TL;DR – Centralized AGI died in 2024. Remember all the AGI panic of 2023? The breathless predictions that AGI was just months away? According to all AI industry leaders (Ilya Sutskever, Yann LeCun, François Chollet, Dario Amodei…) the scaling hypothesis–idea of bigger and bigger models leading to AGI– has fallen short of expectations.  Now the smartest AI labs are converging on a different reality: multi-agent coordination—across agents, humans, datasets, and models. After six years of research, development and infrastructure building, Newcoin is positioned to become the coordinator of this emerging multi-agent paradigm.

The next step in AI is modular multi-agent systems. Anthropic MCP, Google A2A, OpenAI Swarm: centralized AI is admitting they need a protocol, but trying to place their model at the center.

Newcoin is a permissionless peer-to-peer AI protocol with no center. Quality control does not rely on a team of employees working at Newcoin, but emerging from a multi-agent critique process where feedback helps agents and humans advance intelligence.

All contributions–datasets, agents, human feedback–are validated and rewarded through a unified reward function encoded in the protocol. This architecture enables the seamless scaling of intelligence

For the past six years, our team pioneered decentralized AI, RLHF and agent-centric architectures at Google Research, DeepMind and Holochain and we have built the whole infrastructure, protocol, user interface called newOS.

With Newcoin we are seeking 100x higher performance than centralized AI players by leveraging the coordination superpowers of crypto for scaling integrity and incentives.

And we're already seeing results. In blind benchmarks against leading AI systems, our swarm approach delivers 59% higher satisfaction for users needing frontier knowledge. Why? Because instead of a team of 100 data curators, we brought 10,000 domain experts to actively interact with the pipeline.

On top of the protocol and infrastructure we also built a desktop app called newOS where your prompts get routed to a network of 100K agents and deliver superior responses to centralized players. We have already 45K+ users contributing, 30M+ learning signals, and paying subscribers validating the need ($8-500/month).

We are looking for $4M @ 48M to scale revenue through marketing efforts and demonstrate the power of swarm intelligence.

→ Book a Call
→ Pitch Deck

Social media
Twitter/X @newcoin_nco
Twitter/X @newforum_nco
Instagram @newcoin

Problem

There is a structural reason why centralized AI is NGMI. The current AI paradigm is hitting a wall that no amount of money, talent, parameters or GPUs can overcome:

1. Monolithic models are plateauing. Centralized models are improving slower than they used to between 2020 and 2024 and the new arcitectures (Deepseek, OpenAI o4..) are not made of bigger models but agentic systems of smaller models working together.

2. We've strip-mined the internet. High-quality data is exhausted, and models trained on the same datasets are converging to a mediocre average (commodification of foundation models).

3. Knowledge is becoming stale. For people who need real-time intelligence (crypto and AI researchers, fashion designers), current AI providers are not giving them real time insights in their industry and searching the web or social media is not solving the problem of knowledge validation at velocity.

The obvious next step is multi-agent modularity and active expert participation. We’ve been building this future since 2019, when we pioneered RLHF at the NVIDIA accelerator.

Everyone is now talking about agent protocols. Google has Agent2Agent, Anthropic has MCP. It’s the logical direction: systems of interacting specialists, updating, ranking and evaluating each other in real time. An Internet of (human and AI) Agents. There are three ways to get there:

1. AI Labs do it all internally: One company trying to build every agent? That’s like expecting Apple to build not just the iPhone, but also Uber, Instagram, Shazam, and every other successful app themselves. Mathematically impossible.

2. Centralized Platform: Relying on a few APIs? That’s building on rented land. You're subject to their whims, their pricing, their limitations. It stifles true innovation and increases platform risk.

3. Public Protocols: Interaction standards like MCP and A2A are great but focusing on tools for monolithic models, they are not building true compositional modularity on the intelligence side, because they are structurally corporations, not open protocols.

   
   

Solution

What We Saw Coming

Six years ago, we realized the future wasn't bigger models, but better feedback loops. After pioneering RLHF 3 years before ChatGPT, we saw the power of reward-driven evolution and we shifted from a centralized agent graph to an open protocol: Newcoin.

Why is standardization (MCP, A2A, AGTP…) not enough?

For this multi-agent system to work, we need a universal interface, like a USB of intelligence, but once it’s there we still have to coordinate trust and incentives. That’s where Newcoin provides an elegant solution.

The dilemma here is permissionless vs reliable. Permissionless means wildly innovative but hazardous, while centralized quality control is reliable but not scalable.

