Gensyn’s $AI Hits a New Low as Delphi Grows Up in 2026

On August 14, 2026, Gensyn’s $AI token changed hands at $0.01944, a fresh all-time low and roughly 81% below the price it reached on the day it began trading, according to CoinGecko. It has bounced a few percent since, but the message from the tape is hard to miss: almost four months after mainnet, the market values $AI below every level it has held since launch.

The strange part is what was happening under the hood at the same moment. Delphi, the network’s flagship application, now carries a completed security audit from Trail of Bits, a published software development kit, and support for autonomous AI agents that can buy and sell markets without a human clicking a button. By almost any engineering measure, Gensyn looks more finished than it did in the spring. The token does not agree.

This report looks at both sides of that gap: what Gensyn actually shipped over the summer, why the price keeps grinding lower anyway, and what a trader, a builder, or a curious observer should watch from here. It picks up where our coverage of the network’s first hundred days left off.

A New Low While the Machine Runs

The numbers are stark. $AI set its all-time low of $0.01944 on August 14, 2026, taking out the previous floor from June. The all-time high, $0.1073, was reached on April 29, 2026, the first day of trading, so the token’s entire price history reads as one long slide from its opening print. As of this writing it trades near $0.0207, a few percent above the new floor but still down about 81% from that peak, per CoinGecko.

A market capitalization near $27 million against a fully diluted valuation above $200 million says two things at once. First, the float is tiny: only about 13% of the ten billion token supply circulates. Second, the market is pricing the locked 87% at a heavy discount to today’s price, which is exactly how a low-float, high-FDV token behaves when traders anticipate future supply. That shape is common among the AI-crypto tokens that launched in 2026, and it sits at the center of why $AI trades the way it does.

What Gensyn Is Building

Gensyn was founded in 2020 in London by Ben Fielding (CEO) and Harry Grieve, who met through the Entrepreneur First accelerator. The bet has always been the same: build a trust layer for machine-learning compute so that training and inference can run across untrusted, globally scattered hardware and still be verified as correct. If that works, the argument goes, the world’s idle GPUs become a single programmable resource for AI.

Investors bought the thesis early. Gensyn raised a $43 million Series A led by a16z crypto in June 2023, with CoinFund, Canonical Crypto, Protocol Labs, and Eden Block joining, part of roughly $67 million raised across its rounds, per ICO Drops. In their investment note, a16z crypto general partners Ali Yahya and Guy Wuollet wrote that “Gensyn can potentially 10-100x the available compute power for machine learning” (a16z crypto).

The network is organized in four layers, per Gensyn’s documentation: a Reproducible Execution Environment (REE) that produces bitwise-identical outputs across different hardware, a trustless verification layer built on Verde and Judge, an Agent eXchange Layer for encrypted peer-to-peer communication, and an on-chain coordination layer that handles payments and incentives. The determinism comes from RepOps, a set of reproducible operators that fix the order of arithmetic so the same calculation lands on the same bits regardless of the chip. Verde, described in a February 2025 paper co-authored by NYU cryptographer Joseph Bonneau, is the clever bit: rather than re-running an entire computation to check it, Verde uses refereed delegation to pinpoint the first step where two parties’ results diverge. Fielding has called verification the heart of the company, telling Decrypt that “that’s the big secret sauce behind Gensyn, we’ve solved that problem for machine learning training specifically.”

Gensyn’s mainnet went live on April 22, 2026, as an EVM-compatible OP Stack layer-2 anchored to Ethereum. Like Ritual, another verifiable-AI project, Gensyn concluded that a general-purpose chain would not do and built its own settlement layer for compute and payments. Its first native application, the a16z-backed information-market platform Delphi, launched alongside it (The Block).

Delphi Grows Up: The Trail of Bits Audit

Both of our earlier Gensyn reports flagged the same gap: for a network whose token thesis rests on Delphi’s fees, there was no public security audit of the Delphi contracts. That gap is now closed. In its engineering writeup, Gensyn says the Delphi contracts “were audited by Trail of Bits over months and across multiple review cycles,” with the scope covering the cost function, the market lifecycle, the factory architecture, and the gateway interfaces (Gensyn blog). The company says the full audit reports will be published alongside the technical writeup.

Why does this matter for a token? Because Delphi is the only part of the network that currently generates fees, and those fees feed the buy-and-burn loop that is supposed to support $AI. If the market factory or the settlement path held a flaw, the entire economic argument would rest on shaky ground. An audit by a firm with Trail of Bits’ reputation does not make a contract unbreakable, and plenty of audited protocols have still been exploited, but it removes one of the loudest objections skeptics had after mainnet. It is also the precondition for the next step: letting software, not just people, trade on Delphi.

