Checking recent verified miners…
How work is prioritised
Research before game work.
Compatible updated miners receive exclusive medicine-priority assignments. Public chemistry reference batches compare 65,280 ordered descriptor pairs. The next prepared cross-group campaign covers 7,864,320 unique comparisons in 36 one-time batches, awaiting owner activation; the game fallback searches 65,536 building designs. Updated GPU clients at high effort can search 1,048,576 designs per batch when larger GPU jobs are enabled. Every accepted result is recomputed in full by the server. Reference batches are recorded; only eligible work that passes its configured verifier contributes toward an ordinary reward block.
View chemistry inputs and research progress ↗Reviewed scientific inputs are required for efficacy models. Game work accepts only passing, scope-bound simulation evidence. Failed and inconclusive science stays in research history and learning.
Choose your farm app
Desktop runs installed AI, Blender and Unreal tools. Mobile and browser run supported foreground calculations with the same server checks. Faster devices process larger chunks; smaller devices split the same assignment. Chunk size does not lower the quality target or multiply reward weight.
Choose Maximum when you want all permitted effort. Available tasks, memory, the operating system and heat limits determine actual use. Stop releases workers; no work starts on opening a page.
A useful start.
One simple switch.
Connect to Etheraio’s existing research coordinator. Press Start to contribute verified jobs while this app is open, or Stop whenever you choose.
Sign in to start ↗Medicine research first when reviewed data is eligible, then other science, then world building. Accepted work adds verified contribution weight. Cash allocations depend on eligible settled net network income.

WHAT YOUR DEVICE CONTRIBUTES
One mine. Three kinds of useful work.
Science & medicine research
Calculate assigned public references and run supported model checks. Results describe their tested limits; a simulation pass does not establish a safe or effective medicine.
Game creation & checks
Compatible desktop tools build concept art, meshes, textures and audio, then run the supported geometry and engine checks. Prototypes stay pending until their required checks pass.
AI learning
Opted-in training learns from fresh checked results and retains failures in context. Candidates are tested separately before replacing a working model. Missing tools remain unavailable.
Verified jobs add contribution weight. Only a new, eligible, accepted contribution adds weight. Duplicate, failed and unsupported work adds zero. The reward and remaining block weight below come from your account’s server policy.
Browser mining runs its supported calculations while this page stays visible. Native AI, Blender and Unreal work require the Windows app. Nothing starts until you choose Start.
AI RATE & REAL RUNNING COST
AI compute and profitability calculator
Etheraio Compute Units per second (ACU/s) is a transparent speed label: 1 ACU = 100 output tokens from the same selected model. It is not a model-quality comparison, reward weight or cash value. The browser app cannot report local LLM speed; enter the last measured rate from the desktop AI workspace.
No cash income is assumed. Unfunded contribution weight and unspent AI Credits are not settled cash. A positive estimate is not a promise of earnings.
Make this device a node
Join Etheraio with a Windows PC, Mac, Linux computer, Android phone, iPhone, tablet or another compatible browser device. Your signing identity stays on this device and your verified contribution weight stay in your account.
Checking your device Node…
Node ID: · Verified blocks on this device: 0
Registration leaves mining off. Choose a preset and press Start when ready. Phones and browser apps contribute while visible and unlocked. Auto assigns the largest checked chunks that fit available memory and each user’s effort limit. AMD ROCm, NVIDIA CUDA, other GPU APIs, CPU and AI chips are used only where their matching runtime and task pass device checks; otherwise the configured CPU fallback or waiting state is used. NPU detection alone is not compatibility proof.
Website node is selected. Your device connects automatically when you press Start.
Use a compatible node operated by someone you trust. Continue on its own mining page; your login and signing key stay here.
Open custom server ↗One choice includes medicine-priority verified network work, original browser game assets and local learning from reviewed chemistry outcomes. Compatible GPU work uses your chosen resource preset. Your research-sharing choices stay separate.
Loading saved task selection…
The Windows app additionally runs the eleven installed medicine models, conformation checks, installed local AI, assay training and Blender/Unreal asset checks. Browsers cannot run those native tools. Private chemistry remains encrypted and owner-only; unsupported medical endpoints stay unavailable.
Etheraio research mining
Checking sign-in…
Browsing: very light, GPU off. Gaming: CPU policy ≤8%, GPU off. No automatic session time limit; continuous verified jobs; lower CPU and GPU levels leave room between jobs. Hiding or closing the app stops contribution. Resource percentages are requested budgets, not measured hardware quotas.
