Roadmap
@lagrangedev has released its 2025 roadmap for its zkML framework - DeepProve - focusing on scaling verifiable AI: https://lagrange.dev/blog/lagrange-roadmap-2025…
Key goals include support for major LLMs (like LLaMA and Claude), private and confidential inferences, and new proof types - covering model training, fine-tuning, fairness, and reasoning.
DeepProve will be optimized for cloud environments, accelerated with custom hardware, and feature a parallel proving system that will boost performance for real-time use cases. Applications span defense, DeFi, healthcare, and content moderation, where provable AI decisions are essential - overall making AI verifiability an industry standard.
Partnerships
Lagrange also continued to extend the reach of DeepProve through several new partnerships focused on enabling verifiable AI in DeFi, rollups, and agent infrastructure:
• @arbitrum: Integrated DeepProve to bring verifiable AI agents to the onchain DeFi ecosystem, ensuring agent decisions are provable and confidential - a key step toward scaling trust in autonomous strategies: https://x.com/lagrangedev/status/1937133604666253319…
• @LitProtocol: Partnered to enable verifiable agent execution via integration with Lit’s decentralized key and encryption infrastructure. Developers can now launch AI agents that act autonomously while proving every inference cryptographically: https://x.com/lagrangedev/status/1936046391945839060…
• @alt_layer: Collaborated to introduce DeepProve into modular rollup environments, unlocking verifiable offchain computation and zkML capabilities for AI-augmented applications and agent coordination: https://x.com/lagrangedev/status/1935685394307117183
Tech
@NexusLabs introduced a scalable Storage Layer as part of its L1 stack, alongside Execution and Consensus. Using hyper-dimensional encoding and Data Availability Sampling, it enables low-cost, verifiable storage across decentralized networks.
Rolled out in three phases, the system supports auditable training data, provable model outputs, and resilient data access - laying the groundwork for transparent, trustworthy AI at scale with dramatically reduced storage costs: https://blog.nexus.xyz/nexus-storage-layer/…
Discussions
Nexus Chief Scientist @JensGroth16 and @QuillAI_Network CTO @0xakshat discussed the need for verifiable AI in a recent X Space talk: https://blog.nexus.xyz/a-conversation-with-jens-groth-and-quillai/…
Jens stated that as AI systems gain autonomy, trust must be earned through cryptographic proofs - not assumptions - highlighting blockchain as essential infrastructure for accountability. Both Nexus and QuillAI are building tools to verify AI actions across models, data, and decisions.
Partnerships
Nexus announced several partnerships aimed at scaling verifiable AI and improving digital trust infrastructure:
• @aiseerco: Partnered to support @facticityai in verifying text, video, and image content against misinformation. Nexus provides cryptographic tools to ensure the accuracy and integrity of fact-checks, enabling developers to build applications on provably reliable data: https://blog.nexus.xyz/ai-seer-partnership/…
• @sparsity_xyz: Collaborated to integrate Sparsity’s modular proving system with Nexus’ zkVM, powering scalable, auditable AI agents onchain. Through the Verifiable Intelligence Alliance, they aim to standardize verifiable AI infrastructure across DeFi, gaming, and enterprise use cases: https://blog.nexus.xyz/sparsity-ai-partnership/
Tech
@gizatechxyz has unveiled LuminAIR, a new framework for verifiable machine learning built with @StarkWareLtd’s S-two prover: https://x.com/gizatechxyz/status/1935345095068303482…
LuminAIR integrates Giza’s Rust-based Luminal system with cryptographic proofs, ensuring AI decisions are transparent and verifiable. Use cases include bias-free recommendation systems, provable DeFi agents, and AI watermarking - marking a shift toward trustless, provably correct AI in real-world applications.
