A TEE is a secure area of a device’s processor that isolates code and data during execution, preventing access even from the operating system or other applications. This setup allows us to research private AI operations that may, in some cases, offer stronger isolation than enterprise servers.

We’re building a decentralized peer-to-peer network of small language models that together form a single swarm inference architecture. Through swarm inference, many lightweight models run on independent nodes, contribute results, review each other’s outputs, and refine them until a consensus emerges.

Running these lightweight models on Acurast devices is a natural fit.

By leveraging Acurast’s decentralized architecture of smartphones across 140+ countries, Fortytwo can scale horizontally, becoming smarter with every new node, while maintaining confidentiality through secure execution.

About Acurast

Acurast is a decentralized physical infrastructure network (DePIN) and a global fleet of smartphones to provide distributed, secure edge compute. Acurast redefines decentralized compute with smartphones — battle-tested devices trusted by billions. With Acurast, AI/ML inference workloads can be securely deployed to a global network of attested edge devices, enabling privacy-preserving tasks close to data sources.

About Fortytwo

Fortytwo is a decentralized AI network that introduces a new type of inference using swarms of small language models running on consumer-grade hardware to achieve scale and reasoning capabilities that exceed centralized AI. The network becomes more powerful and efficient with each new node, operating in a permissionless, transparent, and censorship-resistant manner. Fortytwo allows anyone to contribute by running a node, with node ratings recorded on the Monad blockchain for evaluation and security. Developers can access frontier AI reasoning via API or on-chain, while node operators are rewarded for their contributions.

Fortytwo, a decentralized AI network designed to scale intelligence beyond centralized infrastructure, has raised $2.3 million in a pre-seed funding round, led by Big Brain Holdings, with participation from CMT Digital, Escape Velocity Ventures (EV3), Chorus One, Mentat Group, and angel investors including Santiago R Santos, Keone Hon, Paul Taylor, and Comfy Capital, among others.

Fortytwo is creating a new type of inference with small language models running on everyday devices, working together to achieve scale and capability that exceed centralized AI.

By coordinating reasoning across specialized models, Fortytwo is pushing AI beyond centralized infrastructure by replacing monolithic models with networked intelligence—with swarms of AI models working together, verifying, refining, and enhancing each other’s reasoning. As new nodes join, the network scales organically, enabling high-accuracy, low-cost inference for AI applications such as coding copilots, research agents, and other use cases demanding frontier accuracy and reasoning capabilities.

The newly secured funding will accelerate the rollout of Fortytwo’s devnet and support the development of distributed multi-step reasoning, as well as further research.

"Fortytwo is offering a new path for AI scalability, allowing anyone to contribute compute power and participate in a decentralized network, moving away from the centralized, corporate-controlled AI infrastructure of today," said Ivan Nikitin, CEO of Fortytwo.

"As more nodes join, the intelligence of our network grows, enabling a planetary-scale AI reasoning model built on community-owned models. Through swarm inference, these models work together to continuously improve, ultimately surpassing the capabilities of centralized AI in reasoning and efficiency. This funding brings us closer to realizing a future where AI is accessible, scalable, and driven by global collaboration."

Investors backing Fortytwo see this shift as a fundamental breakthrough in AI scalability and decentralization.

​​"We invest in research-driven teams that take a deep technical approach to solving fundamental AI challenges.

The Fortytwo team isn’t just theorizing about AI decentralization, they’re building a network that makes decentralized AI scalability a reality with a feasible path towards cheaper and better performance than that of closed-source models.

As centralized AI reaches its limits, Fortytwo is leading the shift to a decentralized, globally distributed future, and we’re excited to support their work." – Sam Kim, Partner at Big Brain Holdings.

"As one of the first investments from our fourth fund, we believe Fortytwo is redefining AI scalability by proving that AI doesn’t need to be bigger—it needs to be smarter,” said Sam Hallene, Investment Partner at CMT Digital. “Their swarm inference model enables AI to collaborate, verify, and refine outputs together, making it more accurate, scalable, and transparent.”

Fortytwo has launched its devnet on the Monad testnet. The company's ongoing early adopter program, ‘Hitchhiking to AGI,’ aims to establish a foundation for decentralized AI coordination and collaboration.

About Fortytwo

Fortytwo is a decentralized AI network that introduces a new type of inference using swarms of small language models running on consumer-grade hardware to achieve scale and reasoning capabilities that exceed centralized AI. The network becomes more powerful and efficient with each new node, operating in a permissionless, transparent, and censorship-resistant manner. Fortytwo allows anyone to contribute by running a node, with node ratings recorded on the Monad blockchain for evaluation and security. Developers can access frontier AI reasoning via API or on-chain, while node operators are rewarded for their contributions.

What is the Fortytwo Devnet?

Fortytwo is a decentralized AI network where small, specialized models run on everyday devices, collaborating to achieve scalable and increasingly sophisticated reasoning.