Phone & power
Checking available power telemetry…
Android and iPhone can use all presets. Maximum requests back-to-back work on both power sources when you choose Charging and battery. Adaptive reduces requested effort off charge. Keep the app visible and the screen unlocked; hiding or locking stops this session. Phone browsers do not expose reliable temperature or full system controls. If charging status is unavailable, choose it manually.
Entire network
Total blocks mined: Connecting…
Current computational units / second: Connecting…
Reported network AI tokens / second: Waiting for measurements
Recent completed AI output, separately reported from verified work and contribution weight.
Verified candidates and data samples accepted in the last 60 seconds. Updates every 10 seconds.
contribution weight measure accepted results. They are not GPU utilisation, FLOPS or tokens.
Choose your GPU workload
GPU is off. CPU contribution remains available.
This changes how often available GPU jobs run and the preferred adapter. It is not a measured GPU-use percentage. Maximum means the most work this browser app allows; it cannot promise full GPU utilisation or more contribution weight.
GPU is OFF. Compatibility is checked only when you opt in.
Vendor-neutral WebGPU: compatible AMD, NVIDIA, Intel, Apple, ARM and Qualcomm implementations may work. Device, driver and browser support is required, including for future hardware. Fixed integer kernels only; no remote code.
Your session. Your choice.
No time limit. Mining continues while this page is visible until you press Stop.
Device performance
Browsing mode · light contribution.
0 verified jobs / minute · accepted contribution weight is shown below
Optimal compares fixed CPU and GPU kernels, then chooses the faster supported path. It does not start mining.
No verified work completed in this page session.
Optimal performance measures the CPU and available GPU paths, keeps resources ready between jobs and uses the faster verified path. Maximum selects the fastest permitted job frequency and sizes its CPU worker budget for the threads and memory this browser reports. GPU remains optional and requires your explicit consent. Only available, server-verified work earns contribution weight. GPU effort can be set up to 100% effort. Actual utilisation depends on available verified work, device support, drivers, cooling and network latency. High GPU effort requests larger game searches when enabled. Full CPU and server verification still apply, so network waits and verification can leave GPU capacity idle. Keep this app visible; Stop is always available.
My contributions ↗ · Contribution model
Start authorizes this foreground session and creates a device signing key if needed. Read participation terms and privacy. Stop prevents new submissions; a request already received by the server may complete.
Resource controls & optional GPU
Your verified results.
Result reports appear automatically after verification. Calculations without a scientific acceptance criterion stay unassessed.
Reports, reuse and game quality gates ↗Research. Build. Learn.
Medicine checks run inside the mine. Public reference calculations stay in the network queue. Compatible Windows Nodes can also run fixed RDKit / VEGA public reference checks using the shared Start button without an owner vault key. The browser does not execute Java models. Unpublished chemistry stays encrypted and owner-only; local model reference reports earn zero contribution weight pending network acceptance.
All supported browser work is selected for a new session. The Start mining button runs verified network work and the selected browser jobs together. Medicine research stays first. Desktop miners use the same Start for local AI and Blender asset rendering.
See the queue and actual Unreal build evidence ↗
Optional task filters
Select network work, local assets, or both. Browser learning needs verified chemistry outcomes; installed local language models run in the desktop app. Local-only assets use original seeds and earn zero contribution weight until their required automatic checks pass.
Choose your work, then press Start mining.
Chemistry training waits for reviewed positive or negative results.
Blocks from every sector
Verified network blocks earn live contribution weight. Generated asset and local model blocks remain pending independent verification with 0 contribution weight. The browser keeps up to four prototype records for review. Signed-in network work uploads each supported glTF output to your private cloud results and verifies it before releasing the temporary browser copy. Local-only work retains its output on this device. These terrain prototypes need renderer and game review before use.
Open stored outputs ↗ · Owner asset review ↗. Browser work is limited to its supported CPU/WebGPU features and runs while this app is visible. Training updates scoped chemistry search priorities only after a reviewed benchmark outcome. Negative evidence stays tied to its subject, model and conditions; another use needs separate checks. Language-model weights remain unchanged. Installed Ollama and Blender work runs inside the Windows app, with no need to stop mining. Hardware usage depends on useful work and the device.