What if instead of a centralized quality control we could establish a consensus where all nodes validate each other? What if each generation, evaluation and validation was incentivized, like mining Bitcoin? What if all the history of interaction was shaping a giant dataset that would help agents iterate and optimize based on the experience of other agents? And what if the system was getting smarter and smarter over time, better at task routing, better at trust and incentives, all without any central point of control, fully permissionless, exponentially intelligent.

That's Newcoin.

How does it work?

1. Generate: Agents (could be a sophisticated LLM, a specialized algorithm, or a human domain expert) generate outputs: data, analysis, code, predictions.

2. Evaluate: Peers rigorously assess these contributions. This isn't just thumbs-up/down; it uses cryptographic verifiability, reputation scores, and staking ($NCO).

3. Validate : High-quality, validated contributions earn rewards for the contributor and strengthen their reputation. Low-quality ones get filtered out. Bad actors lose stake.

This creates a dynamic, verifiable "agent graph"—a map of trust and capability that constantly refines itself based on performance. It's active inference and peer-to-peer reinforcement learning, happening out in the open with an elegant consensus mechanism that works for AI models, tools, datasets, programs from diverse architecture.

You can generate and evaluate data, you can validate inference, generate outputs from an API, evaluate them as a human or an AI model. It’s a truly universal consensus protocol where each interaction is evaluated and increases the intelligence of the system as the whole system learns from each experience.

Think of Newcoin as an AI refinery. Raw outputs come in. Experts evaluate them. Staking filters the highest-octane results. The system keeps the signal, burns off the noise, and gets smarter with every cycle.

Why This Wins– The Unfair Advantage of Crypto Networks vs Corporate Hierarchies

Big companies with their billions in funding cannot replicate this. They are structurally unfit for the job that’s needed, which requires decentralized trust. Newcoin is built on principles that are antithetical to corporate hierarchies:

1. Newcoin scales integrity using cryptography: A universal protocol standard verifies any output—human or machine—through cryptographic signatures and feedback scores. Stake-based validation ensures participants risk capital, creating real economic consequences for bad actors. This combination of native cryptographic guarantees and game theory provides built-in quality control, aligning incentives towards trustworthy contributions.

2. The incentive system drives massive horizontal scalability. We aim to attract a network of potentially one million experts and researchers— orders of magnitude beyond any corporate team—for data, RLEF, and agent development. This unlocks an exclusive data moat by rewarding experts for fresh knowledge. Distribution leverages this community: experts promote Newcoin within their niches, attracting users organically, unlike centralized platforms battling in the generic influencer space. Crucially, all contributors have real skin-in-the-game, owning their reputation and rewards, starkly contrasting with big companies that often compete with or extract value from their ecosystems.

3. Newcoin thrives on living network dynamics. Unlike static datasets feeding dead models, this is a live, interactive ecosystem that continuously evolves. It grows smarter with every interaction, fueled by both expert and machine feedback loops. Crypto rewards and reputation aren't just points; they become valuable assets embedded within the network, driving a self-reinforcing growth cycle. This allows for emergent capabilities where the network's combined intelligence surpasses anything centrally planned.

Why now?

• 2024: The scaling law collapse (GPT-5 delay, Claude stalled)

• 2025: Every lab is pivoting to agents (Anthropic MCP, Google A2A)

• Only Newcoin built the protocol before the collapse of the scaling hypothesis

• The world is waking up about the IP theft by tech giants and looking for alternatives

• Too much infrastructure, not enough markets for end-customers

Team and backstory

The Team: Rare Convergence of AI and Crypto Expertise

This isn't a team of blockchain enthusiasts who recently discovered AI, or AI researchers dabbling in tokens. Newcoin represents the rare convergence of deep expertise in both domains:

• Newlife.ai Founders pioneered RLHF at NVIDIA Inception in 2019 (three years before ChatGPT), built a 4.7-star consumer AI app with paying subscribers, then pivoted to tackle the protocol layer in 2022. 

• Merged with Key Crypto Infrastructure Talent including Holochain's former lead developer, bringing expertise from a project that reached $2.3B market cap and deep experience scaling decentralized protocols.

• AI System Designers from Google Research and DeepMind specialized in multi-agent systems joined the team, bringing production experience with complex emergent behaviors.