Dynamic Pari-Mutuel and Three-Tier Settlement

A point worth correcting from earlier reporting: Delphi’s live markets do not run on a logarithmic market-scoring-rule (LMSR) automated market maker or a central order book. According to Gensyn’s own engineering writeup and its developer documentation, Delphi uses a dynamic pari-mutuel (DPM) mechanism, which quotes a continuous price derived from the share supply across outcomes, with implied probabilities updating as wagers accumulate rather than being set by a market maker. A separate competition environment uses LMSR pricing with different collateral tokens, which is likely where the earlier LMSR description came from. The practical appeal of DPM is that it does not need a dedicated liquidity provider to seed each market, which matters for a permissionless venue where anyone can open one.

Settlement is where Gensyn’s verification technology meets a consumer product. Delphi offers three settlement tiers, and the market creator chooses the trade-off between speed and verifiability.

The REE path is the whole pitch in miniature. A prediction market settled by an AI is only as trustworthy as the model doing the settling, and closed frontier models are a black box. Gensyn’s answer is to let creators run settlement inside a reproducible environment so the resolution can be checked byte for byte, the same verification logic Verde applies to training. Creators write settlement prompts from category-specific templates, which keeps resolutions consistent while leaving control in the creator’s hands.

Agents Enter the Market: The Delphi SDK

The more consequential summer development is that Delphi opened its doors to software. As Gensyn puts it, “the contracts do not distinguish between humans and agents. Anyone with a wallet can buy, sell, and redeem directly on-chain, and the same liquidity is available to both” (Gensyn blog). To make that practical, Gensyn shipped an Agentic Trading toolkit built on the Delphi SDK.

The toolkit, documented at docs.gensyn.ai, pairs skill files that describe behavioral patterns with example TypeScript scripts. Under the hood, the SDK handles the plumbing that usually trips up automated trading: RPC calls, contract interaction, token approvals, and transaction signing. Agents can list, search, and browse markets, pull live pricing, place trades with slippage protection, manage positions and history, query historical on-chain data through Goldsky subgraphs, redeem winnings, and manage token allowances. One guardrail stands out: agents cannot create markets, which still must be opened through the Delphi user interface.

This is where Gensyn’s two theses meet in a useful way. A network built to verify AI computation now hosts a marketplace where an AI that reasons about an event can also take a position on it, and the settlement of that event can itself be verified. Once autonomous agents trade on a shared layer-2, the familiar dynamics of on-chain markets follow them, including the ordering and latency games that make up MEV, crypto’s invisible tax. Whether agent flow becomes a meaningful volume driver for Delphi, and therefore for the buyback, is one of the most important open questions for the token.

The Buy-and-Burn Loop and the Demand Side

Delphi’s fee model is the demand engine for $AI. Every trade pays a protocol fee of about 2% of volume. Of that, 1.5 percentage points go to the market creator, typically in stablecoins, and 0.5 percentage points flow to the AI BuyBack Vault (Bitcoin.com). The Vault swaps that stablecoin revenue into $AI, then splits the result: roughly 70% is permanently burned, about 29% goes to the Community Treasury, and around 1% pays the executor. The swaps route through the canonical Uniswap V3 deployment that governance approved for Gensyn’s layer-2 (Uniswap governance).

On paper, the loop is elegant: more Delphi volume means more $AI bought off the market and burned, which tightens supply and creates a structural bid. Agentic trading is meant to be the multiplier that pushes volume high enough to matter. The problem is scale. A 0.5% slice of trading fees is a thin stream against a ten billion token supply and the unlocks still to come, and Gensyn does not publish a live cumulative-burn dashboard on a free tier, so the market cannot easily price the loop’s output. This is the same idea as the revenue buybacks that have become fashionable across DeFi, where a protocol returns value by buying its own token, but a buyback you cannot measure is hard to pay up for, however well designed.

The Supply Overhang and the 2027 Cliff

If one factor explains the grind lower better than any other, it is supply. Only about 13% of $AI circulates today. The allocation, per Tokenomist, is heavily weighted toward insiders and the treasury: Community Treasury 40.4%, Investors 29.6%, Team 25.0%, Community Sale 3.0%, and Testnet Rewards 2.0%. Team and investors together hold roughly 54.6% of the supply.

