Gensyn

What it does

Gensyn is a compute protocol designed specifically for machine learningMLMachine Learning. The branch of AI where systems learn patterns from data instead of being explicitly programmed with rules. Modern AI (LLMs, image generation, recommendation systems) is almost entirely machine learning.Like teaching a child to recognise dogs by showing them thousands of pictures of dogs, instead of writing down a precise rulebook for what makes a dog. The child learns the pattern from examples rather than from instructions.Read more → trainingTrainingThe one-time process of teaching a neural network to perform a task by showing it massive amounts of example data and adjusting its internal weights until the outputs are good. Training builds the model; inference uses it.Like the years an apprentice spends learning a trade. You don't see any of the actual work, just thousands of repeated mistakes gradually becoming competence. By the end, the apprentice can do the job. The training was invisible, but the skill is now permanent.Read more →. ML training only, with a verification layer to prove the work was done correctly. General compute, rendering and storage are out of scope.

The architecture has four layers. At the bottom, the SAPO algorithm ensures ML computations produce identical results across different hardware, whether that is an NVIDIA A100, an RTX 4090, or an Apple M-series chip. Above that, the Verde verification system uses Reproducible Operators (RepOps) for bitwise-identical ML outputs, with a two-level bisection game for dispute resolution. The peer-to-peer layer handles gradientGradientIn machine learning, the direction a model's parameters need to be adjusted to reduce its prediction error. Training is a long process of computing gradients and nudging the parameters in the direction the gradient suggests.Like a ball rolling down a hill. The slope of the hill at each point tells the ball which direction to move. The gradient is the slope. Training is letting the ball roll downhill many times until it settles into a low point.Read more → sharing without centralised coordinators (NoLoCo for gossip-based gradient averaging, CheckFree for fault-tolerant recovery, SkipPipe for efficient gradient sharing). The coordination layer is a custom Ethereum L2 rollupL2Layer 2. A blockchain that runs on top of an L1 to provide cheaper or faster transactions while inheriting the L1's security. L2s batch many transactions and post compressed proofs back to the L1.Like an express lane built on top of a busy motorway. The express lane handles its own traffic at high speed, but it still feeds back into the main motorway and uses the motorway's bridges and tolls for security.Read more → that handles participant registration, task submission, proof verification, and payment settlement.

Four roles operate the network: Submitters (who create training tasks), Solvers (who provide compute), Verifiers (who check the work), and Whistleblowers (who challenge incorrect verifications). The actual ML training happens off-chain. The proofs, payments, and dispute resolution happen on-chain.

The project was co-founded in May 2020 by Ben Fielding (PhD in AI/ML from Northumbria, research in evolutionary neural architecture search) and Harry Grieve (MA Economics from Aberdeen, Master of Public Affairs from Brown, former Director of Data Research at Cytora). They met through Entrepreneur First’s London cohort. The company, Gensyn Limited, is registered at Companies House (#12601008) in London. Team size is 44 with 20-plus open roles.

Funding totals $66.74 million across four rounds: pre-seed ($1.1M, January 2021), seed ($6.5M, March 2022, led by Eden BlockBlockA batch of transactions added to a blockchain at a set interval. Each block cryptographically links to the previous one, creating an append-only chain that can't be rewritten without redoing all the work since.Like a page in a ledger. Every page has a fixed number of entries, every page references the previous page, and once a page is filled and signed off it can't be edited without visibly invalidating every page that came after. The chain is just a very long series of these sealed pages.Read more →), Series A ($43M, June 2023, led by a16z crypto), and a public token saleICOInitial Coin Offering. A token sale where a project sells tokens directly to the public, usually before any product exists. ICOs dominated 2017-2018 funding and are now mostly replaced by airdrops, IDOs, or fair launches.Like a company selling shares to the public before going public, except with no SEC oversight, no audited financials, and often no product at all. The 2017 ICO boom showed why those guardrails exist in traditional finance.Read more → ($16.14M, December 15-20 2025 via English auction on Sonar, 7,412 participants, clearing at $0.0473/tokenTokenA digital unit of value or access rights tracked on a blockchain. Tokens can represent ownership in a project, a right to use a service, a share of future revenue, or simply a tradable asset with no underlying claim.Like a physical poker chip a casino issues. The chip itself has no value. What makes it worth something is what it lets you do at the casino, what the casino has promised, and how much other people will pay you for it.Read more → and $473M FDVFDVFully Diluted Valuation. The market cap a token would have if every token that will ever exist were already in circulation. FDV is what the project would be worth if all locked, vesting, or unminted tokens were trading today.Like valuing a startup based on what every share would be worth if all the unvested employee options had already been exercised. The number is bigger and uglier than the official market cap, but it tells you the true ceiling.Read more →). Other backers include CoinFund, Protocol Labs, Canonical Crypto, Maven 11 Capital, and Galaxy.