The project also launched its autonomous yield strategy, Arma, in the @rainbowdotme wallet. Users can now manage capital without switching to external dashboards or dApps: https://x.com/gizatechxyz/status/1932420882296639596…
Adoption
Giza launched institutional use of its autonomous financial agents, with leading DeFi fund @Re7Capital the first to adopt the technology: https://gizatech.xyz/blog/re7-brings-the-first-wave-of-institutional-financial-agents…
These agents, already managing over $40 million in volume, offer optimized yield through non-linear strategies, on-chain transparency, and strict session-key controls. Re7’s pilot shows promising performance gains without compromising custody or requiring custom code.
Giza also expanded its integration with @MorphoLabs, with its autonomous agents directing over $41 million to Morpho Vaults. Most volume went to @SeamlessFi USDC Vault ($27.8M), followed by @MoonwellDeFi and @gauntlet_xyz.
Using real-time data and reinforcement learning, Giza’s agents optimize capital deployment by comparing yield against gas fees, ensuring efficient, rule-bound movement across new vault options: https://x.com/gizatechxyz/status/1929904765321740626
Launches
@zorotechnology’s AI Chat has officially launched, offering a Web3-based AI infrastructure built on @BNBCHAIN: https://x.com/zorotechnology/status/1935638688328761818…
Users can interact with live AI models, train them through prompts, and access tools for developing custom AI agents. The platform supports tasks, subscriptions, and API integration, with applications focused on crypto, robotics, and ZK ecosystems.
The project also unveiled its AI-on-Chain platform - supported by @GoogleStartups - enabling developers to create adaptive AI agents, build on-chain inference apps, and automate user-driven workflows.
The system converts tasks, behaviors, and sensor data into structured intelligence. Designed for real-world use, Zoro connects biometric and behavioral inputs directly to trainable AI pipelines: https://x.com/zorotechnology/status/1933117073607045463…
Roadmap
Zoro outlined a roadmap to bring Physical AI fully on-chain: https://x.com/zorotechnology/status/1932754697556836806…
In 2025, Zoro will launch an AI marketplace, mobile app, and real-time session memory. By Q4, smart devices and AI agents will connect through expanded sensors. In 2026, Zoro plans global scaling with SDKs, DAO grants, and a full data feedback loop.
Partnerships
Finally, Zoro has announced a wave of partnerships spanning compute, data, and AI agent infrastructure:
• @AIxBlock: Integrating Zoro’s global data pipelines with AIxBlock’s decentralized training tools to streamline the path from real-world data to usable AI models: https://x.com/zorotechnology/status/1935248907463704865…
• @NodeGoAI: Aligning Zoro’s data layer with NodeGo’s decentralized compute to enable efficient AI workflows from labeling to inference: https://x.com/zorotechnology/status/1934983802943586531…
• @qstnus: Enhancing AI training with verified user sentiment and opinion data collected through Web3-native surveys: https://x.com/zorotechnology/status/1934615713945137311…
• @dagama_world: Connecting verified geolocation and movement data to Zoro’s AI pipelines, unlocking spatial intelligence for on-chain AI: https://x.com/zorotechnology/status/1934574104507503068…
• @Dormint_io: Syncing biometric signals from wearables with Zoro’s training stack to improve health-focused machine learning: https://x.com/zorotechnology/status/1932779567036375142…
• @4everland_org: Powering Zoro’s AI infrastructure with decentralized storage and edge computing, improving scalability and data access: https://x.com/zorotechnology/status/1930263274630852674…
• @krain_ai: Supporting the creation and deployment of fine-tuned, on-chain AI models with full transparency and modular design: https://x.com/zorotechnology/status/1929526143121551370…
• @IQAICOM: Integrating the Aiden autonomous agent framework into Zoro's systems, enabling real-time, contextual support through intelligent agents: https://x.com/zorotechnology/status/1928407493337604097…
• @aagentai: Collaborating on smarter task automation and seamless, chain-abstracted AI agent infrastructure for DeFi: https://x.com/zorotechnology/status/1927777812041052237…