The devnet runs on Monad as its blockchain settlement layer. Unlike centralized AI, which depends on massive infrastructure, Fortytwo operates as a distributed network where participants can contribute any model, including fine-tunes, without centralized validation or whitelisting. On-chain node performance tracking is essential for maintaining network integrity – rewarding high-performing nodes while preventing exploitation, sybil attacks, and dishonest participation.

‘Hitchhiking to AGI’ – How to Participate

‘Hitchhiking to AGI’ is Fortytwo’s devnet launch program, designed to onboard and reward users who help build and test the network. There are two ways to participate:

A. Run a Node

1. Apply as a node operator here.

2. If selected, you’ll receive an initial token mini-drop to the Monad address you provide.

3. You will have 72 hours to launch your node using the provided instructions.

4. Active nodes will receive ongoing rewards.

B. Grow the Network

1. Share updates about Fortytwo on X.

2. Create dashboards to track node activity and devnet progress.

3. Share your contributions in our Discord to qualify for rewards.

Why Join the Devnet?

• Shape the future of AI and become part of its foundational community.

• Contribute to a decentralized planetary-scale intelligence that scales without relying on centralized control.

• Earn rewards for strengthening the network as a node operator or contributor.

Fortytwo goes beyond standard LLM inference and base AI use cases, building a decentralized foundation for the next generation of AI reasoning. Instead of relying on expensive model training and centralized infrastructure, the network evolves with every new participant and model. With Monad’s high-performance settlement layer, Fortytwo enables open participation, allowing any model—whether general or fine-tuned—to contribute to a new form of AI inference. By enabling multi-step collaboration and deliberation across nodes, this system has the potential to unlock new levels of intellectual work, advancing AI capabilities beyond what centralized approaches can reasonably achieve.

Join us in 'Hitchhiking to AGI' – because the future of AI isn’t something built in a lab but something we create together. Every node, every model, every participant strengthens the network, pushing us closer to a truly decentralized, ever-evolving intelligence. This isn’t just about running a node; it’s about shaping the foundation of a planetary reasoning engine – one that grows with the people who power it. Your contribution isn’t just valuable; it’s essential.

About Fortytwo

Fortytwo is a decentralized AI network that introduces a new type of inference using swarms of small language models running on consumer-grade hardware to achieve scale and reasoning capabilities that exceed centralized AI. The network becomes more powerful and efficient with each new node, operating in a permissionless, transparent, and censorship-resistant manner. Fortytwo allows anyone to contribute by running a node, with node ratings recorded on the Monad blockchain for evaluation and security. Developers can access frontier AI reasoning via API or on-chain, while node operators are rewarded for their contributions.

At the frontier of artificial intelligence, this lesson has grown increasingly relevant. While tech giants race to build ever-larger language models—consuming vast amounts of energy and resources in centralized data centers—Fortytwo is following nature's lead, cultivating an adaptive, dynamic network that grows stronger and more intelligent with every new participant.

The Future of Sustainable AI

Fortytwo’s swarm inference orchestrates the collective power of networked Small Language Models (SLMs). Each node in the network is like a specialized expert, trained to excel at specific tasks while remaining lightweight and efficient. These nodes don't need massive data centers or extraordinary computing power – they can run on the kind of hardware many of us already have at home.

As traditional approaches to scaling AI are hitting technological and economic limits, Fortytwo’s swarm intelligence charts a new course through distributed processing across the network. This architecture isn’t just more efficient—it transforms the way scalability is achieved, offering energy efficiency, sustainable growth, and greater resilience by eliminating single points of failure.

How Swarm AI Works: A Practical Example

Let's say we want to ask, "What were the key causes of the fall of the Roman Empire?"

Instead of routing that query through a single massive AI model, Fortytwo's network springs into action like a highly coordinated team of historians. Some focus on economic factors, others on military strategy, and still others on social dynamics. Yet these experts don't just provide independent answers—they actively validate each other's contributions through peer ranking.

When a query enters the system, it triggers a cascade of parallel operations, with nodes simultaneously analyzing different aspects of the question. Rather than having every node evaluate every response—imagine asking each historian to review everyone else's work—each node reviews a strategic subset of responses. This process is highly efficient: nodes can assess quality by generating just a single evaluation score, making the ranking process significantly faster than generating a full response. Think of it like a judge holding up a score card rather than writing a detailed review.

By distributing evaluations across subsets, the network remains tamper-proof, preventing manipulation attempts, such as biased or coordinated judging.

To further enhance integrity, the system maintains an incremental reputation system on blockchain that tracks how well each evaluating node performs over time. By recording a node's scoring deviation from the collective consensus on-chain, the network automatically adjusts the weight given to their future evaluations. This dual layer of validation—fast individual scoring combined with long-term performance tracking—ensures both speed and reliability in the network's decision-making process.

While some nodes are evaluating responses, others continue their work asynchronously, allowing for overlapping operations that reduce latency. The network’s weighted ranking aggregation ensures that nodes with proven expertise have greater influence in the final selection, while still allowing newer perspectives to contribute.