Current status: the Gensyn L2 is live, with Delphi (AI-settled prediction markets) launched on 22 April 2026 as the first product on the network. The $AI token went through TGE on 29 April 2026, hitting an all-time high of $0.1034 the same day before falling roughly 68% in the following three days. Trading is now active on Bitget, KuCoin, Kraken, MEXC, Uniswap V3/V4 and Binance Alpha, with a Binance perpetual added shortly after. Daily volume is around $26 million in early May 2026. The training network itself, the original product the protocol was designed around, is still pre-mainnet. Testnet (Phase 0, launched March 2025) continues training 0.5B-1.5B Qwen variants.

Value proposition

The pitch is simple. ML training is expensive, GPUGPUGraphics Processing Unit. Originally designed to render video game graphics, GPUs turned out to be exceptionally good at the massively parallel math that AI models need. Modern AI training and inference runs almost entirely on GPUs.Like a factory with 10,000 workers doing the same simple task in parallel, versus a CPU which is more like 10 workers each doing different complex tasks. AI training involves doing simple math a million times per second on a million numbers, which is exactly what the GPU factory is designed for.Read more → supply is constrained, and most of the world’s compute sits idle. Gensyn connects underutilised hardware into a unified training network with cryptographic verification that the work was done correctly.

What makes Gensyn stand apart from the dozen other “decentralised GPU” projects is the research depth. Seven papers published on arXiv: Verde (verification), NoLoCo (gradient averaging), CheckFree (fault tolerance), SkipPipe (gradient sharing), SAPO (deterministic execution), and two more on reinforcement learningRLReinforcement Learning. A training paradigm where an AI agent takes actions in an environment, receives rewards or penalties for the outcomes, and learns a policy that maximises long-term reward. Used heavily for aligning modern LLMs.Like training a dog with treats. Good behaviour gets a treat. Bad behaviour gets nothing or a reprimand. Over many repetitions the dog learns which behaviours produce treats and starts doing them on purpose.Read more → swarms and diverse network ensembles. The Verde system, which enables bitwise-reproducible ML across heterogeneous hardware, is a novel contribution to the field. a16z’s investment thesis called them “nobody better” at combining AI and crypto.

The testnet shows meaningful traction. Gensyn reports 165,000-plus users, two million models trained, and 90 million transactions. RL Swarm lets anyone with a GPU or 32GB of RAM run a node and participate in collaborative reinforcement learning training. The barrier to entry is low.

Three products are live or in active use. Delphi is a predictionInferenceRunning a trained AI model to produce an answer. Inference is what happens when you type a prompt into ChatGPT and get a response. The model takes your input, computes a best guess, and returns it.Like asking an expert for their opinion. The training was the decades they spent becoming an expert. The inference is the 30 seconds it takes them to answer your specific question.Read more → market where users stake on which AI models will perform best on benchmarks. CodeAssist trains coding models locally on your own edits. BlockAssist does the same for blockchain queries. All three demonstrate practical applications of the core protocol.

The counter-narrative is equally clear. The training network itself is still on testnet. The models being trained are 0.5B to 1.5B parameter Qwen variants, not frontier-scale workloads. The litepaper projected V100-equivalent pricing at $0.40/hour (80% cheaper than AWS), but no real compute pricing exists yet. Delphi is the first product to ship to the L2, but it is a prediction market, not a training service. Competitors like Akash and io.net have production networks handling real workloads today. Gensyn is still proving that its verification system works at scale.

The ML-specific focus is both a strength and a limitation. General compute networks can serve ML workloads plus everything else. Gensyn is ML-only. If the verification layer and distributed training primitives work at production scale, the specialisation is a moat. If not, it’s a constraint.

Tokenomics

$AI is an ERC-20 token on the Gensyn L2. Total supply: 10 billion. The token sale closed in December 2025, with tokens distributed to buyers in early 2026. 1.31 billion tokens (13.05%) are now in circulation as of June 2026.

