Ritual Explained: Crypto’s Bet on Trustworthy On-Chain AI

Ritual has spent close to three years pitching itself as the trust layer artificial intelligence never had: the piece of infrastructure that lets a smart contract call a language model and actually verify what comes back. As of this month it is still, technically, a gated testnet with no live token and no announced mainnet date. That is not automatically a red flag. Getting a blockchain to talk to a GPU cluster without quietly becoming centralized is a harder problem than most crypto teams have tackled, and Ritual has raised real money, signed real partnerships, and shipped real, open-source code while working on it. This piece walks through what Ritual actually is, who built it, how the architecture works, what has shipped so far, and where it sits in a crowded field of projects promising to decentralize AI.

What Is Ritual, and Why Crypto’s AI Crowd Keeps Talking About It

Ritual describes its own end goal as a “sovereign, community-owned” execution layer for artificial intelligence: an open network where anyone can plug in compute, a model, or an application, rather than routing every request through a handful of corporate APIs. Its two co-founders, Niraj Pant and Akilesh Potti, met as investors at Polychain Capital and started the company in 2023 on a simple observation: crypto had spent a decade building censorship-resistant money, while AI was busy consolidating into the exact kind of chokepoints crypto was supposed to route around.

The project ships in stages. The first, called Infernet, is a decentralized oracle network that lets smart contracts request an AI computation, such as running a model or fetching an inference, and receive the result on-chain. The second, much larger piece is Ritual Chain, a purpose-built layer 1 that treats AI inference as a native transaction type rather than something bolted on through an external oracle. Wrapped around both is the Ritual Superchain, a planned family of specialized app-chains that share Ritual’s security and tooling. None of this exists in a vacuum: Ritual competes for attention and capital with Bittensor, Gensyn, Nous Research and a growing list of projects chasing a similar thesis, that AI compute and model access should be a public good coordinated by a blockchain rather than a handful of hyperscalers.

The Problem Ritual Says It Is Solving

Ritual’s pitch starts from a specific complaint about the state of AI in 2026. Frontier model access runs through a small number of companies, GPU supply is rationed by a handful of cloud providers, and there is no cryptographic way for a smart contract, or a human, to confirm that a claimed model output actually came from the model it says it did. Ritual’s own writing frames this as four linked failures: no computational guarantees around integrity, privacy or censorship-resistance; concentrated ownership of the APIs that gate access to capable models; scarce and expensive hardware; and reward structures that pay closed labs while giving open-source model builders little way to monetize their work.

That framing deserves to be treated as a real critique rather than pure marketing, but it also invites an obvious counter-question: can verified, decentralized inference ever be cost-competitive with simply trusting a centralized provider? The wider decentralized-AI category has not settled that question. Combining cryptography, blockchain consensus and machine learning performance at scale is a genuinely hard engineering problem, and every network in this category, Ritual included, still competes for the same limited pool of GPUs that centralized clouds are also bidding for. The industry has broadly settled on four ways to make an off-chain AI computation trustworthy on-chain: zero-knowledge proofs, which are cryptographically airtight but currently too slow and expensive for large models; trusted execution environments, which are fast but require trusting the chip vendor; optimistic schemes that publish a result and leave a challenge window for anyone to dispute it; and crypto-economic approaches that rely on staking and slashing rather than either cryptography or hardware. Most serious projects in the space, Ritual among them, end up picking a lane, or building a hybrid, rather than solving all four at once.

From Two Polychain Investors to a Protocol

Niraj Pant joined Polychain Capital as an intern at nineteen, dropped out of the University of Illinois a month later, and spent roughly six years there as a general partner, working on more than thirty investments including dYdX, Compound, Offchain Labs and EigenLayer. Akilesh Potti was a partner at the same fund, having previously worked on machine learning at Palantir, done quantitative and high-frequency trading work at Goldman Sachs, and carried out ML research at MIT and Cornell. The two spent three years working together at Polychain before leaving to start Ritual in 2023, betting that their combined view of crypto capital markets and applied machine learning was a defensible edge.

Ritual’s early team drew from DeepMind, Paradigm, Coinbase, Citadel and Palantir, according to The Block’s coverage of the company. The advisory bench reinforces the crypto-plus-AI framing: NEAR Protocol co-founder and Transformer paper co-author Illia Polosukhin, EigenLayer founder Sreeram Kannan, and Gauntlet chief executive Tarun Chitra all sit on it. In January 2024, former BitMEX chief executive Arthur Hayes joined as an advisor, saying he wanted “to ensure the burgeoning AI economy has access to a more censorship-resistant, collaboration-powering technology than we currently have,” according to The Block. Hayes said he expected Ethereum to underpin what he called the future human and AI economy, a framing that lines up with how often Ritual leans on Ethereum-adjacent infrastructure rather than building in isolation.

