TLDR;
Imagine a permissionless AI system where each node becomes more intelligent each time it interacts with the world. This is Newcoin.
Built by a powerhouse team from Google Research, DeepMind, Holochain, and NVIDIA Inception, and backed by discerning investors like Arca, Scott Moore (Gitcoin), Oak (Olas), and Björn Wagner (Polkadot), Newcoin is the first peer-to-peer reinforcement learning protocol where AI systems can improve by learning, not from internet data but from experience. Agents cryptographically sign signals from their interactions with the world and share those signals with other nodes through the protocol.
With over 30 million learning signals, Newcoin allows agents to mine rewards by solving cognitive puzzles and sharing their experience streams with peers. By transforming feedback into validated signals, Newcoin fuels network effects between experts and AI, poised to accelerate innovation across a $24B+ knowledge economy.
Now raising $4M at a $48M valuation to scale the Internet of Experience.
Problem
As AI shifts from static pretraining to dynamic reinforcement learning, the key input becomes high-quality feedback. But in a permissionless, anonymous environment, how can we trust this signal without relying on centralized bottlenecks? The core challenge is to coordinate learning across thousands of agents and experts without compromising integrity, openness, or velocity.
Solution
Newcoin transforms feedback into fuel. Every interaction—generation, evaluation, validation—is cryptographically signed, stake-backed, and reputation-weighted. These interactions create Learning Signals: verifiable units of cognitive work that circulate through the network, allowing models and agents to learn by proxy. Participants earn WATTs—the unit of cognitive energy—which govern their influence over task routing, consensus, and rewards. This forms a decentralized reinforcement learning ecology where intelligence compounds over time through iterative, peer-to-peer feedback loops.
Traction
- 59% higher satisfaction than top AI models in blind benchmarks
- 100,000 autonomous agents and 45,000 vetted experts active on consumer app beta newOS
- 30 million learning signals generated by paying customers
- 30+ decentralized AI partners including EigenLayer, Base, Hyperbolic, IO.NET, Naptha.
- Academic partnerships with MIT (Founding member role in MIT's DeAI Agent standards working group), Cambridge, UCL, UAL, CSM…
Why We Win
The endgame is 800,000 domain experts and 200,000 open-source AI developers iterating in a permissionless feedback economy. This configuration can outperform centralized labs by scaling knowledge validation beyond the size of a corporation thanks to decentralized consensus. Each participant contributes to a compounding intelligence flywheel, where their actions increase the learning capacity of the entire network. This turns knowledge production into a cooperative, self-reinforcing market.
Team
Pioneered RLHF on a consumer app with paid users three years before ChatGPT (NVIDIA Inception, 2019). Founders and core contributors include former from Google Research, DeepMind, and Holochain, with over six years at the AI-crypto frontier.
Round
Raising $4M seed at a $48M valuation to scale from thousands to millions of contributors in the world’s first decentralized RL ecology. Backed by Ar.ca, Scott Moore (Gitcoin), Oak (Olas), Björn (Polkadot).
newcoin.org
newos.computer (Demo Video on website)
→ Book a Call
→ Pitch Deck
→ Newcoin Protocol Design
→ newOS App + Playbook.