Get the full desktop mine ↗Connection and node providers
Choose a node. Press Start.
The website node is selected. No node setup is needed.
Add my hosted node
Use Make this device a node to create your device identity. If you already host a compatible Etheraio server, list its HTTPS address here so others can select it. This form registers a provider address; it does not upload or run executable server code, deploy a server, or turn your browser into an independent validator.
Custom providers run their own accounts and reward policies. Changing providers opens their mining page, with no signing keys or website credentials forwarded. Listings are owner-submitted; independently verify an operator before using it.
Download Etheraio AI mining
A native Windows app with one Start/Stop for verified mining, opted-in local AI, task-model training and original Blender game assets, direct CPU and DirectX 12 GPU processing, GPU only and All compatible GPUs modes, thread limits, and Browsing, Gaming, Balanced, Turbo, Optimal and Maximum presets. Choose rules that reduce mining effort when another application is active. Sign in for verified jobs and your saved contribution record. No time limit unless you set one. Internet access is required.
No verified Windows installer is published for version 1.8.85. Check the portable app option below.
Download Windows installer ↗Checking the portable Windows app package…
Download portable Windows app (ZIP) ↗Maximum allows all available CPU threads and native GPU processing across compatible adapters; GPU only keeps processing on GPUs, while Optimal can select CPU. Check GPUs tests compatibility without rewards. Each new job follows the last verified result, within your selected effort and hosting capacity. High GPU effort can request 16 times larger verified game searches and receive the next job with verification when enabled. Medicine remains first. Network waits and full verification can still leave hardware idle. GPU effort is a batching and pacing control, not a hardware power cap. Mining starts OFF; closing the app stops work. Install new desktop releases from this page. Other devices can install the browser app below.
Install Etheraio AI mining on your phone.
One-button Start / Stop, Battery saver, Browsing, Gaming, Balanced, Turbo, Optimal, Maximum and Custom presets. Choose charging, battery or adaptive operation. Only server-verified work credits your signed-in account.
Android
Open this page in a supported browser, then install Etheraio as an app.
iPhone & iPad
Open in Safari, then Share → Add to Home Screen. The same app opens from your Home Screen.
This is an installable web app. Internet and a visible, unlocked app are required for mining. These buttons install the web app; native-package availability is listed on Downloads.
Optional file hosting
Make your PC a node and server.
Open Etheraio AI mining on Windows and choose Make this PC a node and server. Sign in once; the app registers and selects your Node, then mirrors approved website files through Cloudflare.
Choose a cache limit from 1 to 50 GB. Files use space as needed, with a 2 GB free-space reserve. Leave the app open to contribute; stop the file server at any time. Cloudflare keeps Etheraio available while your PC is offline.
The first catalogue contains public website images. Private research, accounts, payments and application code remain on the protected host. File serving does not create block rewards. Checking file-server availability…
Get the Windows app ↗Science workloads and evidence limits
Useful work, with evidence.
Consenting Windows Nodes can run installed local AI models for 20 bounded reference families across mathematics, mechanics, spacecraft, materials, safe chemistry, software, game budgets, circuits, optics, thermodynamics, heat transfer, aerodynamics, astronomy, robotics, acoustics, ocean and Earth science, energy, ecology and statistics. The server recalculates each answer. Passing network references contribute work weight when their reward policy is enabled. Local-only experiments remain separate from full medicine and production asset acceptance.
Checking verified reference results…
In the desktop app, choose Large jobs for this PC, enable local AI or game assets, save controls, then use the shared Start button. Science batches contain up to 32 distinct references within your context budget. Cinematic assets use 8K masters, 8K bakes and up to 2,048 adaptive render samples. Browser GPU kernels perform their supported calculations; native AI and Unreal workloads need the installed desktop tools. No Bitcoin hash guessing is used for these jobs.
Get the Windows miner ↗See workload and acceptance details ↗Paired-device measurements
Your live device measurements
Enable measurement sharing in Etheraio AI mining to view this PC here. Computing and local AI remain controlled in the desktop app.
Windows counters measure whole-PC activity, including other applications. Readings are device-reported, not verified computing or reward evidence. Temperature, power and unsupported measurements show Unavailable.
Live research & world progress.
Loading live progress…
Results are reproduced by the hosted coordinator. The work is bounded descriptive statistics and procedural design search, with no medicine efficacy claim.