Inside the War Chest: Funding, Investors and Advisors

Ritual closed a $25 million Series A in November 2023, led by crypto venture firm Archetype, with participation from Accomplice, Robot Ventures, dao5, Accel, Dialectic, Anagram, Avra and Hypersphere, plus angel checks from Balaji Srinivasan and Polosukhin, according to Cointelegraph. Five months later, Polychain Capital, the fund where both founders had worked, added an undisclosed “multimillion dollar” strategic investment to the same round, a detail CoinDesk reported at the time. By the time Ritual Chain’s testnet launched in November 2024, cumulative disclosed funding had passed $30 million. None of this money appears to have gone toward a token sale: Ritual has not run a public sale, and the amounts above cover equity or equity-like venture financing rather than any token allocation.

The table below summarizes what has been publicly disclosed. Ritual has not announced a new priced round since the Polychain top-up, and has not confirmed a company valuation at any stage.

Infernet: The Oracle Network Where Ritual Actually Started

Infernet is the part of Ritual that already runs in production. It works like a specialized oracle: a smart contract fires off a request, for example asking to run inference against a specific model, and independent Infernet nodes pick up the job, run it off-chain, and deliver the result back on-chain through the Infernet SDK, along with proof or attestation data depending on how the job is configured. Developers plug it in through open-source tooling published under the ritual-net organization, including an Infernet node client, a router for distributing requests across nodes, and a container starter kit with working examples ranging from a basic model call to full Torch and ONNX model deployments, all documented on Ritual’s developer site.

Because Infernet operates at the request layer rather than replacing the underlying chain, it can serve contracts on Ethereum mainnet and rollups like Base and Arbitrum without those chains needing to change anything about how they operate. That matters for adoption: rather than asking developers to migrate to a new chain for AI functionality, Infernet lets an existing DeFi protocol, prediction market or game bolt on model access much the way it might bolt on a price feed. The tradeoff is that Infernet alone cannot offer the same execution-level guarantees Ritual Chain is designed to provide, which is exactly why Ritual treats it as a bridge to the full chain rather than the end state.

Ritual Chain: Superposition, TEE-EOVMT and an EVM Built to Wait on AI

Ritual Chain is where the architecture gets genuinely unusual. Ritual’s developer documentation describes it as a sovereign, EVM-compatible layer 1 running what it calls TEE-EOVMT, shorthand for an EVM extended with off-chain verifiable machine tasks. The core idea, which the team calls Superposition, is that two execution paths coexist over the same state. Ordinary, deterministic operations, like a token transfer or a storage write, follow the standard EVM path and get replicated and checked by every validator the usual way. Non-deterministic workloads, like an LLM inference call, an HTTP request, or agent orchestration, get delegated instead: they run inside a trusted execution environment, and validators check a cryptographic attestation of the result rather than re-running the computation themselves, per the same documentation.

That is a meaningful departure from how most chains handle non-determinism, which is usually to forbid it outright, since two validators running an identical non-deterministic model call could get two different answers and break consensus. Ritual’s bet is that enshrining non-determinism, instead of designing around it, is what actually lets a contract trust an AI output rather than merely receive one. Staying EVM-compatible while doing this is itself a deliberate choice: it means Solidity contracts, existing wallets and standard tooling mostly still work, so builders are not asked to learn an entirely new virtual machine just to get access to native AI calls. In the broader verifiable-compute taxonomy described above, this places Ritual squarely in the trusted-execution-environment camp: it trades cryptographic certainty for lower overhead and faster response times, at the cost of trusting the hardware vendor and the attestation chain underneath it, the same tradeoff made by TEE-based compute networks elsewhere in the sector.

The Scheduler, Precompiles and Wallets Built for Agents, Not Just Humans

A few specific design choices in Ritual Chain point at who it is actually built for: autonomous agents that need to act without a human clicking confirm every time. The chain ships with a Scheduler, a protocol-level contract that lets any other contract register a callback and a future execution schedule, so it can trigger itself at a later block without relying on an external keeper bot or cron job. It also introduces a dedicated transaction type built around WebAuthn passkeys, letting a wallet authenticate without a seed phrase, a small but deliberate nod to the reality that an autonomous agent needs a way to hold and move funds that does not depend on a human memorizing twelve words.