Distribution:

• Community Treasury: 40.4% (20% unlocked at TGETGEToken Generation Event. The moment a project's token first becomes tradeable. TGE is when vesting clocks usually start, when liquidity hits exchanges, and when public price discovery begins.Like the IPO day for a startup. Everything that happened before TGE was private valuations and paper agreements. Everything after is the public market deciding what the thing is worth in real time.Read more →, remainder linear over 36 months). Funds ecosystem development, grants, and programmes. Controlled by the Gensyn Foundation.
• Investors: 29.6% (12-month cliffCliffA waiting period at the start of a token vesting schedule during which no tokens unlock at all. After the cliff ends, tokens begin releasing according to the vesting schedule.Like a probationary period at a new job. You don't get your stock options on day one. You wait 12 months to prove you'll stick around, then everything starts unlocking normally.Read more →, then 24-month linear unlock). Pre-seed, seed, and Series A investors. Cannot stake during lockup.
• Team: 25.0% (12-month cliff, then 24-month linear unlock). Cannot stake during lockup.
• Community Sale: 3.0% (immediate unlock for non-US buyers; 12-month lockup for US buyers and those opting for 10% bonus multiplier). 300 million tokens sold via English auction at the clearing price.
• Testnet Rewards: 2.0% (distributed based on testnet participation level and sale bid amount).

The insider allocation is 54.6% (investors plus team). That is high. The 12-month cliff and 24-month linear unlock provide meaningful protection, and locked tokens cannot be staked. But once the cliff expires (approximately April 2027), 24 months of continuous insider selling pressure begins.

The community’s direct access is limited: 3% public sale plus 2% testnet rewards equals 5% of total supply. The remaining 40.4% community treasury sounds generous but is controlled by the Foundation, which is initially team-aligned.

Token utility spans five functions: compute payments, verification stakingStakingLocking up a cryptocurrency to help secure a blockchain network, usually in exchange for rewards. The locked tokens act as a security deposit that can be taken away if the staker misbehaves.Like putting down a large rental deposit for an apartment. You get the money back if you behave, you earn interest while it's locked, and the landlord takes it if you trash the place.Read more →, Delphi prediction market stakes, governance voting, and a programmatic buy-and-burn mechanism from transaction fees. Delphi is the first of these to go live: trading fees on the prediction market are roughly 2%, with 70% of the protocol cut permanently burned and 29% to community treasury. The compute-payment leg of the design still depends on the training-network mainnet, which is not yet live.

The public sale used a transparent English auction on Sonar. The token began trading on 29 April 2026 and now sits on Bitget, KuCoin, Kraken, MEXC, Uniswap V3/V4 and Binance Alpha, with a 20x Binance perpetual added shortly after. Daily volume is around $26 million per CoinGecko in early May 2026. Price action has been brutal: ATH of $0.1034 on TGE day, down roughly 68% within three days.

How to participate

Run an RL Swarm node. The lowest barrier entry point. Install Docker, Python 3.10-plus, and Node.js 22.x. GPU option: NVIDIA RTX 3090, 4090, 5090, A100, or H100 with 12GB-plus VRAM and CUDA 12.4-plus. CPU option: ARM64 or x86 with 32GB-plus RAM (slower but functional). Nodes train Qwen2.5 models (0.5B to 1.5B parametersParametersThe internal numbers (weights and biases) inside a neural network that get adjusted during training. A 70-billion-parameter model has 70 billion adjustable internal numbers encoding everything it has learned.Like the synapses in a human brain. Each parameter is a tiny dial that gets nudged a little during training. With enough dials, the network can represent surprisingly complex patterns. The total parameter count is roughly how much "brain" the model has.Read more →) as part of a distributed reinforcement learning swarm. Trained models are published to your HuggingFace account. Testnet rewards come from the 2% allocation.

Predict on Delphi. The prediction market lets you stake on which AI models will outperform on benchmarks. Markets run in rounds (four completed December 2024 to February 2025, next scheduled March 2026). Stake on expected winners, earn rewards when markets settle.

Build. Contribute to open-source repos (rl-swarm, codeassist, blockassist, rl-swarm-contracts). All MIT licensed. Deploy custom swarms. Publish models to HuggingFace. The documentation is solid, covering protocol design, testnet setup, and product guides. Seven arXiv papers provide deep technical context.

