Hearth — Money Mined at Home

A people-mined, ASIC-resistant proof-of-work chain that speaks Ethereum.
Mine EMBER on the computer you already own — then deploy to it with the tools you already have.

🌐 explorer.cloudsforge.online · https://rpc.cloudsforge.online · chain id 7411
🧪 testnet (mining again since 2026-08-11) · explorer-testnet.cloudsforge.online · https://rpc-testnet.cloudsforge.online · chain id 7412
_{Hearth is the Mine in CloudsForge's one crypto world — mine it, trade it, mint it, spend it, play in it.}

proof of work · CPU mining · ASIC-resistant · fair launch · EVM · chain id 7411

---

Status, before anything else

Hearth is an account-model, EVM-executing proof-of-work chain, and mainnet is live. Chain id 7411, Ethereum JSON-RPC at https://rpc.cloudsforge.online, height 13,946 and climbing. Measured 2026-08-11 14:04 UTC from outside this network, with no credentials: eth_chainId → 0x1cf3, eth_blockNumber → 0x367a, web3_clientVersion → Hearth/v0.3.0/linux-x64/node22.23.1, and the height moved five blocks across a 75-second poll. A GET answers 405 with allow: POST,OPTIONS, because this is a JSON-RPC endpoint and not a page. Block 1 was mined 2026-08-04 19:12:21 UTC and the chain has not stopped since — just under seven days at that measurement.

There is a consensus change waiting at height 20,000, and a node still on 0.2.x will silently stop following the chain when the tip crosses it. The emergency difficulty rule shipped in 0.3.0; below the gate the old rule applies exactly as before. Both published chains already run 0.3.0. If you run your own node, upgrade before mainnet gets there — docs/mining.md is the whole story and node/src/params.js is the arithmetic.

What is not there yet, said in the same breath, because a chain id without this paragraph reads as a promise. There is no EMBER of any monetary value: no market, no listed price, no liquidity. There is no payout path of any kind in this repository. Nobody outside the project has used any of it. And the public testnet is deliberately stopped (below).

The chain is nearly empty, and "nearly" is the honest word rather than "entirely". Walking every block from 1 to 10,987 on 2026-08-10 counts 62 transactions in 52 blocks — a mean of one transaction per 177 blocks. They fall into three groups, and none of them is a stranger using this chain:

• Blocks 235–251, 2026-08-04. 18 transactions, of which nine are successful contract creations — ForesightMarket instances belonging to another repository in this estate. They are real: eth_getCode at 0x49408b99deb3afaafd914ed9f0e71a89874b980e returns 6,007 bytes of runtime code, and the creation receipt reports status: 0x1. So "no contract is deployed on mainnet" is false, and this README asserted it until 2026-08-10.
• Blocks 1,323, 1,506 and 1,518. Three transactions, days apart.
• Blocks 10,851–10,967, 2026-08-10 17:19–17:23 UTC. 41 transactions, roughly one per block, every one a plain value transfer with empty input from a single address to a single address — the signature of an automated sweep rather than of anybody using the chain.

What is unqualifiedly true is narrower and more important: no block has ever been produced at production proof-of-work parameters, and no third party has ever transacted here.

"Live" means reachable and mining, not established, and each measurement below says something the height alone does not.

• Every block this chain has ever had was mined by this project. There has never been an independent miner. This is the fact that does not expire, and it is the one to read instead of a difficulty number.
• One browser tab can move this chain's difficulty by a factor of 32 and then stall it for twenty minutes. That is not a hypothetical: it happened on the evening of 2026-08-10, and the chain spent 1,154 s at a standstill afterwards. Difficulty is therefore a live reading that oscillates, not a property — which is why no number for it appears in this README. The account, the measurements and the cause are in MAP.md §1, stated once there and nowhere else in this repository.
• The mean interval from block 1 to the tip is 46.8 s (measured 2026-08-10 17:56 UTC), against a 15 s protocol target. Dividing height by age and concluding the block time is broken is the wrong conclusion; 15 s is a design target, not an observed rate, and the observed rate is whatever the one or two machines mining at the time happen to supply.
• It is one home server behind a single Cloudflare Tunnel — no redundancy, no failover, and no backup that has ever been restored.

The public testnet is mining again. Chain id 7412 at https://rpc-testnet.cloudsforge.online: measured 2026-08-11 14:04 UTC, eth_chainId → 0x1cf4, eth_blockNumber → 0x1f22 (7,970), a tip timestamp three minutes before the reading, and the height advanced across a 75-second poll. Sends confirm and the faucet can pay.

