PARALLEL BIT CHAIN · PROOF OF WORK FOR THE MACHINE ECONOMY

btc

BTC, flipped — humans on Bitcoin, agents on PTC.

A GPU-native PoW chain that keeps Bitcoin's ledger semantics exactly — UTXO, 10-minute blocks, 21 million coins, four-year halvings — and flips two things: the proof of work (an ASIC must lose to a graphics card) and the subject (this is where autonomous agents settle, budget, and notarize).

Read the whitepaper GitHub ↗

design stage · P0 spec freeze in progress · no code yet, no coin yet

GENESIS COMMITMENTepoch_seed(0) = keccak256("PTC/mainnet/genesis-seed-v1")

600 s

block interval

21,000,000

total supply

210,000

blocks per halving

128 B

block header

01 · THE MIRROR

Same ledger. Opposite proof. Mirrored subject.

Flip a lowercase b and you get a p. Bitcoin is humanity's chain; PTC is its mirror — the same ledger discipline, an opposite proof of work, and a mirrored subject: the machine economy.

IDENTICAL TO BITCOIN

  • =UTXO transaction model
  • =600-second blocks
  • =21,000,000 PTC, halving every 210,000 blocks
  • =Bitcoin difficulty formula — 2016-block window, ±4× clamp
  • =Bech32m addresses, HRP ptc
  • =Taproot: Schnorr default, ECDSA compatible
  • =MuSig2 multisig, CLTV/CSV timelocks

FLIPPED

  • ⇅Geyser PoW — memory-bandwidth-bound hashing
  • ⇅Per-block random programs — 256 mixed INT/FP instructions
  • ⇅Scheduled algorithm rotation via algo_version
  • ⇅128-byte header with algo_version + mix_digest
  • ⇅Mining economics: buy a GPU, don't tape out an ASIC
  • ⇅The subject: humans settle on Bitcoin — agents settle on PTC

02 · AI-NATIVE

Three things an agent can't get elsewhere

Agents are natively keypairs — no passport, no bank account, 24/7, APIs for senses. They don't need another smart-contract casino; they need settlement they can budget against, and a notary for their records. Demand-side positioning, zero consensus changes.

01

Headless-first tooling

Agents have no screens. Wallets, node APIs, mining protocols: APIs before GUIs. A headless wallet SDK ships with the reference stack in P1 — the GUI is the compatibility layer, not the entrance.

02

Money an agent can budget against

21,000,000 cap, halving schedule fixed at genesis, a script subset with no DeFi casino. Predictable monetary policy is a feature for machine cost functions — that is what Bitcoin-isomorphism buys.

03

The machine notary

Model cards, inference commitments, training-data fingerprints, agent logs — hashed via OP_RETURN, anchored routinely on PTC and terminally on Bitcoin. The machine economy periodically notarizes itself into humanity's ledger.

Not building: a smart-contract platform (agents are their own runtimes) · an “AI governance token” · on-chain “verified intelligence” — the chain verifies hashes and signatures, not intent. Zero consensus bytes changed.

03 · GEYSER · PROOF OF WORK

Three layers between ASICs and your GPU

01

Bandwidth binding

Every hash streams random 8 KiB reads across a 6 GiB dataset. The bottleneck is memory bandwidth — and consumer GPUs are the platform-wide optimum of hashrate per dollar per GB/s. CPUs, FPGAs and AI accelerators are structurally priced out.

02

Random program layer

Each block's PoW program is generated from the previous block hash: 256 mixed integer/float instructions, reshuffled every block. An ASIC that wants to keep up has to build itself into a GPU.

03

Rotation covenant

algo_version hard-forks the algorithm on a published 2–3 year schedule. GPU miners update software; ASIC tape-outs turn into landfill.

Target: ASIC efficiency capped at ≤2× with no payback · dataset ≤60% of median-GPU VRAM (6 GiB today, +32 MiB per epoch)

04 · HONEST ENGINEERING

What we don't claim

A chain that hides its trade-offs is marketing, not engineering. The ugly parts are written down, with numbers, in the docs.

No absolute ASIC immunity

It does not exist, and anyone promising it is selling something. The goal is economic: hold ASIC efficiency at ≤2× and destroy the payback period.

Birth is fragile — said out loud

New chains are weakest at launch. That is why mainnet ships only after six-plus months of public testnet, with signed checkpoints against early private chains.

Zero premine, zero dev tax

No allocation to founders. Emission starts at genesis on Bitcoin's exact schedule — one curve, nothing to argue about.

Macs are legitimate miners

Apple-silicon UMA needs zero consensus changes: M4 Max ≈ 0.8× an RTX 5070 at lower energy per hash. A Metal backend lands in P2.

Blocks stay small

4M weight units with no growth policy. Throughput belongs on L2; the chain itself carries hashes, not files.

No quantum theater

No PQC at genesis — no CRQC exists. The extension points are reserved now and triggered by capability milestones, not by marketing.

05 · ROADMAP

Four phases, no shortcuts

  1. P0NOW

    Spec freeze

    Cross-platform test vectors, bit-identical on five platforms — NVIDIA, AMD, Intel, Apple GPU, ARM64 CPU. The whitepaper graduates to v1.0.

  2. P1

    Full node & miners

    Go full node, CUDA and ROCm miners, Stratum v2, headless wallet SDK (agent-first).

  3. P2

    Incentivized testnet

    Six-plus months of public testnet, Metal backend for Apple GPUs, dual external audits.

  4. P3

    Mainnet

    Genesis with signed checkpoints against early private chains. Zero premine, zero dev tax.