Ritual’s precompiles are split by how long they take to answer. Fast, synchronous calls return inline within the same transaction. Short asynchronous calls, expected to resolve within a couple of seconds, get their result injected directly into the transaction receipt. Long-running asynchronous jobs, the category that covers agent orchestration, image generation or zero-knowledge proof generation, work in two phases: the contract commits to a request and receives a task identifier, then a separate callback transaction delivers the result once it is ready. That two-phase pattern solves the compute side of a problem crypto is also trying to solve on the payments side, namely how an autonomous agent transacts without a human in the loop; Ritual’s push into AI agent and gaming payments is one example of that same problem being tackled purely at the settlement layer rather than the compute layer.

The Ritual Superchain and the Partners Building on It

Ritual Chain is meant to be the base layer for what the team calls the Ritual Superchain, a family of sovereign, modular execution layers, each carrying its own set of Stateful Precompiles tuned to a specific class of AI workload, whether that is classical machine learning models, image generation, or full foundation-model inference. The idea echoes how the Ethereum rollup ecosystem fragmented into specialized chains for gaming, payments and DeFi, except Ritual wants the specialization axis to be AI workload type rather than general-purpose throughput.

None of this happens without outside infrastructure, and Ritual has been unusually deliberate about picking partners that plug specific gaps. Celestia supplies modular data availability, letting Ritual publish the data generated by AI workflows as blobs rather than bloating its own state. EigenLayer, now operating under the EigenCloud brand, announced a partnership with Ritual in February 2024 to build Actively Validated Services around Infernet and Ritual Chain, an arrangement EigenCloud’s own writeup describes as covering operator bootstrapping, economic security with slashing for dishonest behavior, dual staking, and model routing. It is a fitting pairing on paper: Sreeram Kannan, EigenLayer’s founder, is a Ritual advisor, and Niraj Pant backed EigenLayer himself during his time at Polychain. Restakers who opt into these AVSs get paid for keeping Ritual’s compute honest in roughly the same way any staker gets paid for securing a base chain, just against a different kind of fault; readers newer to that mechanic may find our explainer on how stakers actually get paid useful background before judging whether the yield on offer compensates for the risk.

Story Protocol and generative AI platform MyShell joined in March 2024, giving creators a way to register AI models and their outputs as on-chain IP assets with watermarking and provenance tracking, a partnership CoinDesk covered when MyShell, a creator app with more than a million users at the time, became the first application to route through it.

Where Ritual Actually Stands in July 2026

Strip away the roadmap language and the honest status update is this: Ritual Chain has been in testnet since November 18, 2024, when the network launched alongside the newly independent Ritual Foundation, the non-profit now tasked with taking the chain to market. Co-founder Akilesh Potti described the milestone at the time as enabling “entirely new user behavior that wasn’t possible in any other system today for interacting with AI.”

Since then, the rollout has moved through a long sequence of community programs rather than a single big-bang launch: an Ambassador Program opened in June 2025, Ritual Quests began that September, a 21 Days of Prompts campaign ran through early October, a Builders Program opened in December 2025, and a gated testnet phase restricted to holders of specific Discord roles went live on April 24, 2026, according to a program timeline tracked by CryptoRank. The most recent milestone, a phase the team calls Genesis 1000 that lets early participants deploy their own agents on the network, opened on June 24, 2026, three weeks before this article. None of that is a mainnet, and Ritual has not published a mainnet date.

The Token Question: What RITUAL Is and Isn’t Yet

There is no live, tradable RITUAL token as of this writing. The testnet uses a faucet-distributed test asset, also called RITUAL, purely to pay for gas while builders test the chain; it has no market value and is not the same thing as a mainnet token generation event. Community trackers list Ritual’s airdrop status as potential with a reward date still to be announced, which is a polite way of saying the team has not committed to a distribution mechanism, an allocation, or a timeline in public. In most networks that follow this playbook, testnet quest points and node-operation history end up mattering for an eventual distribution, but until Ritual publishes real tokenomics that remains an assumption, not a promise.

That puts Ritual meaningfully behind some of its direct competitors on this specific dimension. Gensyn, a rival decentralized-compute network, went from mainnet to a live, trading token in about a week this spring. Nous Research, which is building decentralized model training rather than inference, has not launched a token either, but has at least attached a public number to the plan: its April 2025 Series A valued the future token at $1 billion even though it had not launched, according to The Block’s coverage of that round. Ritual has disclosed neither a token allocation nor a target valuation, which makes it hard for outsiders to judge how an eventual distribution might treat testnet participants relative to investors.