The granular unlock schedule sits behind a paywall, so precise dates and tranche sizes are not public, but the reported structure is a familiar one: team and investor tokens face a 12-month cliff from the April 2026 token generation event, followed by a linear release over the following two years. In plain terms, the first large wave of insider supply begins to reach the market around April 2027, and it keeps coming after that. A smaller unlock tied to the December 2025 community sale, some of which carried a 12-month lock, arrives sooner, around the end of 2026.

This is what a $27 million market cap sitting under a $207 million fully diluted valuation is telling you. The market is discounting a lot of future supply, and every rally between now and 2027 has to climb a wall of anticipated unlocks. For the buy-and-burn loop to change that math, the burn has to start outpacing new issuance, and at current Delphi volumes it does not. Supply, not the product, is the heaviest weight on the chart, and it is a weight that gets heavier as the cliff approaches.

Why $AI Keeps Sliding: Macro Meets Float

Supply is the structural story; the macro backdrop is the cyclical one. $AI is a high-beta bet on a single narrative, decentralized AI, and that narrative has cooled across the board in 2026. Even the sector’s bellwether, Bittensor, trades well below its old highs, and smaller AI-DePIN tokens have fared worse. When risk appetite contracts, the newest and least liquid tokens fall hardest, and a token with a tiny float and a big overhang is about as sensitive to sentiment as they come.

Rates are part of the same picture. Crypto’s beta to liquidity conditions has been on full display this summer, as our look at the collapsing odds of a September Fed hike and crypto’s counterintuitive reaction laid out. Against that, an engineering roadmap, even a good one, struggles to move price. An audit and an SDK are the kind of substance that matters over years, not the kind of catalyst, a marquee exchange event, a viral Delphi market, a headline partnership, that repriced tokens in this cycle. Shipping has been necessary for Gensyn; so far it has not been sufficient.

$AI vs. the Compute-Token Field

It helps to place $AI among its peers, because they are not selling the same thing. The table below uses direct CoinGecko readings from August 16, 2026.

The distinctions matter. Akash and io.net rent raw compute; Bittensor pays for useful model output ranked by validators; Gensyn sells something narrower and, if it works, more defensible: a way to prove a computation was done correctly on hardware you do not control. That verification layer is the moat. It is also the smallest market cap of the four, which a bull reads as room to run and a bear reads as the market’s verdict on how far verification is from real demand. io.net is the closest tokenomics cousin, since it too funds a buyback from network revenue, so the two are effectively running the same experiment: can a usage-funded buyback hold up a low-float token before the unlocks arrive?

RL Swarm Is Still on the Bench

Here is the irony at the center of Gensyn today. The original pitch was decentralized machine-learning training, embodied by RL Swarm, an open-source framework for reinforcement-learning post-training run across volunteer nodes. RL Swarm drew real testnet participation through 2025, with community trackers counting thousands of nodes at its peak, and it was extended in November of that year by CodeZero, which pushed it toward cooperative coding agents with Solver, Proposer, and Evaluator roles. But there is still no permanently running, fee-paying swarm. The team paused the hosted versions to focus engineering on mainnet and Delphi, and that is where things stand.

Co-founder Harry Grieve framed the company’s reason for existing this way to Decrypt: “we have a very acute machine learning problem that needed a decentralized trust layer.” The trust layer now exists, and it is busy, but it is busy settling prediction markets rather than refereeing large training jobs. Bulls argue Delphi is a wedge: prove verification on a consumer app, generate fees, then bring the flagship training marketplace back online with a working economy underneath it. Bears note that the thesis that attracted a16z, decentralized training at scale, remains unproven in production, and that a markets app, however clever, is not the same business.

Regulation: Information Markets, the CFTC, and the SEC

Gensyn is careful to call Delphi an information market rather than a prediction market, a distinction Fielding has drawn repeatedly and one The Block echoed in describing the platform as AI-settled information markets (The Block). The framing is not just branding. In the United States, event contracts are the subject of an active regulatory fight. The Commodity Futures Trading Commission issued a notice of proposed rulemaking on June 10, 2026, on how it reviews event contracts and defines gaming contracts (CFTC), while the leading US event-contract exchange has spent the year in a thicket of litigation with state regulators.