Trade or hold AI. The token began trading on 29 April 2026 across Bitget, KuCoin, Kraken, MEXC, Uniswap V3/V4 and Binance Alpha. A Binance perpetual with 20x leverage was added shortly after. Daily volume is around $26 million per CoinGecko in early May 2026. Note the post-TGE volatility: AI hit an all-time high of $0.1034 on launch day before falling roughly 68% in three days. Compute-fee revenue and the buy-and-burn at scale still depend on the training-network mainnet that has not yet shipped.

Honest assessment

What works

The research is real. Seven arXiv papers from a team that includes academic researchers and cryptographers is exceptional for a crypto project. Verde verification (bitwise-reproducible ML across heterogeneous hardware) is a substantive technical contribution that no competitor has replicated. The founding team is credible: Fielding’s PhD work directly feeds into the protocol design, and Grieve brings quantitative finance expertise. Both are doxxed, verified through Companies House, and have been building since May 2020.

The testnet engagement is solid. 165,000-plus users and two million models trained (per Gensyn’s reported metrics) suggests real grassroots interest, even accounting for airdropAirdropDistributing tokens for free to eligible wallets, usually to reward early users, bootstrap a community, or decentralise token ownership away from a small group of insiders at launch.Like a supermarket handing out free samples to people who already shop there. The samples cost the supermarket nothing to print. The goal is to convert casual shoppers into loyal customers by giving them something tangible to talk about.Read more → farming. The RL Swarm framework lets you actually run distributed reinforcement learning on consumer hardware today. That’s tangible, not vapourware.

Open-source credentials are strong. MIT licence across core repos. 1,699 stars and 639 forks on rl-swarm. Six models published on HuggingFace. The research-to-product pipeline is visible: academic papers become protocol components become testnet features.

What does not work

The training-network mainnet, the project’s whole reason for existing, is still not live. No security audits despite $66.74 million raised over nearly six years. Delphi is shipped and the token is trading, but a prediction market is not the compute product investors paid for. The first three days of trading saw a 68% drawdown from the TGE-day peak, which is normal for tokens with thin float, but it makes price discovery noisy. At some point, “still on testnet for the actual product” stops being patience and starts being a red flag.

The token distribution is insider-heavy. 54.6% to team and investors is above average for the space. The 40.4% community treasury is controlled by the Foundation, whose council is “initially mapped to core team members.” In practice, the team controls the vast majority of tokens at launch.

Governance doesn’t exist. No DAO, no on-chain voting, no governance forum. The Foundation describes “progressive decentralisation” as a plan, not a feature. Two directors (Fielding and Grieve) make all decisions. For a project raising from the public via token sales, the absence of any community governance mechanism is a notable gap.

Testnet trains small models. Qwen 0.5B to 1.5B parameters is useful for reinforcement learning research but nowhere near the frontier-scale training (70B-plus parameters) that would justify the “network for machine intelligence” branding. Scaling from testnet to production is an unproven leap.

The risk

Execution risk dominates. Transitioning from a testnet training small RL models to a mainnet handling real economic value and production-grade ML workloads is a massive engineering challenge. The Verde verification system adds computational overhead to training; at production scale, this overhead needs to be manageable.

Competition isn’t waiting. Akash has had a production compute marketplace since 2020. io.net is delivering millions of compute hours. Even within the ML-specific niche, Bittensor subnets compete for the same workloads. Gensyn’s verification moat only matters if they ship mainnet before the market decides the problem is solved.

Regulatory risk is real. A UK-registered company conducting a public token sale with US buyer restrictions faces securities classification risk in multiple jurisdictions. The 12-month US buyer lockup suggests the team is aware of this risk.

The public sale FDV prices perfection. Investors paid for a fully functioning mainnet, real adoption, and sustained revenue. None of these exist yet. If mainnet slips or adoption is slow, the valuation repricing will be painful.

No security audits have been published for the smart contracts or protocol, despite $66.74 million in funding and a testnet processing 90 million transactions.

My position

I don’t hold $AI tokens. Trading is now live across multiple exchanges, so taking exposure is straightforward, but I want to see what the training-network mainnet actually delivers before forming a price view. The research depth and technical differentiation are real, and Delphi shipping ahead of compute is at least a working product on the L2. For now, this is a credible research bet with significant execution, competition, and post-TGE volatility risk. Running an RL Swarm node remains the most interesting hands-on way to engage; the token offers a position that prices in a future mainnet I haven’t yet seen.

Freedom Score: 52/100

Gensyn scores 52/100 (D grade). Full methodology at Freedom Score Methodology.