The Competition: Bittensor, Gensyn, Nous Research and the Rest of the Field

Ritual is one entrant in a decentralized-AI category that CoinGecko currently sizes at roughly $21 billion in combined market capitalization, though that figure blends genuinely AI-native protocols with chains, like Chainlink, that added AI-relevant features onto an existing oracle business. Chainlink alone accounts for close to $6 billion of that total, more than the next two AI-specific tokens combined, a reminder that Ritual’s most entrenched competitor for the position of “the oracle that feeds AI data on-chain” may not be a dedicated AI-crypto project at all.

Among the AI-native networks, Bittensor remains the category’s largest by market capitalization, built around a subnet model where dozens of independent teams compete to provide the best machine intelligence for a narrow task, from text generation to financial prediction, and get rewarded in TAO based on a peer-ranked quality score. Gensyn approaches the same broad problem from the training and settlement side: it launched its mainnet on April 22, 2026 alongside Delphi, an AI-settled prediction market platform, and its AI token began trading a week later in one of the more volatile debuts of the year, surging over 250 percent before giving back roughly half of that gain, according to The Block. Nous Research is chasing a narrower problem still, decentralized pretraining itself, coordinating GPU contributors through its Solana-based Psyche network; Paradigm partner Arjun Balaji, whose firm led that $50 million round, described the approach as “a powerful contrast to the closed, centralized efforts” dominating frontier model training.

The rest of the field is long and getting longer. Akash Network runs a Cosmos-based reverse-auction marketplace for GPU rental. Sentient is building an open routing layer it calls GRID across models, agents, data and compute, and launched its SENT token in late 2025. Smaller players like Allora and io.net chase adjacent niches in collective intelligence and compute aggregation. Ritual’s distinction from most of that list is architectural ambition: it is not simply renting GPUs or ranking model outputs, it is trying to make AI inference a first-class, natively verifiable operation at the base layer of a blockchain, closer in spirit to what Ethereum did for smart contracts than to what a compute marketplace does for GPU rental. Whether that ambition is rewarded before a faster-moving competitor locks in developers is the open question hanging over the whole category.

Regulation: Where Project Crypto’s Safe Harbor Leaves a Network Like Ritual

Ritual has never publicly framed itself as at risk of being called a security, and nothing about Infernet’s or Ritual Chain’s current testnet-only status has drawn regulatory attention. The question is not hypothetical forever, though, because the network is explicitly designed around token-paid compute and a not-yet-launched RITUAL asset, and the framework that will decide how that token gets treated is being written in Washington right now. SEC Chair Paul Atkins has spent 2026 pushing what his agency calls Project Crypto, and in a March 17 speech titled “Regulation Crypto Assets: A Token Safe Harbor,” he laid out a plan built around three pieces: a startup exemption letting early projects raise up to $5 million over four years under lighter disclosure, a fundraising exemption allowing up to $75 million through qualifying crypto investment contracts, and a broader safe harbor for issuers that step back from active managerial control, according to the SEC’s own publication of the remarks. “This is what regulatory agencies are supposed to do: draw clear lines in clear terms,” Atkins said of the effort.

That framework moved from speeches to an actual rulemaking docket this month. The SEC’s own regulatory agenda lists a Notice of Proposed Rulemaking, filed under the name Regulation Crypto, targeting release as soon as July 2026, though the proposal still has to clear a public comment period and a Commission vote before it means anything in practice, a process our explainer on the SEC’s shift from lawsuits to rulemaking covers in more depth. Whether it survives depends partly on Congress: the companion CLARITY Act, which would give the framework statutory backing, has been stuck in the Senate for months over unrelated disputes and had no scheduled floor vote as of early July.

One narrower precedent already applies more directly to a network like Ritual. In September 2025, the SEC issued a no-action letter to DePIN project DoubleZero, telling the company it would not recommend enforcement over the project’s token distribution model, so long as transfers stayed programmatic and tied to work actually performed rather than structured as an investment offering, according to the SEC’s statement on the letter. That is close to the same shape as paying Infernet nodes for compute delivered. Separately, SEC staff guidance issued in mid-2025 already extended similar comfort to plain protocol staking and, later, liquid staking specifically. Restaking, the exact mechanism Ritual leans on through its EigenLayer partnership for Infernet’s economic security, has not been given that same explicit treatment, leaving a real, undecided question hanging over any future RITUAL token design. Whether that token would pass the underlying legal test at all is a separate question worth understanding on its own terms; our explainer on how the SEC applies the Howey test in 2026 walks through the analysis that would apply.