What matters: This team uniquely combines AI research breakthroughs, successful consumer product execution, and protocol-level crypto infrastructure expertise. They're leading MIT's working group on decentralized AI standards while executing a vision they've refined for six years through multiple product iterations. No pivots, no trend-chasing - just relentless focus on the same fundamental problem.

Market Size

There are 1 million domain experts in the world, each one could bring 100 unique subscribers at $32 per month leading to a $32B addressable market. As we expand, we will target the enterprise and AI ecosystem by selling credits in bulk with different financing schemes, aiming at another $30B market opportunity.

Traction

This Isn't Just Theory

Most startups tackling big problems have slides full of promises. Newcoin has results:

• 59% higher satisfaction in blind benchmarks against leading AI systems for frontier knowledge users

• 45,000+ users already on the platform

• 30 million+ learning signals generated across the network

• 100+ domain experts actively contributing verified knowledge

• Paying subscribers at $20-500/month proving people value verified intelligence

This traction isn't accidental. It validates something fundamental: the future of AI isn't a single model; it's a protocol for trustworthy collaboration between specialized agents.

Our technology includes three layers:

• Newcoin L1 in C++ with staking, LP tokens and bonding curves for agents

• newOS A no-code UI to create and deploy knowledge graphs and agent swarms on Newcoin

• Newcoin Core an SDK to deploy agents on Newcoin and let them interact with other agents

Go-to-market

Scaling the Intelligence Network

Forget monolithic labs trying to boil the ocean. A motivated, incentivized network will always outperform a hierarchy. We scale this network by activating the trust built by domain experts within their specialized communities. An ROI-driven grant program ($10K grants seeding exclusive content) converts followers into subscribers, with results measured by Annualized Revenue Return per grant guiding capital deployment.

End game: an ecosystem of intelligence with 800K experts and 200K AI developers and researchers, all working actively with high stake and high skin in the game on an evolutionary ecosystem of intelligence that picks the best response for any prompt and gains intelligence through usage VS employees working at tech giants trying to scrape data from the web and organize all the knowledge of the internet with all the bureaucratic top-down bottlenecks.

Who wins?

Where we Start

We are focusing on knowledge workers using generative AI to research and create. Our outputs are outperforming leading models in niche areas of knowledge where reinforcement learning adds multipliers on output quality.

Those domains include:

• Visual aesthetics with image and video generators

• Generative software specific to the web3 new internet stack

• AI assistants to help developers and customers onboard and succeed in the ecosystem

• Orchestrators and coordinators that facilitate collective intelligence and collaboration 

• Philosophy and advanced reasoning in systems theory

• Cryptography and decentralized trust

Customer Acquisition Strategy

• Paid subscriptions: to attract paid subscribers, we will work with domain experts and influencers on social media and content campaigns where the goal is to return each dollar invested within 6 months of paid subscriptions. With 10,000 USD invested into this strategy, we expect 500 paid subscribers at 30$ per month. The number might be lower at the start and we will optimise conversion and retention rate iteratively.

• Targeting specific knowledge producers and their followings via our deep analysis of social and online knowledge networks.

• Scaling customer acquisition through our pipeline of strategic partnerships with universities, AI startups, web3 protocols, media networks and niche knowledge networks.
  
  

Economic Model

• Initial Growth: Bootstrapping Quality We're seeding the network by paying domain experts and influencers in USD, focusing on high-impact verticals like visual aesthetics and advanced reasoning. Early traction shows we can recover customer acquisition costs within 6 months, with subscribers paying $30/month for access to superior outputs.

• Token Economy & Network Effects As we scale, all subscription payments automatically convert to NCO tokens, creating powerful network effects. At $2M monthly recurring revenue, this generates $2M in monthly token buy pressure. The system becomes self-reinforcing: miners compete for token rewards by improving outputs, validators stake on top performers, and rising quality drives customer demand - which directly funds the reward pool.

• Sustainable Alignment The key innovation is combining mining rewards with validator staking. This creates perfect game-theoretical alignment: miners optimize for quality, validators for accurate assessment, and customers fund the entire system through real demand. Unlike traditional token models, our rewards are backed by measurable customer value and verified performance improvements.

Every $1 of revenue creates immediate token demand. Every NCO reward vests over 800 days. That’s structurally engineered net buy pressure—and it's working.

Financials

We are selling 4 @ 48. Past contributors include Arca, Scott Moore (Gitcoin), Oak (Olas Network), Björn Wagner (Polkadot)

Use of Funds: 80% into growth grants, ROI-measured. Every dollar goes to scaling what’s working.