Infrastructure decentralisation (10/20): The testnet is permissionless. Anyone with qualifying hardware (GPU or 32GB-plus RAM CPU) can run an RL Swarm node. The network supports heterogeneous hardware from gaming GPUs to data centre A100/H100s. The Verde verification system enables trustless work verification without centralised intermediaries. However, the coordination layer runs on a custom Ethereum rollup whose sequencer and operator details are not disclosed. Testnet infrastructure relies on Alchemy for authentication. The Gensyn L2 is now live with the Delphi prediction market and a tradeable token, but the training-network mainnet itself is not, so live economic security is partial.

Governance decentralisation (5/20): Governance is fully centralised. Gensyn Limited (Companies House #12601008) controls all protocol decisions. Two directors (Fielding and Grieve) are the sole decision-makers. Post-TGE plans include an elected council “initially mapped to core team members” with on-chain proposals. This is a stated intention, not a delivered feature. No DAO, no community voting, no governance forum.

Token distribution fairness (6/15): Insider allocation is 54.6% (29.6% investors plus 25% team). VestingVestingA schedule that locks up tokens allocated to insiders, investors, and team members, releasing them gradually over months or years. Vesting prevents insiders from dumping on public buyers immediately after launch.Like a new employee's stock options at a startup. You don't get all the shares on day one. They unlock over four years so you stick around and do the work rather than cashing out and leaving.Read more → mitigates concentration risk: 12-month cliff plus 24-month linear unlock, and locked tokens cannot be staked. The community’s direct access is 5% (3% public sale plus 2% testnet rewards). The 40.4% community treasury is controlled by the Foundation, which is initially team-aligned. The public sale used a transparent English auction. Not a fair launchFair LaunchA token launch where everyone has the same access from day one. No private sale, no insider allocation, no VC discount. Tokens are distributed by mining, staking, or open public sale at a single price.Like a 100m sprint where everyone starts behind the same line at the same time. Some runners are faster, but nobody gets to start 10 metres ahead because they paid extra. The race is decided by the run, not by who bought the best position.Read more →.

Censorship resistance (8/15): The protocol design is permissionless. Any device can contribute compute. Verde verification removes the need to trust individual providers. The Ethereum rollup inherits some censorship resistance from L1L1Layer 1. A base blockchain that runs its own consensus mechanism, executes transactions, and settles its own state. Bitcoin, Ethereum, NEAR, and Solana are all L1s. Anything built on top of an L1 is technically a Layer 2 or higher.Like the foundation of a building. Nothing else can exist on top until the foundation is solid. Different L1s make different tradeoffs for what kind of building they can support.Read more → settlement. However, the rollup sequencer is undocumented, creating a potential censorship vector. The team could theoretically upgrade or pause contracts. No censorship incidents on testnet.

Data sovereignty (10/15): The architecture provides meaningful user control. CodeAssist trains models locally on user machines. RL Swarm nodes train local models with results published to user-owned HuggingFace accounts. The litepaper describes optional functional encryption for training on encrypted data. The peer-to-peer architecture avoids centralised data aggregation. Only proofs are submitted on-chain, not raw training data or modelModelA trained neural network that takes inputs (text, images, audio) and produces outputs (more text, classifications, generated content). In DeAI the model is the thing that actually does the work.Like a very experienced apprentice who has spent years watching thousands of masters make furniture. They can't explain how they know when a joint is right, but they can make a chair that looks and functions like a Chippendale. The training is invisible. The output is what matters.Read more → weights.

Open source and transparency (13/15): Core repositories are MIT licensed: rl-swarm (1,699 stars), codeassist (699 stars), rl-swarm-contracts, blockassist. Seven academic papers on arXiv with full methodology. Six models published on HuggingFace. Smart contractSmart ContractA program stored on a blockchain that runs automatically when its conditions are met. Smart contracts are how blockchains do anything beyond just transferring tokens — DeFi, NFTs, DAOs, and DeAI infrastructure all run on smart contracts.Like a vending machine. You put in the right input and it produces the expected output, no human operator required. The rules are fixed in the machine itself, anyone can use it, and nobody can stop a transaction in the middle.Read more → code is publicly auditable. The research-to-product pipeline is transparent. Deductions: no published security audits, rollup infrastructure code is not visibly open-sourced, and the litepaper is explicitly outdated with no full replacement published.

