The Lagrange Roadmap: AI’s Next Frontier (2025)

Artificial intelligence is transforming every aspect of modern life. From healthcare and defense to finance and digital infrastructure, AI is becoming the silent force behind the decisions that shape our world. But as the power of AI grows, so does our need to be able to trust it—progress in AI must be matched by progress in verifiability.

At Lagrange, we believe that the future of AI must be verifiable by design. Verifiability isn’t a nice-to-have, it’s the foundational standard that will determine whether AI can be safely deployed across critical sectors. That’s why we built DeepProve: the world’s leading zkML system for verifying AI outputs, bringing trust and safety to our AI-powered world.

The Journey So Far

Lagrange’s DeepProve began with a bold question: How can we verify the integrity of AI? More practically: What if every AI decision could be proven? Since then, we’ve made that vision real, expanding DeepProve’s abilities to prove inferences for LLMs and deploy verifiable AI across real-world applications.

Early on, we proved that cryptographic verification doesn’t have to live in the lab—it can work at scale, and it can power real products. That’s exactly what we showcased with Turing Roulette—the world’s first live, gamified AI prediction challenge, powered by DeepProve. Players competed with AIs and each other to determine whether or not their opponent was machine or human. The game drew over 500,000 unique players who generated over 3.75M proofs, demonstrating DeepProve’s ability to verify real-time inferences.

Since then, DeepProve has rapidly expanded—both in capability and adoption. Our technology is already being integrated by major players across Web2 and Web3, including:

• Enterprise AI partners like NVIDIA, with more major partnerships on the way
• AI x Crypto leaders like Sentient, 0G, Gaia, Mira, Billions and OpenLedger, with 20+ integrations in the pipeline

At the same time, our research team has continued to push the frontier of cryptography. To date, we’ve authored four papers pioneering novel mechanisms in zero-knowledge research, including Reckle Trees, DARA, and Dynamic zk-SNARKS, all of which were accepted into top academic and industry conferences.

DeepProve is more than an emerging technology. It is the cryptographic backbone for an AI-powered world that is safe, decentralized, and trustworthy. Our mission is to make proof-based verification the default standard for how machine learning is deployed across industries—so that every individual, anywhere in the world, can trust that the AI shaping their society is acting in their best interests.

Here's how we’re making that vision real…

The Journey Ahead

Product & Research

→ Verifiable Inferences for LLMs

DeepProve is scaling to support the world’s most important machine learning models. Support for LLaMA, Gemma, Claude, Gemini, and other large foundation models are on the way. This leap will enable:

• Verifiable inference for the most popular multimodal models
• zkML for DeFi, healthcare, defense, and more
• Confidential proofs, ensuring that proprietary models and sensitive inputs remain private

DeepProve’s flexible graph architecture makes it extensible by design—capable of supporting a wide range of AI models and workloads as the frontier of AI evolves. In other words, DeepProve’s internal design makes it not only powerful for today’s AI, but for whatever comes next. Moving forward, DeepProve will enable the cryptographic verification of popular LLM models used across major enterprise applications.

→ Expanded Proving Capabilities

Inference is just the beginning. Lagrange is conducting novel research to pioneer new types of proofs for the next generation of AI systems. Soon, DeepProve will support a variety of proof types that are essential to safe AI, regulatory compliance, and trust in critical infrastructure:

• Proofs of Training: Proofs that enable private, collaborative training with verifiable outcomes, attesting to the correct execution of the entire model training process. This will allow for verifiable model training across multiple parties (e.g., hospitals, banks, government agencies) and within single parties whilst protecting the privacy of sensitive datasets.
• Proofs of Fine-tuning: Proofs that confirm a pre-trained model was updated on an authorized dataset with specific training constraints. This verifiably ensures that the updated AI model is specialized to handle specific tasks and use cases.
• Proofs of Fairness: Proofs that verify a model’s outputs satisfy a predefined fairness constraint (e.g. demographic parity, equalized odds) without revealing the model or the training data. This is imperative for verifying that AI meets ethical and bias standards—for example, when deploying AI models across finance, education, and HR, where compliance with fairness regulations must be demonstrated verifiably.
• Proofs of Reasoning: Proofs that trace and verify why a model made its decision without revealing sensitive internal weights or user inputs. This will prove critical for systems in defense, healthcare, and policy, making it possible to deploy interpretable LLMs in regulated settings where decisions must be justifiable and the logical flow of a model’s decision-making verified.

→ Hardware Acceleration & Cloud Integrations

To support real-world adoption, DeepProve will be optimized for both centralized and decentralized environments.

We're working with ecosystem partners (including Intel and NVIDIA) to:

• Accelerate proofs with custom silicon and GPU primitives
• Integrate with cloud providers and data centers
• Enable developers to access DeepProve through simple APIs (plug-and-play)

→ Horizontal Scaling with Parallel Proving

The global adoption of verifiable AI will depend on its speed and scalability. In order to scale DeepProve, we’re pioneering new research for the development of a fully parallelized proof system that:

• Splits large inference traces into parallel tasks and distributes them across the Prover Network
• Aggregates proofs into a single, succinct proof for high throughput and minimal latency
• Reduces proof generation time by 5-10x (with more to come)

This will make DeepProve suitable for low-latency environments like high-frequency trading, autonomous vehicles, and smart contract execution.

