AI & Crypto, Past, Present and Future – Day 90, Onchain AI: The New Era of Decentralized Intelligence
How open-source LLMs, ZK-powered verification, and decentralized networks are reshaping the programmable stack for AI in crypto.
AI is moving from the cloud to the blockchain. This new wave of development is not just about integrating models or building chatbots. It is about running open-source language models onchain, verifying AI computation with cryptographic proofs, and creating decentralized networks for training, inference, and data exchange. Today’s article breaks down the most important live experiments, why they matter, and what the next generation of onchain AI could unlock.
1. Open-Source LLMs Go Onchain
Recent months have seen a push to bring large language models (LLMs) directly onto blockchain platforms. Rather than relying on closed APIs or centralized inference, teams are experimenting with ways to deploy, execute, and verify model outputs in public environments. This unlocks new levels of transparency, composability, and censorship resistance. It also opens the door for smart contracts and DAOs to call AI models as part of their core logic, making protocols smarter, more responsive, and able to reason over complex, real-world data.
2. Giza: Verifiable AI Inference Onchain
Giza is building a platform to run and verify AI models within blockchain environments. By leveraging zero-knowledge proofs, Giza allows smart contracts to trust that a given AI computation was performed correctly, even if the actual model run is offchain. This means protocols can rely on AI-powered outputs, such as predictions, classifications, and optimizations, without having to trust a centralized server or oracle. Giza’s approach is already enabling onchain use cases for generative art, DeFi risk scoring, and verifiable automation.
3. Ritual: Decentralized AI Training and Inference
Ritual is building a decentralized network for AI compute, training, and inference. Instead of relying on a single provider, Ritual allows anyone to contribute resources, train models, and serve inference tasks across a distributed network. This approach is designed to increase resilience, reduce costs, and democratize access to powerful AI. By combining decentralized infrastructure with token incentives, Ritual is creating the foundations for a truly open AI economy, where models, data, and compute are all shared and governed onchain.
4. Modulus Labs: AI and Zero-Knowledge Proofs
Modulus Labs is focused on making AI inference provable using zero-knowledge (ZK) technology. Their work enables smart contracts and DAOs to verify that an AI model was run with a specific input and produced a specific output, all without revealing the model weights or raw data. This solves a key problem for using AI in DeFi, identity, and privacy-sensitive applications, where trust and verification are critical. Modulus Labs has demonstrated ZK-proven inference for small models and is now working on scaling these techniques for real-world LLMs and more complex use cases.
5. Why This Matters
Transparency: Onchain AI and verifiable inference make it possible to audit, trace, and trust model outputs. This is crucial for financial, governance, and compliance use cases.
Composability: Smart contracts and DAOs can use AI models as plug-in logic, enabling new forms of automation, risk management, and user experiences.
Censorship resistance: Decentralized AI networks reduce single points of failure, allowing critical infrastructure to stay online and open, even in adversarial environments.
6. Key Challenges
Scalability: Running large models onchain, or even verifying them with ZK proofs, is expensive and complex.
Privacy: Protecting user data and model weights while enabling verification is an active area of research.
Incentives: Decentralized training and inference networks need robust mechanisms to ensure quality, uptime, and fair rewards for participants.
The most successful projects are those that strike a balance between performance, transparency, and usability.
7. What’s Next
More onchain AI experiments: Expect to see more DAOs, DeFi protocols, and dApps plugging in verifiable models for everything from credit scoring to NFT curation.
Hybrid approaches: Offchain compute with onchain verification and settlement will become the norm for high-value or compliance-sensitive use cases.
Open-source model marketplaces: Communities will publish, share, and govern AI models as public goods, with cryptoeconomic incentives for quality and performance.
8. Reflection
Onchain AI is not just hype. It is a fundamental reimagining of how intelligence, value, and infrastructure will interact in the programmable economy. Projects like Giza, Ritual, and Modulus Labs are building the rails for transparent, composable, and censorship-resistant AI in crypto. As these systems mature, the line between smart contract and smart agent will begin to blur, opening up new frontiers for what blockchains can automate and secure.
Hasta manana
Cpt