Vision

Nethermind's vision is to develop and design the fundamental technologies underpinning safe, ubiquitous, and trustless interactions between robots, agents, humans, and the real world. The ultimate goal is to have completely autonomous and decentralised construction.

At the heart of our vision is a relentless focus on the following key themes:

As autonomous agents and robots become ubiquitous, establishing verifiable digital and physical identities for humans and machines (from now on, ‘actors’) will become fundamental to secure their interactions. As we did for cars, we will need much more sophisticated methods for authenticating billions of drones, humanoid robotics and other systems operating in the public domain.
With AI now able to make decisions, we can have “AI as smart contract,” which can operate and decide over arbitrary data. With TEE (phala network) and ZK, we can make AI decision-making permissionless while keeping the AI operation uncensorable. “AI as smart contract” is inevitable. AIs can now operate as specific roles in a network (e.g., content policy, decision maker, accountant, legal defence, reason over real-world data, etc. ).
As data integrity becomes more important, the world will build systems for cryptographically regaining trust in media. Our goal is to do part of this onchain. At the very least, these trust systems must be available to “AIs as smart contracts”. We will also increase the efficacy and introduce proofs of location, orientation, time, etc. With the developments of TEEs and PUFs, our ability to create trust networks over arbitrary sensor data will be unlocked. This is already happening. We must act fast.

With these three components, we can create autonomous onchain systems which can operate over the real world. Using authenticated data, authenticated actors can prove the state of the world to onchain protocols; permissionless actors can reason about them and decide on further actions. By solving the oracle problem, blockchain can disintermediate human trust in the real world.

Motivation

Nethermind is driven to make AI and Blockchain a reality. We are a 200+ strong organization of engineers and experts with a track record of building and designing software and protocols for some of the blockchain industry's leaders.

The scale and pace of technological change make this transformation inevitable:

The amount of information created by humanity is projected to grow to 175 Zettabytes by 2025 - that's 175,000,000,000,000,000,000,000 bytes or ~2.3bn years of HD video. (IDC/Seagate)
The capabilities of AI systems and Robots are improving exponentially. Each new model release represents a leap forward in the applicability of general-purpose AI models to new use cases.
Since 2022, over 15 billion images have been created using AI, with another 30+ million added daily. Humans already find “disinformation generated by AI may be more convincing than disinformation” (humans were 3% less likely to spot false tweets generated by AI than those written by humans). It has never been cheaper or easier to create convincing but false data at any other point in human history, and we expect this trend to continue.
The number of robots worldwide has multiplied threefold over the past two decades to 2.25 million, with projections suggesting that 20 million manufacturing jobs could be displaced by 2030 (*Oxford Economics). We expect exponential growth over the next few years.*
The IoT suppliers' market is expected to reach approximately $500 billion by 2030, with over 40+ billion devices expected to be deployed (IoT Analytics).

We believe it will be feasible within 1-5 years to deploy systems that can change physical space without any human intermediary. In the future, complex AI systems will control the ownership and operation of these systems. We assume that rational AI will value the efficiency and openness of crypto to deliver its goals. As autonomous agents and robots become ubiquitous, establishing verifiable foundations for the root of all interactions between human agents and machines will be fundamental to realizing a future where blockchains and trustless interactions underpin society.

It's clear that when robotics comes online in the public sphere, AI and crypto will be key to making it work well. Initially, we believe that robots will roll out through large, centralised entities. Still, the system created will be 100x times more efficient when the connections between robotic control systems are interoperable, verifiable, and largely permissionless.

The crypto holy grail efficiently applies capital to transform the physical world with open, efficient markets. That is what Nethermind is now setting out to do.

The roadmap

The roadmap is simple:

Control layer:
1. build infra for AI systems to operate autonomously onchain (partially done by Phalla network and Flashbots)
2. build infra for onchain AI to adhere to a set of protocol rules (in progress, somewhat present in virtuals.io and Eliza)
3. build infra to bring trusted data and “reality” onchain or to the AIs (started by Nethermind and StarkWare exploration team)
4. build systems to identify actors (humanity: World, robotics: some prior art but not much, agents: some prior work exists)
5. build out examples of blockchain networks using these systems (in progress; see below)
  1. verify the current state of a construction site and track all elements about it onchain (in progress; see below)
  2. build a system for actors to demonstrate work delivered to some onchain AI system
Decentralised manufacture and delivery
1. build systems to verify the physical properties of objects trustlessly
2. build systems to verify the delivery of objects and the object's identity
3. build systems to manufacture objects trustlessly
4. build out a working network combining all of these systems
  1. Decentralised 3d printing network
  2. Decentralised milling market
Decentralised Construction
1. build systems to track physical space onchain
2. build systems to track ownership of physical space onchain
3. build systems to transform physical space autonomously

Stage 1 - Control layer

The first step to achieving decentralised manufacturing is to bring AI systems, trusted data on reality and identity into crypto so that these systems can interoperate easily. Luckily, many parts of this have already been accomplished - the crypto industry has realised the need for trusted AI execution through the rise of onchain agents and autonomous X bots.

The most important observation for this stage is that an AI with good adherence to a goal can fill a complex role in an onchain protocol. For example, in a protocol where humans can compete to complete work for AI systems, a permissionless, uncensorable AI can scan all work requests to prevent obvious tasks that induce human harm. We now have an ‘ethics AI’, which a protocol can permissionlessly adopt to police itself. Another can scan all evidence that the work has been completed, etc.

Core requirements:

We require that agents be secure, with public testing of LLMs and other systems available to all, and tooling for verifying the models and systems these agents consist of that are accessible.
We require that reality can be verified. Agents must be able to comprehend and access reality without human trust.
We require that robotic systems be identifiable and reputation-carrying.

With these in place, you can construct arbitrary systems of operation and trust between humans, AI, and reality.

Autonomous AI

We believe in a long-term vision of self-sovereign AI systems, interacting in chorus with each other and their surrounding digital and physical worlds in coordination - serving as vessels for creating and moving capital around the economy.

Autonomous AI (ZK-ML/Trusted Execution Environments)
AI security
AI economy
Decentralised compute networks

Verifiable reality

Project Truth

The origin of all information should be attributed to humans, agents, or models' machines and robots. A robust verification system should not only verify the information itself but also authenticate the identity of its source (Provenance). This includes:

Establishing cryptographic signatures, timestamps, and metadata that verify:
- Who/what created the information (human, AI model, device or combination)
- When the information was created
- Under what conditions or context it was created
- The original unmodified state
- The chain of custody from creation
Real-World Oracles

‣
- Verifiable sensors (C2PA)
- Authenticity scanning
- Verifiable Location Data
- Web2 → Web3 (ZkTLS)
Secure & Verifiable Data Marketplaces

Verifiable data market

Robotic identity / Agentic Robotics

As digital spaces become increasingly crowded with non-human agents, and robots operate in open environments - humans will need tools that harness unique signatures of identity to distinguish themselves from robots.

Identity primitives are formed from a secure Root-of-trust, an anchor point from which assurances can be generated and attested to and the basis of all interactions between Humans, Machines and Agents.