→ Private Inferences for Confidential AI

We’re evolving DeepProve to enable private inferences where users can submit encrypted inputs and receive encrypted outputs with a proof verifying that the result is correct. This will open up a variety of critical use-cases for:

• Regulated industries like healthcare and finance
• Government use cases involving classified data
• Any user who values privacy by default

Ecosystem

→ Real-World Deployment for Mass Adoption

DeepProve is already being deployed in the real world. As more and more industries adopt AI-first systems, verifiability will be crucial to ensuring that the innovations we develop are safe and free from error. Here are the key areas DeepProve will expand to:

• Defense: Modern defense infrastructure is rapidly adapting to the use of AI for autonomous systems, fleet management, and cyber operations. DeepProve brings cryptographic guarantees to these mission-critical systems—proving that drone decisions, battlefield AI, or satellite intelligence systems operate securely and correctly. It also enables encrypted coordination between agent systems while preserving command confidentiality.
• DeFi: As decentralized finance becomes increasingly agent-driven, the need for verifiable strategies is critical. DeepProve ensures that AI agents trading on-chain can prove their actions were executed correctly—without revealing their underlying strategies. This enables new levels of trust in high-frequency trading, market-making bots, and decentralized asset management.
• Finance: From algorithmic risk models to compliance AI, the financial sector is under immense pressure to prove that its decision-making systems are trustworthy. DeepProve allows institutions to deploy AI with built-in transparency—verifying decisions like loan approvals, fraud detection, and portfolio optimization, all without compromising sensitive user or institutional data.
• Healthcare: In medical AI, trust can be a matter of life or death. DeepProve allows hospitals and health tech platforms to validate AI diagnoses—ensuring they’re generated from FDA-compliant models on clean datasets—while keeping personal health information private. This opens the door for broader adoption of AI in clinical diagnostics, imaging, and personalized medicine.
• Media & Moderation: In the era of misinformation and synthetic content, platforms need to know if AI-generated media is accurate—or if moderation tools are acting fairly. DeepProve powers content verification and moderation systems that can prove their integrity, detect deepfakes, and ensure consistent, bias-free enforcement across massive online communities.

→ Industry-Leading Partnerships & Integrations

Lagrange is actively collaborating with the pioneers shaping the future of AI, cryptography, and decentralized infrastructure. As our ecosystem grows, so will the number of integrations, SDKs, and supported use cases. You can expect continued rollouts across major enterprise partners in Web2 and Web3.

• Web2 Giants: Our strategic alliance with NVIDIA unlocks deep access to the world’s most advanced AI hardware and research tooling. With upcoming partners, we’ll be accelerating DeepProve at the silicon layer, bringing zero-knowledge proofs into direct alignment with the next generation of high-performance chips. These partnerships are foundational to scaling verifiable AI into real-world enterprise applications—and we’re just getting started.
• Web3 Infrastructure Partners: Onchain, we’re working closely with protocols like 0G Labs, Gaia, and others to deploy DeepProve into rollup ecosystems and modular networks. These integrations bring verifiability into the heart of decentralized applications, powering transparent DeFi execution, fair gaming systems, and secure, auditable AI tooling for onchain governance and autonomous agents.

→ Staking Rewards for The $LA Token

The $LA token aligns the interests of clients, provers, and token holders for decentralized proving in the Lagrange Prover Network, powering the proofs needed for verifiable AI via DeepProve. Put simply, proof demand = token demand—proof fees are converted to $LA and distributed to provers, reinforcing the demand for $LA alongside protocol emissions and subsidies. Upon launch, the $LA token listed on major exchanges including Coinbase, Binance Futures, Binance Alpha, OKX, Upbit, Bybit, HTX, Bitget, Bithumb, Gate, Kucoin, MEXC, Crypto.com and more.

The Lagrange Foundation is continuing to build value for token holders and will be rolling out the final pillar of decentralization: empowering $LA holders to delegate their tokens to provers across the network. This mechanism will directly shape the network’s computing power and play a critical role in its economic design.

‍Community

The future of AI won’t be built by engineers alone—it will be shaped by a global community that demands transparency, accountability, and trust. At Lagrange, community is core to our mission. That’s why we’re investing heavily in culture, access, and on-the-ground momentum.

→ Global Expansion

Our community in Asia—especially Korea ♥︎—is growing rapidly, with some of the most engaged $LA holders in the world. We’re actively working with leading enterprise AI and infrastructure partners across the region to expand our footprint across Korea, Japan, China, and beyond. We’re also doubling down on regional events, translated content, and in-person activations to maintain and grow the community’s momentum. Expect to see Lagrange show up in more APAC cities throughout 2025, with language-localized programs and region-specific integrations on the horizon.

→ Experiences for $LA Token Holders

The official Discord for the Lagrange Foundation has become the beating heart of the $LA ecosystem. With token-gated roles across three key tiers—Goldfish (1,000 $LA), Whales (10,000 $LA), and VIPs (100,000 $LA)—$LA holders can now unlock access to private channels, protocol updates, and exclusive experiences. Stay tuned for $LA-only private gatherings, limited-edition merch, curated rewards, and community raffles you won’t find anywhere else.

→ Online and IRL Events

Throughout the year, we’ll be curating more unforgettable events that bring the Lagrange community together. From online summits to the world’s largest Web3 and AI conferences, our activations are designed to connect and amplify the people shaping the future of verifiable AI.

Catch us at:

• Korea Blockchain Week (Sept 22–28), 
• TOKEN2049 Singapore (Oct 1–2) 
• Devconnect Argentina (Nov 17–22)

A New Standard for AI

Lagrange isn’t just building technology—we’re defining a new standard for how AI should operate in the world. By proving the outputs of AI models, we are proving that AI itself can be deployed safely, reliably, and securely across every domain that matters: healthcare, defense, finance, media, governance, and beyond. Through technological innovation, world-class research, and a global ecosystem of partners, we are pioneering the future of safe AI.

The future of AI starts with ZK. The future of humanity starts with Lagrange.