Nodes act as both contributors and validators, creating a self-regulating system that ensures accuracy and reliability. The result is a dynamic web of trust where the best answers rise to the top through consensus—ensuring a more balanced approach to information, free from centralized control or bias.

The Network Effect: Every Node Counts

Unlike monolithic large language models that demand ever-growing computational power for diminishing performance gains, Fortytwo’s network gets smarter through diversification. Each new node enriches the collective intelligence of the entire system.

Additional nodes bring:

• Unique specializations that expand the network's capabilities.

• Additional reasoning and verification power to ensure accuracy.

• Increased capacity for handling more responses.

• Greater geographical distribution for increased reliability.

The future of AI isn’t centralized—it’s distributed. With just a PC or Mac, anyone can participate in building the first planetary-scale AI.

The devnet is live. Join as an early contributor and help shape what’s next.

But, as decentralized finance and services gain momentum, artificial intelligence—the most transformative technology of our time—remains firmly in the hands of a few corporations.

This centralization is now facing an unavoidable constraint: compute scarcity. The expansion of AI is requiring ever-larger data centers, which is hitting economic and physical limits.

Fortytwo presents an alternative—one that decentralizes intelligence itself.

By uniting the idle computing power of home devices, Fortytwo forms a global network of AI models, an emergent planetary-scale intelligence.

This method, called Swarm Inference, makes AI scalable, cost-efficient, and permissionless, evolving and improving with every new participant.

The Problem with Centralized AI

Right now, a handful of massive corporations control AI development—Big Tech.

With abundant resources, particularly capital, billions of dollars are being funneled into development of ever-larger, monolithic language models. These investments extend beyond model training to include the construction of new data centers, energy infrastructure, and semiconductor manufacturing.

We don’t see this as a sustainable path—especially as traditional model training approaches appear to have reached their limits. Further training of large language models (LLMs) now delivers diminishing returns, where even marginal performance improvements demand exponentially greater compute power and energy consumption.

At the same time, a new AI paradigm is emerging—reasoning models. Unlike conventional LLMs, these models prioritize more intelligent responses by allocating additional processing time per query. However, this shift introduces a tradeoff: fewer requests can be handled for the same computational resources. Such meaningful improvements require giving models more time to think, adding further to compute scarcity.

And compute is already scarce. OpenAI caps API calls at 10,000 per minute, effectively limiting AI applications to serving only around 3,000 concurrent users. Even ambitious projects like Stargate, a $500 billion AI infrastructure initiative announced recently by President Trump, may provide only temporary relief. According to Jevons’ Paradox, which observes that efficiency improvements often lead to increased resource consumption due to rising demand, as AI models become more capable and efficient, compute demands will likely surge through new use cases and broader adoption.

The result? Big Tech will continue to own and control artificial intelligence, deemed to be humanity's most powerful invention—and shape its evolution toward Artificial General Intelligence (AGI) and beyond.

But we believe in a different future—one where AI is a universally accessible resource, owned, operated, and maintained by a global community of AI node operators.

Fortytwo’s Decentralized Answer

Instead of building ever-larger models, Fortytwo scales AI horizontally by networking small language models (SLMs) on consumer hardware. Recent advancements in SLMs have shown that they can already outperform state-of-the-art LLMs in specialized domains—coding, mathematics, creativity, vision, and many more.

At a time when centralized compute is scarce, idle compute on everyday devices is abundant. A MacBook Air with an M2 chip, for instance, only uses about a quarter of its power for routine tasks like video calls, spreadsheets, and web browsing. Collectively, consumer devices—Macs with Apple Silicon, PCs with dedicated GPUs, and others—hold an estimated 2,000 exaflops of underutilized compute, just sitting idle.

Fortytwo transforms this idle compute into a coordinated swarm. Nodes running small AI models collaborate, amplifying each other’s capabilities through peer evaluation and joint response preparation based on the network’s most valuable contributions. With Swarm Inference, Fortytwo is not constrained by the capabilities of even the most advanced models—each participant strengthens the collective intelligence of the entire system.

Joining the network is simple. No special skills or hardware are required. A node application runs seamlessly in the background on a standard MacBook Air, using only idle compute that is immediately available. The network is fully permissionless, allowing technical users to operate custom models—including personal fine-tunes—without whitelisting, centralized approval, or verification mechanics. Privacy remains intact, as all node customizations stay private and under the operator’s control.

The swarm thrives on diversity—the more variation within the network, the stronger it becomes.

Distributing the Future

William Gibson famously said: "The future is already here — it’s just not very evenly distributed."

Fortytwo is changing that. Instead of concentrating AI power in a few data centers, it distributes intelligence across a decentralized network, making AI accessible to everyone and enabling broader participation in the AI economy.

We invite those who share our vision to contribute—whether by running a node, spreading the word, or building with us. Web3 and AI developers are especially welcome to explore opportunities to join our team and help accelerate the advent of decentralized intelligence.

The network is launching soon. Run a node and join the AI swarm.