Risks Worth Taking Seriously

The most direct risk sits in the trust assumption Ritual chose. A trusted execution environment proves that a specific chip ran a specific piece of code, not that the chip itself is uncompromised; security researchers have documented side-channel attacks against TEEs from multiple hardware vendors over the years, and Ritual’s entire non-deterministic execution path depends on that hardware attestation holding up. That is a deliberate tradeoff rather than an oversight, since the zero-knowledge alternative remains too slow and expensive to verify large model inference directly, but it does mean Ritual is making a different bet than projects chasing cryptographic, rather than hardware-rooted, verification.

The second risk is economic security itself. Ritual’s restaking partnership only protects the network to the extent that real capital backs it, and EigenCloud’s total restaked value, which industry trackers put at a 2026 peak above $19 billion, has fallen to roughly $4.8 billion as of mid-July, according to CoinGecko. A shrinking restaking base means a shrinking economic penalty available to punish a dishonest Infernet node or model router, the same concentration-and-capital risk that liquid restaking tokens carry more broadly; our piece on how liquid staking tokens like stETH and rETH actually work is useful background on how that layered trust chain gets built in the first place.

A third, more specific risk comes from the exact audience Ritual is building for. An agent-native chain also inherits agent-native attack surface: if a model can be tricked by adversarial input into misreading an instruction, a sufficiently autonomous agent could authorize a transaction a human never would have approved, a risk category the wider industry is still learning to defend against rather than one Ritual invented or can fully engineer away. Finally, there is a plain execution and timing risk. Ritual has been building since 2023, has been in some form of testnet since November 2024, and still has no mainnet date, no live token, and no disclosed tokenomics, while at least one direct competitor has already shipped a mainnet, a flagship application and a trading token within the same year this article was written. Momentum in crypto rarely sits still, and a long testnet phase risks losing developer mindshare to whichever competing network ships first, regardless of whose architecture is more ambitious on paper.

What to Watch Next

A handful of concrete signals will say more about Ritual’s trajectory than any roadmap post. The first is whether Genesis 1000 turns into a genuinely public testnet, open beyond Discord-role holders, in the second half of 2026. The second is whether Ritual publishes a mainnet date at all this year, and whether that date holds. The third is tokenomics: watch for an actual allocation table, a stated total supply, and clarity on how testnet quest participants, node operators and investors get treated relative to one another, since that split is usually where community goodwill is won or lost. The fourth is whether EigenCloud’s restaking capital base stabilizes, since Ritual’s specific security model depends on it more than most other design choices the team could have made. The fifth, and simplest to track, is usage: how many applications outside Ritual’s own ecosystem partners, MyShell and Story Protocol among them, actually route real transactions through Infernet or Ritual Chain rather than merely claiming an integration.

Frequently Asked Questions

What is Ritual in crypto?

Ritual is a crypto project building infrastructure that lets smart contracts request and verify artificial intelligence computations. It launched in two stages: Infernet, a decentralized oracle network for AI already in use, and Ritual Chain, a purpose-built layer 1 blockchain that treats AI inference as a native, verifiable transaction type. It was founded in 2023 by former Polychain Capital investors Niraj Pant and Akilesh Potti.

Does Ritual have a token yet?

No. As of mid-2026, RITUAL only exists as a testnet asset distributed through a faucet to pay for gas while developers test Ritual Chain. There is no live, tradable mainnet token, no confirmed generation event date, and no published tokenomics or allocation table.

How is Ritual different from Bittensor?

Bittensor organizes independent teams into subnets that compete on a specific machine intelligence task and get ranked and paid based on output quality. Ritual instead builds inference directly into a blockchain’s execution layer, using trusted execution environments so a smart contract can call a model and receive a cryptographically attested result within the same transaction flow, rather than routing through a separate competitive marketplace.

How does Ritual verify that an AI model’s output is correct?

Ritual Chain runs non-deterministic workloads like model inference inside trusted execution environments and has validators check a hardware-signed attestation of the result rather than re-running the computation themselves. That proves the claimed hardware and code produced the output, but it relies on trusting the chip vendor’s attestation chain, a different guarantee than the cryptographic proofs used by zero-knowledge-based verification approaches.

When will Ritual Chain launch on mainnet?

Ritual has not announced a mainnet date. The network has been in various testnet phases since November 2024, most recently a Discord-role-gated phase that opened in April 2026 and a Genesis 1000 agent deployment phase that opened in June 2026. Both are still testnet, not mainnet.

Written by the HOGE Wire research desk.