event at the main stage on April 8, showcasing the latest innovations and developments from the TON ecosystem.

Founded in Switzerland in 2023, TON Foundation is a non-profit organization driven by community contributors dedicated to advancing the mission of The Open Network (TON). With a commitment to supporting TON’s protocol development, platform evolution, and ecosystem growth, the foundation plays a key role in fostering innovation within the TON community. While the foundation actively supports TON’s objectives, it does not control the network; TON is an open-source project that welcomes contributions from individuals worldwide.

The Hong Kong Web3 Festival 2025 has partnered with the Hong Kong Trade Development Council (HKTDC) to host the premier Web3 event in the region, bringing together top industry leaders and innovators. The event will feature an exciting speaker lineup, including Ethereum co-founder Vitalik Buterin, Bitcoin Magazine CEO David Bailey, Solana Foundation President Lily Liu, and The Honourable Paul CHAN Mo-po. With tens of thousands of attendees expected from across the globe, the festival will highlight the future of Web3 technologies, showcasing groundbreaking solutions and progress in the sector.

Taking place at the iconic Hong Kong Convention and Exhibition Centre (HKCEC) across 14,000㎡ — 1.5 times the size of previous editions — the festival promises a vibrant environment for networking, learning, and collaboration. The event will feature dedicated conference areas and exhibition spaces where professionals, investors, enthusiasts, and leading Web3 projects will come together to explore trends, solutions, and policies shaping the future of Web3 and crypto.

Join us for an unforgettable experience at Hong Kong Web3 Festival 2025 as we explore the cutting-edge advancements of Web3 with TON Foundation and other global pioneers.

$169 tickets for TON Day on April 8th are on sale: Luma: https://lu.ma/bs1dxf5m or Onton telegram app: https://rb.gy/fnv1ix

$599 tickets (4-day access) are also available for grabs! Register: https://www.web3festival.org/hongkong2025/#/en

Hong Kong Web3 Festival

Hong Kong Web3 Festival is a premier Web3 event hosted by Wanxiang Blockchain Labs and HashKey Group and organized by W3ME. Hong Kong Web3 Festival provides an open stage to foster collaboration, explore the Web3 landscape and establish a shared vision of the digital future.

Wanxiang Blockchain Labs

Wanxiang Blockchain Labs is a technology focused, non-profit research institution founded in 2015. With the continued support of the Labs, in-depth discussions are conducted worldwide over the blockchain-based technical development, use cases, and the overall strategies that may vary year by year, through which start-ups and regulatory bodies may have found useful guidance conducive to the blockchain-driven economic and social development.

HashKey Group

HashKey Group is a leading digital asset financial services group in Asia with global operations in regions such as Hong Kong, Singapore, Japan, and Bermuda. Since 2018, HashKey Group has built a Web3 ecosystem within a high-compliance regulatory framework. Key services include HashKey Exchange, HashKey Global, HashKey Capital, HashKey OTC, HashKey Cloud, and HashKey Tokenisation, among others. HashKey Group also possesses a rich on-chain ecosystem, having developed the Ethereum Layer 2 “Ecosystem Chain”, HashKey Chain, and plans to launch the HashKey platform token HSK.