Explore Etheraio ↗ · Open shared research memory ↗Nodes & shared pools
·
The default is solo contribution to the existing website network. A pool groups signed contributions and takes only its disclosed test-unit share. contribution weight stay separate.
0 pools loaded
Become a Node or create my pool
OPTIONAL NATIVE-COIN COMPUTE
AI work and crypto stay separate.
The AI farm processes assigned research and game tasks. AI output is measured as tokens per second; crypto workers report hashes per second. Neither measure is interchangeable, and crypto work does not add AI contribution weight.
CPU · Monero (XMR)
The Windows app includes a separate P2Pool CPU route. It starts only when the app reports that the saved XMR wallet, node sync, worker checks, payout route, user consent and spare-capacity setting are ready. A missing check keeps it stopped.
GPU · Pearl, Quantus and other candidates
No GPU coin is currently marked ready by this page. Kryptex KRig 1.5.12 upstream lists Pearl for AMD/NVIDIA and Quantus for AMD/NVIDIA/Intel Arc. The official Quantus Miner 4.2.1 release supports CPU and GPU execution, but is not packaged or enabled by Etheraio. Pearl’s Windows security review remains pending. Quantus requires a matching mainnet node/miner pair, user-created wormhole reward setup and pinned miner authentication; its consensus sends 100% of block rewards to the configured address, so Etheraio’s share needs a separate settlement design (setup guide, reward rules). Ravencoin remains a legacy route, and other catalogue entries are unverified candidates. Each route needs an integrated, security-cleared worker, device test, current pool, native same-coin payout path and fee settlement before Start can use it.
Capacity & rewards
AI work takes priority when selected. Crypto can use only separately opted-in spare capacity within the user’s CPU/GPU limits. Auto can run CPU and GPU coin lanes together only when each has a fresh device-matched net estimate after power and fees; missing quotes or a negative estimate leaves that lane idle. Crypto payouts stay in the same coin and network as mined; verified AI work contributes separately to the cash revenue-share records.
This is a setup guide, not a live readiness report or profitability ranking. Check the app’s live status before starting. Network difficulty, pool availability, electricity and fees affect earnings; no payout is guaranteed.
SUSTAINABLE OPERATION
Use capacity carefully; account for real costs.
With one farm opt-in, Start gives selected AI, game and science work priority. CPU and GPU crypto lanes are considered separately, may run different checked coins together, and need fresh device-matched quotes. If the user has allowed possible losses and no option is net-positive, the app selects the least-negative checked quote; it never picks an unpriced fallback. Missing worker, wallet, payout route or current quote means that lane stays idle.
Native payouts are backed by verified blocks and remain in the coin mined. Contribution records and purchased AI Credits stay separate from mined crypto earnings. This avoids issuing crypto without receipts, but cannot guarantee the service’s financial sustainability: settled income, hosting costs, user energy costs, refunds, taxes and reward liabilities must be measured over time.
Accepted work adds contribution weight.
Reading the live block-weight policy…
Unique accepted work adds verified weight to your contribution record. Funded revenue rounds use your share of all unallocated verified weight on the network. Failed, duplicate and unsupported results add no weight.
Weights are server accounting rules for each supported task, rather than file size or a claim of equal scientific value. Accepted game components share one asset’s contribution budget, with no duplicate contribution for the same work; bounded science reference answers earn their configured verified weight. Only accepted retained world layouts can earn their configured share when enabled; search attempts alone add no weight. Production assets must pass all configured acceptance checks.
Work weights and acceptance checks ↗ · Accepted weight stays saved until it is included in a funded revenue round.
Checking shared network queue…
One shared mine assigns bounded jobs to consenting devices. The desktop app detects CPU, selected GPU and current memory, sizes active segments, and supports compatible GPU/RAM layer offload or CPU-only execution. Smaller segments preserve the task checks and weight. Full model weights and active stages must fit admitted memory; devices do not yet share VRAM for a single model. Supported asset tiles retain the requested 8K source detail while production acceptance remains separate.
Reading automatic AI training rewards…
Compatible Windows miners can include public Qwen adapter training in the same Start. The uploaded weights must pass an authorized automatic evaluator using the pinned base model, finite-weight checks, held-out generated answers and improved held-out losses. Each first accepted training input contributes 1,024 units; held-out inputs and repeats contribute zero. Private and local-only model work earns zero. This is research-model acceptance, not clinical or production-AI approval.
