Why PowMachina

Mining without a middleman.

Every share you find goes into the chain, and the chain pays you for it. No pool, no operator, nobody holding your coins.

CPU miner GPU miner share block

01 · The problem

Proof of work ended up with a few pools.

A small miner who mines alone waits weeks for a block, so they join a pool. The pool evens out the luck — and in exchange it holds their payouts, takes a fee, picks the transactions and casts their vote.

Thousands of miners, a handful of operators. Every one of them is a single point that can be hacked, pressured or simply disappear with the balance.

02 · The idea

The pool is part of the protocol.

A share is a block header that met a target 3,000 times easier than a block’s. Miners publish them like transactions, about a hundred a second across the network.

Every block carries the shares it has seen in its settlement, and every full node checks each one. A share has eight blocks to make it in, so one block that leaves it out doesn’t make it disappear.

03 · The price

A fixed price makes censorship pointless.

The obvious rule — a block splits its reward among the shares it includes — has a flaw: leave everybody else out and your own shares are worth more. In our simulation a miner with 45 % of the hashrate earned 1.4 to 1.8 times its fair share that way.

So PowMachina pays every share a fixed price from a pot that holds about sixteen blocks of reward. Leaving shares out earns nothing; the same censor ends up with 0.99 of its fair share — less than if it were honest.

Simulated with powm-sharesim, which is in the public repository. Run it yourself.

04 · Your work stays yours

Mine through anyone. Nobody can take it.

Your payout address, your finality key and your share’s weight are hashed into the proof of work itself. Change one bit and the share is worthless.

That makes mining through a stranger’s node safe: it can refuse your work, but it can’t redirect it. No accounts, no trust, no withdrawal button.

05 · The payout

Paid by the chain, once a day.

Each settled share adds to your credit, which lives in the chain’s state — every node knows it, and every block commits to it. Once a day the block reward itself pays it out to your address, from 0.05 POWM up.

A small miner with one CPU gets paid like everybody else: less, but steadily, without waiting weeks for a lucky block.

06 · CPU and GPU

Two lanes, half each.

Processors mine RandomX, graphics cards mine a memory-hard algorithm over a dataset that changes every week. Each lane has its own difficulty, and the chain never takes more than two blocks of one lane in a row.

A share is priced against its own lane’s block, so a CPU share and a GPU share of the same work are worth the same. Rented CPU power can’t rewrite the chain alone: it would need the graphics cards too.

07 · Finality

The miners who did the work sign it.

Every block carries its miner’s finality key. The quorum for a block is everybody who mined in the day before it, each lane holding half the weight. Once more than two thirds have signed, about 40 minutes later, no amount of hashrate can rewrite it.

Nobody appoints the quorum and no key list ships with the software. Mining through a public stratum, your key stays yours, so the server can’t vote with your hashrate.

Before a block is final, deep reorgs are already expensive: an hour-deep one needs 2.7 times the network’s work.

08 · What we build on

Nothing here comes out of nowhere.

Each piece has been proposed, published or run before. What we haven’t found anywhere is all of them in one coin, in consensus, with the payouts enforced by every node.

  1. Bitcoin

    2008

    Proof of work, the longest valid chain, and a coin that nobody can print. Everything else here is built on that.

  2. P2Pool & Braidpool

    2011 – today

    Decentralized pools as a share chain next to the coin. They proved miners can pay each other without an operator; we move the shares into the coin's own consensus.

  3. FruitChains

    Pass & Shi, 2017

    Near-blocks paid from the reward, so that small miners earn steadily. The closest ancestor of paying shares on-chain.

  4. HotPoW & Tailstorm

    Keller et al., 2019 – 2023

    Votes, finality and blocks built from many small proofs of work. Research we learned from; here it runs on a live network.

  5. RandomX

    tevador, Monero

    The CPU algorithm, audited and battle-tested for years. Used as it is, with our own salt, not rewritten.

  6. Ethash & ProgPoW

    2015 – 2018

    Memory-hard GPU mining over a dataset that changes every epoch. Our GPU algorithm keeps the idea and caps the dataset at 8 GB.

09 · The numbers

All of it, on one screen.

Straight from the consensus code. The full reasoning is in the Sharepool spec, finality and the chain spec.

Block time
30 s, the two lanes alternating
Lanes
C: RandomX v2 (CPU) · G: MachinaHash-G (GPU)
Same lane in a row
at most 2
Shares
about 100 a second; the smallest is 1/3000 of a block
Share lifetime
8 blocks to be settled
Payouts
once a day, from 0.05 POWM, straight to your address
Finality
> 2/3 of the last day’s miners, 80 blocks deep (~40 min)
Reward
10 POWM a block, halving every ~4 years
Tail emission
0.24975 POWM a block, forever
Start
no premine, no sale; the reward ramps up over ~7 days
Development fee
5 % of the block reward for ~2 years, then none

10 · The testnet

Now it has to work with real people.

Simulations show the design holds. The testnet has to show it under real conditions: many small miners paid fairly, shares crossing continents and bad connections, and people trying hard to cheat. Consensus can’t be patched quietly after mainnet, so this is the time to find what’s wrong.

Testnet coins have no value, and there is nothing to buy.