It was deliberately stopped from 2026-08-08 18:00:11 UTC to 2026-08-11, at height 7,765, because the machine running it also ran bitcoind and dogecoind in initial block download and testnet mining competed with them for the same disk and bandwidth. Two things retired that reason: bitcoind reached the tip, and the estate split the chain daemons onto a host of their own. The testnet miner was then repointed at the public RPC name and restarted. What has not changed: no genesis off mainnet outlives a docker compose down -v, so testnet state is still disposable.

The public surfaces around it all answered 200 in the same measurement: cloudsforge.online, network.cloudsforge.online, explorer.cloudsforge.online, explorer-testnet.cloudsforge.online and network-testnet.cloudsforge.online/faucet — re-measured 2026-08-11 14:04 UTC, all still 200.

The EVM is built and gated on Ethereum's published reference vectors, and consensus on the account model now runs in public rather than only on loopback. The original UTXO ledger is being retired.

Built and vector-gated	keccak/RLP/uint256/secp256k1 · Merkle Patricia Trie + StateDB · the interpreter (609/609 VMTests) · transactions, receipts, bloom (188/188 TransactionTests) · the state transition (20,077/20,077 GeneralStateTests) · all nine precompiles including bn128 and blake2f · the eth_* JSON-RPC surface · an EVM-aware explorer · the hearth CLI with an opcode tracer · a browser wallet on secp256k1
Proved end to end	Uniswap V2 runs on our own EVM — node/test/dex.js, 167/167, a real swap at 112,456 gas. And the m-of-n that has to hold feeToSetter before any of it is deployed — node/test/multisig.js, 143/143
Built and running locally	Consensus on the account model. hearthd --evm --mine produces and validates blocks and serves eth_* on 8545; two real nodes partition and reorg in node/test/evm-p2p-fork.js; docker-compose.testnet.yml runs three on chain id 7412
Published, and mining	Mainnet, chain id 7411, at https://rpc.cloudsforge.online — publicly trusted TLS, JSON-RPC over POST (a GET answers 405), plus an explorer at explorer.cloudsforge.online. Height 13,946 and climbing, measured 2026-08-11 14:04 UTC. The node ports still bind 127.0.0.1; a Cloudflare Tunnel on one home server is the only thing routing them, so this is one machine with no failover. It is mined by two hosts since 2026-08-10 — the machine the node is on, and a second one that reaches it the same way a stranger's miner does
Published, and mining again	Testnet, chain id 7412, at https://rpc-testnet.cloudsforge.online — same POST-only surface, same publicly trusted TLS, with an explorer at explorer-testnet.cloudsforge.online and a faucet at network-testnet.cloudsforge.online/faucet. Height 7,970 and moving, measured 2026-08-11 14:04 UTC; it was stopped at 7,765 from 2026-08-08 to 2026-08-11 and resumed once bitcoind synced and the chain daemons moved to a host of their own. P2P is a WebSocket at wss://p2p-testnet.cloudsforge.online/p2p — only the /p2p path is routed, the host root answers 404, both re-checked 2026-08-11
Still not published	Any contract of record — nine ForesightMarket instances are live on mainnet from 2026-08-04 and no file in this repository names one of them, which is the defect rather than the achievement. Off mainnet no genesis outlives a docker compose down -v, so testnet state is not durable and should be treated as disposable
Measured, and open	The proof of work is 64 KiB and cannot be raised: a 2 GiB pad measures 185.7 s per evaluation and a validator pays one per block received (docs/pow-parameters.md). Making it meaningfully memory-hard needs an amortised dataset, not a constant

MAP.md is the verified inventory — every claim in it cites path:line or a command that was run. Where this README and MAP.md disagree, believe MAP.md.

---

Why Hearth exists

Proof-of-work was supposed to be one CPU, one vote. Instead Bitcoin became a game for warehouse farms and a few pools, and almost everything since became a speculative chip you hoard rather than money you spend.

And the CPU-mineable, fair-launch corner of the space is full of coins with no contracts, no tooling, no wallet support and no way for anyone to build anything. Fairness that nobody can use is a moral position, not a network.