Path to improvement

Three changes would materially increase Gensyn’s score:

1. Launch mainnet with transparent metrics. The testnet has been running for a year. Mainnet launch with a public, real-time dashboard showing active nodes, compute hours delivered, revenue, and verification metrics would be the single biggest credibility signal. Self-reported testnet figures from press releases are not enough.
2. Implement governance now that TGE has happened. Token holders who spent real money on $AI tokens should have governance rights. An elected council “initially mapped to core team members” isn’t decentralisation. Deploying on-chain governanceDAODecentralised Autonomous Organisation. A way to coordinate decisions and manage a treasury using token-weighted voting instead of a traditional company structure. Token holders propose and vote on changes directly.Like a shareholder-run company where every shareholder can vote on every decision, the votes are public, and the company can't do anything the shareholders don't approve. The coordination is messier than a normal company but nobody has unilateral control.Read more → with community-elected council members, proposal submission, and transparent voting would demonstrate actual commitment to decentralisation. The token sale closed in December 2025; this is overdue.
3. Publish security audits. $66.74 million in funding, 90 million testnet transactions, and a public token sale, yet no published audit from a recognised firm. Before mainnet handles real economic value, smart contract and protocol audits are essential. This is a straightforward gap to close with available resources.

Returns Score: 34/100

AI scores 34/100 (F grade). Full methodology at Returns Score Methodology.

Token utility (10/20): The $AI token has been distributed to sale participants but none of the intended utility is live. Planned use cases span compute payments, verification staking, Delphi prediction market stakes, governance voting, and a programmatic buy-and-burn mechanism. These are reasonable utility vectors on paper, but none of them can be evaluated in practice because the mainnet doesn’t exist yet. Scoring generously for design intent, but design intent is not utility.

Value accrual (6/20): The buy-and-burn mechanism from transaction fees is the primary value accrual path, and it is a sensible design. Protocol revenue destroys token supply, creating deflationary pressure proportional to network usage. Delphi is now generating prediction-market fees with 70% of the protocol cut burned, but on-chain burn data has not yet been independently verified and the training-compute fees the model was built around still depend on a mainnet that has not shipped. Held at 6 pending Tier 1 verification of Delphi burn flows.

Supply dynamics (6/20): Ten billion tokens with 54.6% allocated to insiders (29.6% investors plus 25% team). The 12-month cliff and 24-month linear unlock provide meaningful short-term protection, and locked tokens cannot be staked, which prevents insiders from compounding their position during the lockup. But once the cliff expires around January 2027, two years of continuous insider selling pressure begins. The community’s direct access is just 5% of total supply. The 40.4% community treasury sounds generous until you realise the Foundation, which is initially team-aligned, controls it.

Revenue sustainability (2/25): Compute revenue is still zero — the training-network mainnet has not shipped. Delphi (live since 22 April 2026) charges roughly 2% trading fees and is the first product on the Gensyn L2 generating any real protocol income, but the figures are nominal and unverified on-chain. The litepaper projected V100-equivalent pricing at $0.40 per hour (80% cheaper than AWS), which would be compelling if it existed. Held at 2 until either the training-network mainnet ships with paying users or Delphi fees become independently verifiable at scale.

Liquidity and access (10/15): TGE happened on 29 April 2026. The token now trades on Bitget, KuCoin, Kraken, MEXC, Uniswap V3/V4 and Binance Alpha, with a Binance perpetual offering 20x leverage added shortly after. Daily volume is around $26 million per CoinGecko in early May 2026, comfortably enough for retail-scale entry and exit. Coinbase and Binance spot are still missing; the listing surface is strong but not yet tier-one complete. Price action has been ugly (ATH $0.1034 on TGE day, down ~68% within three days), which doesn’t change access scoring but signals thin float.

Path to improvement

Three changes would materially increase Gensyn’s returns score:

1. Launch mainnet and generate real revenue. The entire returns case hinges on the buy-and-burn mechanism working with actual transaction fees. Until compute workloads generate revenue on a live network, every tokenomics projection is theoretical. Revenue is the single metric that would transform this score.
2. Reduce the insider allocation or increase community access. 54.6% insider allocation with only 5% direct community access is poor by any standard. Increasing the testnet rewards allocation, adding a broader public sale tranche, or committing to community-directed treasury governance would improve the supply dynamics score materially.
3. Demonstrate pricing competitiveness at scale. The $0.40/hour V100-equivalent claim needs real-world validation. Publishing transparent pricing benchmarks against AWS, Lambda Labs, and competing decentralised compute networks once mainnet launches would establish whether the economic model actually delivers value to compute buyers.

Score change log