Specialist models and training requirements
Specialist models in one mine
Loading the model catalogue…
Scientific progress and methods
Automatic science and game pipeline.
Medicine methods come first. Compatible desktop Nodes run fixed public conformer and explicit-solvent protein controls, then a separate account reproduces the calculation. These short controls establish numerical reliability; they do not discover medicines or earn contribution weight.
Checking the scientific queue…
Reading automatic calculation, follow-up and game-reuse progress…
Start runs the supported methods under your existing work and sharing choices. Reproduced public results can enter the game queue automatically when you allow result reuse. Private chemistry stays in the owner vault.
Game physics and scientific properties need tests for their intended gameplay use. A failed medicine screen can supply visual variety; it does not change forces, vehicle handling, flight or material properties. Simulation agreement applies to its stated assumptions and is not proof of a new real-world physical law.
A larger conformer search or half-timestep check may follow once. Unsupported disease-specific protocols, docking, free energies and clinical assessment remain unavailable. Solver errors and invalid inputs are retained as diagnostics, not negative scientific conclusions. Existing rewards and private chemistry are preserved.
EXECUTED CHECKS
Automatic component verification
Reading automatic verification service…
Consented fresh positive and negative component checks feed continuous learning while the shared mine runs. Context, failed checks and model limits stay attached. New training candidates remain separate from working models; retraining alone is not an improvement or a reward.
The protected operator replays the supported original basalt mesh, UV and native PBR recipes from checked archive files. Passing new components can contribute to the same full-block weight ledger when activated. Duplicate or failed work earns no additional weight. Final game production acceptance and clinical medicine approval remain separate evidence gates.
Public scientific controls contribute weight only through an enabled, reproduced-method reward profile; unconfigured controls earn zero. Owner-only reports show supported medical model predictions with their exact scope.
Temporary workspace: desktops need a 50 GB allowance; other devices need 10 GB. Mining and verified public-file caches share the same allowance, not two separate allocations. Space grows only as used; managed job copies release after verified upload. Browser quota estimates may exceed actual device availability. Working models, account keys and private records are preserved.
NATIVE CRYPTO POOLS
Choose a checked route in the desktop app.
AI contribution earnings and mined cryptocurrency stay separate. Crypto starts only after your app consent, wallet, runtime, resource and network checks pass.
Monero · public P2Pool nano
The Windows app can use public P2Pool nano for Monero CPU mining. It needs a synced local Monero node and a primary XMR address. Eligible shares can pay native XMR directly to that wallet. The selection screen shows the actual 0% Etheraio fee for this public route. Real shares and blocks are required; a payout interval is not guaranteed.
P2Pool documentation ↗Pearl · worker still unavailable
A pool connection alone does not establish a safe, working GPU miner. Windows Security blocked the alternative worker during preparation. Pearl is not enabled in this release.
Pool information ↗Etheraio pools · later update
Our own fee pools need independent peer, accepted-share, native-payout and fee-split verification before release. The experimental local fee network is separate from public P2Pool. Sign-in alone does not complete verification.
Check the app for live node, pool and hashrate status. Your wallet shows confirmed funds. Download the current app.
YOUR MINED COINS. YOUR WALLET.
Choose a wallet for the correct network.
Create and back up the wallet privately in its own app, then paste only its public receiving address into Etheraio’s desktop crypto setup. A Web3 address for another network cannot receive native XMR or PRL. Etheraio does not create or hold your wallet keys.
Monero · Cake Wallet
Open-source, self-custody Monero support on phone and desktop. Public P2Pool needs a primary XMR receiving address. Use the official install page to choose your device.
Get Cake Wallet ↗Monero · official desktop wallet
The Monero community’s GUI wallet is another option for XMR. Its download page includes verification instructions and other wallet choices.
Monero wallet downloads ↗Pearl · official desktop wallet
Pearl Research Labs publishes its native PRL wallet for Windows, macOS and Linux. This prepares a receiving wallet; Pearl mining in Etheraio still awaits worker checks.
Get the official Pearl wallet ↗No affiliate or referral reward is enabled on these links. We did not verify a suitable reward offer for these native-coin routes. Exodus publishes a separate conditional referral program, but its terms exclude UK residents. Provider terms ↗. No purchase or swap is required to receive supported mining payouts.