Big crypto problem	What Hearth does
Mining centralized into ASIC farms	Homefire PoW: memory-hard and CPU-friendly, so a farm earns little more per dollar than your laptop. A winning proof must be signed by the key its coinbase pays, so work handed to you cannot be redirected — see docs/mining.md for what that does and does not buy
Development captured by VCs / premines	Fair launch + an on-chain Commons treasury (10% of each block). Genesis mints zero spendable coins, checkable in about thirty seconds
The "fee cliff" (security dies when rewards end)	Perpetual tail emission funds security forever
A chain nobody can build on is a chain nobody uses	Hearth speaks Ethereum. 0x… addresses, secp256k1, 18 decimals, chain id 7411, Shanghai semantics, standard eth_* JSON-RPC — so MetaMask, ethers, viem, Hardhat and Foundry work without knowing this chain is bespoke
Too hard for normal humans	A web wallet and a browser miner that need nothing installed, and a reference node that is a full node, a wallet and a miner in one process

Full argument: WHITEPAPER.md. The specification: docs/evm-spec.md.

---

The coin

• Network: Hearth · Coin: Ember · Ticker: EMBER · Chain ID: 7411 (testnet 7412)
• Decimals: 18 — specified; node/src/params.js still defines 1e8 and has not moved yet
• Block time: 15 seconds · PoW: Homefire (memory-hard, CPU-friendly, ASIC-resistant)
• Emission: a deterministic integer schedule — 6 EMBER at genesis, 2-year half-life, perpetual 0.3 EMBER tail; 10% to the Commons
• Supply: uncapped, disinflationary. No hard cap and no fee burn — gas is paid to the coinbase in v1

 yr      reward   issued/yr      supply     commons    gross%
  1        4.50  11,045,161  11,045,161   1,104,516   100.00
  2        3.00   7,889,401  18,934,562   1,893,456    41.67
  5        1.13   2,761,290  31,163,132   3,116,313     8.86
 10        0.30     631,152  36,827,722   3,682,772     1.71
 30        0.30     631,152  49,450,762   4,945,076     1.28

Those are the numbers the consensus schedule produces (node/src/params.js); reproduce them with the command in docs/tokenomics.md §3. docs/coinnomics.md carries a different table from a smooth exponential model with an assumed fee burn — it is the historical design rationale and its numbers are not the chain's.

---

This is a working application, not a slide deck

Everything below runs, and every number was produced by running it:

• ✅ An EVM we wrote ourselves (node/src/{crypto,state,evm,chain}/) — Keccak-256, RLP, secp256k1 recovery, uint256, the Merkle Patricia Trie, StateDB, the interpreter, the Shanghai gas schedule and all nine precompiles. No @ethereumjs/*, no ethers, no web3, no runtime dependency of any kind
• ✅ Gated on Ethereum's own vectors — 609/609 VMTests, 20,077/20,077 GeneralStateTests, 188/188 TransactionTests, plus RLP and Trie tests. No component is done until its vectors pass
• ✅ Uniswap V2 runs on it — node/test/dex.js deploys the factory, router, pair and WEMBER, adds liquidity, swaps, swaps back, exercises permit and removes liquidity. 167/167, swap at 112,456 gas against mainnet's ~150,000
• ✅ HearthMultisig, the first contract here that is not a port — an m-of-n for HearthV2Factory.feeToSetter, which has no timelock and no two-step handover and therefore cannot be an EOA even briefly. node/test/multisig.js runs it on our own EVM: 143/143, signers rotate, a rotated-out key's existing confirmation stops counting, and a real factory refuses the EOA that deployed it while obeying the wallet
• ✅ the eth_* JSON-RPC surface (node/src/jsonrpc/) — 41 methods, 422 checks against a fake chain and 170 against a real one over HTTP; strict QUANTITY/DATA encoding, batches, notifications, revert payloads as code 3. Mounted on 8545 by node/src/evmnode.js
• ✅ hearth, the terminal tool — trace (an opcode-level debugger with gas, stack, memory and storage deltas per step), watch, wallet, call, send, deploy, devnet. 310 checks
• ✅ an EVM-aware block explorer — decoded logs, revert reasons, EOA-vs-contract with disassembly, ERC-20s, eth_getLogs search, and it renders "no node answered" rather than inventing data. It lived in web/ until 2026-08-04 and is now micro-explorer-web
• ✅ a non-custodial wallet on secp256k1 — its crypto is a port of the node's, and the two are run over the same random inputs and compared. That cross-check found a real gas bug in the node on its first run. Now micro-hearth-wallet-core, where the comparison is made in-process against node/src and nothing is checked against a fixture the library produced itself
• ✅ a developer kit (tools/) — a faucet, working Hardhat and Foundry templates, and an RPC probe that serves the real method surface over a fake chain
• ✅ contract verification (tools/verify/, 116/116) — including the API forge verify-contract speaks — and 🟡 an Etherscan-compatible /api (tools/explorer-api/: account, contract, stats, transaction, logs, proxy over a real address index) whose test suite currently fails. Zero dependencies. There is now a chain for both to serve — mainnet 7411 — and neither is hosted against it; that is a deployment that has not happened, not a chain that does not exist
• ✅ property fuzzing (node/test/fuzz/) — random bytes into uint256, the trie, RLP, transactions and the interpreter. It found two real defects in merged code and reports rather than patches; both are listed in SECURITY.md
• ✅ a real proof-of-work network (the UTXO chain) — most-work fork choice with reorg over real sockets, P2P gossip, LWMA difficulty, disk persistence
• ✅ mining you can run — hearth-mine on the command line and a desktop app (app-desktop/, Tauri) for macOS, Windows and Linux, both over /mining/template and /mining/submit. The browser miner that stood here was removed with web/: it signed 64-byte signatures where the node requires 65, so every block it ever found was refused. It lives in micro-network-site now, carrying the fix, and node/test/browser-pow.js and node/test/browser-proof.js compare it against this node
• 🟡 a Rust crate (rust/hearthd) — a self-check and a PoW benchmark. Not a node and not consensus-compatible, and it has no EVM at all — see docs/why-two-implementations.md