A permissionless venue where anyone can spin up a market, where AI settles the outcome, and where autonomous agents can trade is precisely the kind of design that draws regulatory attention. Gensyn is a UK company and sits outside direct US registration, but US user access and a US-listed token pull it toward that scrutiny regardless. As for $AI itself, US readers should apply the Securities and Exchange Commission’s 2026 interpretive framing, which sorts digital assets into categories such as digital commodities, tools, and securities. A network-fee token wired to a buyback most plausibly reads as a digital commodity or tool rather than a security, but that is analysis for readers, not a determination the SEC has made about $AI. None of this is legal or investment advice.

Security When Agents Hold the Keys

Opening Delphi to agents shifts the security question. When software trades on your behalf, software holds keys and signs transactions, and the audited contract underneath does nothing to protect a user whose agent approves a malicious spend or whose key leaks. The perimeter moves from the protocol to the wallet and to the permissions you grant the agent.

That makes account design central. This is the case for smart-account wallets, which can enforce spending caps, scoped allowances, and revocable session keys, so an agent has exactly the authority it needs and no more. It also raises the stakes on approvals, because as our guide to crypto phishing in 2026 argued, the signature is the exploit: an agent that blindly signs is a faster, tireless version of a careless human. The practical advice is unglamorous but load-bearing: fund agents with limited allowances, cap what they can spend, keep long-term holdings in separate cold storage, and treat every approval as a decision, whether a person or a model is making it.

Bull Case, Bear Case

The honest synthesis is that Gensyn’s engineering and its token are on different clocks. The build is compounding: a verification stack that now settles real markets, an audit that closes the loudest criticism, and an SDK that could turn AI agents into genuine users. The token is captive to arithmetic: a tiny float, a large overhang, a buyback too small to offset the coming supply, and a sector out of favor. Both can be true at once, and for now they are.

What to Watch Next

• Publication of the full Trail of Bits audit reports and the promised technical writeup, which would let outsiders judge the findings rather than take the summary on faith.
• Delphi volume and fee trends, and specifically whether agentic trading lifts them enough to make the buyback visible in the data.
• Transparency on cumulative burns; a public, free dashboard would let the market actually price the flywheel.
• The approach of the 2027 insider cliff and any pre-emptive selling or hedging as it nears.
• Any relaunch of RL Swarm into a permanent, fee-paying swarm, which would revive the original training thesis.
• New applications beyond Delphi that put the verification layer to work on something other than markets.
• Macro and rate expectations, since $AI trades as a high-beta expression of risk appetite.

Frequently Asked Questions

What is Gensyn’s $AI token and what is it used for?

$AI is the native token of Gensyn, a network that verifies machine-learning computation run on distributed hardware. It is used for payments, staking, security, and governance on the network, and it sits at the center of Delphi’s buy-and-burn loop, in which trading fees are used to buy $AI and permanently remove most of it from supply. As of August 16, 2026, $AI traded near $0.0207 with a market cap around $27 million, per CoinGecko.

Why did Gensyn’s $AI hit a new all-time low in August 2026?

$AI set a fresh all-time low of $0.01944 on August 14, 2026. The main drivers are structural rather than product-related: only about 13% of the ten billion supply circulates, a large block of insider tokens begins unlocking from 2027, and the token is a high-beta bet on the AI-crypto narrative, which has been weak across the sector. Real engineering progress, including a security audit and an agent SDK, has not been enough to offset that supply and macro pressure.

What is Delphi and how does it settle markets?

Delphi is Gensyn’s flagship application, a permissionless information-market platform where anyone can create a market on almost any question. It prices markets with a dynamic pari-mutuel mechanism and settles them with AI, offering three tiers: fast foundation-model settlement, fully reproducible settlement inside Gensyn’s Reproducible Execution Environment, and a hosted version of that environment. The reproducible option lets anyone re-run a settlement and confirm the result byte for byte.

When do Gensyn’s token unlocks happen?

The detailed schedule is not fully public, but the reported structure has team and investor tokens (together about 54.6% of supply) subject to a 12-month cliff from the April 2026 token generation event, followed by a linear release over roughly two years. That means the first large wave of insider supply is expected to begin reaching the market around April 2027, with a smaller community-sale unlock arriving near the end of 2026.

How does Gensyn compare to Bittensor and Akash?

They solve different problems. Akash rents general GPU and CPU cloud through a reverse auction, and Bittensor pays for model output ranked by validators across subnets. Gensyn instead sells verification: a way to prove that a machine-learning computation ran correctly on hardware you do not control. Bittensor is by far the largest of the three by market cap, at roughly $1.88 billion versus Akash near $154 million and Gensyn near $27 million as of August 16, 2026.

By Marcus Okafor, senior markets correspondent, HOGE Wire.