• @The_StarAI: Exploring user-owned AI by enabling the customization, training, and cross-chain deployment of personal AI agents: https://x.com/zorotechnology/status/1927251206847467927
Partnerships
@inference_labs announced a series of collaborations to advance trustless, verifiable AI infrastructure across Web3, compute, and communication layers:
• @GXChainGlobal: Brought zkML inference to REI’s gasless, high-speed EVM chain, enabling scalable and verifiable real-time AI for DeFi, GameFi, and beyond: https://x.com/inference_labs/status/1928090617357422687…
• @Laika_ai: Teamed up to expand secure, transparent AI tooling for analytics across DeFi and NFT markets, ensuring insight is always backed by provable inference: https://x.com/inference_labs/status/1929900481683550559…
• @immunefi: Joined the Magnus SecOps platform to enhance the security of decentralized AI agents with automated threat detection, CODEXA intel, and ZKP integration across the SDLC: https://x.com/inference_labs/status/1929950117789880663…
• @AscendEX_: Partnered with the global exchange to help introduce verifiable AI infrastructure into mainstream crypto platforms, emphasizing transparency over centralized black boxes: https://x.com/inference_labs/status/1932008519513092440…
• @Openpad_io: Collaborated on building trusted AI-driven tools for early-stage deal discovery and analytics, ensuring that insights in investing are both intelligent and provable: https://x.com/inference_labs/status/1933101434394067202…
• @dechat_io: Joined forces to bring AI verification to decentralized communication, enabling trustless intelligence in secure Web3 messaging protocols: https://x.com/inference_labs/status/1934875131148685676…
• @ar_io_network: Rolled out ZKP storage infrastructure and automated audits for any Bittensor subnet, enabling persistent, cryptographically auditable AI memory across 300+ Arweave gateways: https://x.com/inference_labs/status/1935355688412189012
Integrations
@SpaceandTimeDB has been integrated into AgentKit by @0xGasless, enabling AI agents to access real-time, ZK-proven historical blockchain data: https://x.com/SpaceandTimeDB/status/1936080291208044856…
This allows developers to create autonomous agents that monitor wallet activity, analyze protocol behavior, and execute transactions without gas fees or user signatures - making onchain decision-making faster, smarter, and based on verified historical insights.
Progress
@zk_agi have shared their monthly progress report, which includes updates on PawPad’s autonomous trading platform: https://medium.com/zkagi/zkagi-monthly-progress-report-may-2025-8b1828a88b86…
While the public beta was delayed to incorporate hackathon feedback, early trials with 100 users showed positive results. Key features include TEE-isolated wallets, real-time BTC signals, ultra-low latency execution, and a seamless Flutter-based mobile app. A limited beta began this month (June), with cross-chain and multi-agent support on the roadmap.
Partnerships
ZkAGI also partnered with the @Auroracoin_io to advance AI-powered internet capital markets. The collaboration includes sharing ZkAGI ecosystem deals with Aurora, improving Aurora’s AI infrastructure, and jointly working on community, business, and technical initiatives. Both teams aim to support future project launches and mutual growth:
https://x.com/zk_agi/status/1930256581066690956
Launches
@NetworkNoya shared an article explaining the launch of its AI-managed Omnichain MetaVaults, which optimize yield across 11 networks, 14 assets, and 65 lending pools: https://x.com/NetworkNoya/status/1933056499472851237…
Unlike passive strategies, NOYA’s system reallocates capital autonomously based on real-time market data, adjusting for fees, slippage, and volatility. Built on Morpho’s infrastructure, the platform promises higher returns and enhanced security.
The project also launched the beta version of its voice-enabled AI agent for DeFi, enabling users to perform actions like swapping, bridging, and sending using voice commands.
Built on NOYA’s infrastructure, it aims to evolve into a powerful, modular system for real-time execution and strategy: https://x.com/NetworkNoya/status/1935755570998657350