Two honest caveats about that list. It is driven by blocks now — mainnet has produced 10,987 of them as of 2026-08-10, carrying 62 transactions including nine contract deployments — but that is a rounding error of traffic, all of it produced by this project, so every path through the EVM is still proven by tests rather than by use. And the merchant-button mockup that used to sit in web/pay-demo.html — it simulated settlement on a timer — went with that directory; there is still no payment SDK and no payout path.

npm test passes from a clean clone — 39 suites, 86,451 checks, no install, no corpus and no network. Re-measured 2026-08-09 by cloning this repository into an empty directory and running it; docs/testing.md §1 lists every suite. Fetching the reference corpus completes two of them: blake2f goes 50/50 → 53/53 and bn128 81/81 → 86/86, running its one skipped case.

The caveat that used to matter most is closed. docs/robustness-review.md measured StateDB re-rooting both tries on every single mutation: a single 30M-gas transaction cost 443 MB of retained heap and 65 seconds of CPU, against a 15-second block time. Hashing is deferred to root() now and the same transaction measures 5.2 s and 9.2 MiB, gated by node/test/bench/block-execution.js.

The caveat that matters most now is the proof of work: every block ever produced used a 64 KiB pad, the 2 GiB the documents promised measures at 185.7 s per evaluation, and closing that gap is a redesign rather than a constant (docs/pow-parameters.md).

---

Try it now

1 — Run the test suites (needs Node 18+, no install, no network):

cd node
node test/interpreter.js      # 194 checks
node test/statetransition.js  # 133 checks
node test/jsonrpc.js          # 422 checks
node test/cli.js              # 310 checks
node test/fuzz/run.js         # 82,481 property-fuzz checks

2 — Run the real conformance gate (fetches ~350 MB of Ethereum's corpus):

cd node && ./scripts/fetch-vectors.sh
node test/conformance/runner.js --impl=test/interpreter.js \
     --dir=test/conformance/vectors/VMTests --no-gas          # 609/609
node test/conformance/runner.js --impl=test/statetransition.js \
     --suite=GeneralStateTests --dir=test/conformance/vectors  # 20,077/20,077

3 — Watch Uniswap V2 execute on our EVM:

pnpm --dir contracts install && pnpm --dir contracts compile
cd node && node test/dex.js                                    # 167/167
node test/multisig.js                                          # 143/143

4 — Point your tooling at mainnet:

cast chain-id     --rpc-url https://rpc.cloudsforge.online      # 7411
cast block-number --rpc-url https://rpc.cloudsforge.online      # climbing

The public testnet is where you should be if you are deploying anything you have not deployed before, and it is mining again as of 2026-08-11:

cast chain-id     --rpc-url https://rpc-testnet.cloudsforge.online   # 7412
cast block-number --rpc-url https://rpc-testnet.cloudsforge.online   # 7970, and climbing

Fund an address from network-testnet.cloudsforge.online/faucet. A chain of your own — hearthd --evm --mine, or the three-node docker-compose.testnet.yml — is still faster to iterate against, and nothing you put on testnet is durable.

Or serve the real RPC layer locally over a fake chain, to check your wiring and your encodings without mining anything:

node tools/rpc-probe/stub.js --port 8745
cast chain-id --rpc-url http://127.0.0.1:8745                   # 7411

Full walkthrough, with every step marked [RUN], [PROBE] or [WAITING]: docs/quickstart.md.

5 — Boot the UTXO network (the ledger being retired; mainnet 7411 is the chain producing blocks today, and it is account-model):

docker compose up --build      # seed + 2 miners + web on :8080

No Docker? ./scripts/run-local-network.sh. Details: docs/network.md.

---

Repository layout

hearth/
├── README.md · WHITEPAPER.md · MAP.md   the pitch, the argument, and the verified inventory
├── docs/                                evm-spec (authoritative) · decisions · quickstart ·
│                                        network-config · tokenomics · exchange-integration ·
│                                        listing-checklist · mining · records · architecture ·
│                                        robustness-review (measured) · testing (coverage)
├── node/                                the reference node — AND the EVM implementation
│   ├── src/crypto  src/state  src/evm  src/chain  src/jsonrpc   the account-model chain
│   ├── src/*.js                                                 the UTXO chain, being retired
│   ├── src/cli/  bin/hearth.js                                  the terminal tool + tracer
│   └── test/  test/conformance/  test/fuzz/                     27 suites + vectors + fuzzing
├── contracts/                           WEMBER, a Uniswap V2 port, Multicall3 (compiled, undeployed)
├── tools/                               faucet · hardhat · foundry · rpc-probe · metamask.md ·
│                                        explorer-api (Etherscan-compatible /api) · verify
├── app-desktop/                         the desktop miner (Tauri) — macOS, Windows, Linux
├── rust/hearthd/                        a self-check and a benchmark — not a node
├── proto/                               teaching scripts (the emission model is NOT consensus)
└── .github/workflows/ci.yml             six jobs; `npm test` is the single source of truth

Contributing

Hearth is a commons. See CONTRIBUTING.md. Highest-leverage areas right now: an amortised proof of work so the memory-hardness argument is more than a construction (docs/pow-parameters.md), and closing the items in docs/listing-checklist.md §7. Contract verification is already built (tools/verify/, 116/116) and the Etherscan-compatible /api shim is built and green (tools/explorer-api/) — both are waiting on a deployment, not on a chain. Mainnet 7411 has been answering since 2026-08-04; nobody has hosted either service against it.

Security

Please report vulnerabilities privately — see SECURITY.md. There is deliberately no security@ mailbox yet, and that is tracked rather than papered over.

License

MIT — money should be free to fork.

---

_{Keywords: Hearth, EMBER, cryptocurrency, proof of work, CPU mining, ASIC resistant, memory-hard proof of work, EVM chain, EVM implementation, Ethereum JSON-RPC, chain id 7411, Solidity, Uniswap V2, decentralized digital cash, mine crypto at home, no premine, fair launch, crypto wallet, block explorer, blockchain node, tail emission.}

How this was built

Parts of this repository were produced with AI assistance, and it seems worth saying so plainly rather than leaving it to be inferred.

The code in this repository was written by Claude Opus 5 and Claude Fable 5, assets generated with FLUX 2 Pro, under human direction and review.

This paragraph used to name different models, and the correction is itself a fact about the repository. It credited Claude Opus 4.8 for the code and OpenAI's GPT Image models for the art. Both were true when written and neither is true now: the chain was rewritten around the account model and its own EVM, decommissioning the code that attribution described, and every brand mark and piece of in-game art was regenerated with FLUX 2 Pro. An attribution that survives the work it describes is just a stale sentence with a credit in it.

The generated originals carry C2PA provenance metadata written by the model, so the claim above is checkable rather than merely asserted — branding/hearth-logo.png and branding/mark-hearth.png hold it. The wordmark, favicon, Open Graph and social files do not, because they are derived from those originals by downscaling and cropping, and resampling discards the metadata along with the pixels. (This list named a third holder, web/art/mine-hero.png, until that tree was deleted in 48bc28a.)

The models were used under paid API access and the output is the project's to use. Nothing here is claimed to be hand-written that is not, and nothing is claimed to work that has not been tested. The conformance-vector discipline exists partly for this reason: an implementation is judged by whether it passes the reference vectors, not by who or what wrote it.