Show the work.

Wallets, receipts and miner jobs.

Two validated development cycles added an encrypted ML-DSA-65 wallet, bounded signed-transaction submission with durable receipts, and a versioned CPU miner job/solution CLI. All 49 tests passed, including independent C++ share verification, fee burning, replay and restart recovery.

These are development tools. The miner CLI is standalone and does not yet submit work to the persistent service. The final mining algorithm remains unfinished.

A persistent development chain.

The ledger now saves its chain in SQLite, replays it at startup, and resumes after a restart. All 21 server tests passed, including process-crash recovery and corrupt-state rejection. A move onto the large storage partition preserved the existing chain and identity.

The CUDA SHA-256 baseline produced an independently verified result. The 4 GiB memory probe passed 16 sampled lane checks on an RTX 3050. These are baseline and synthetic results, not final proof-of-work validation.

A signed transfer, end to end.

The local ledger now verifies ML-DSA-65 signatures, validates UTXO ownership and value, applies coinbase maturity, rejects invalid blocks and accounts for fee burning. A four-block demonstration completed a transfer and replayed the export using a fresh instance of the same implementation.

Validation: 16 automated cases passed, including the original hash/accounting checks and new rejection tests for double spends, signature changes, cross-network contexts, invalid ownership, reward overpayment and bad development work.

Limit: this is a local development harness with fixed, trivial SHA-256 work. Peer networking, a public network, final GPU mining and security audit are not complete.

Bandwidth is the hypothesis.

The mining research now explicitly targets H200 and B300 performance alongside RTX 5090 viability. The dataset candidate is 20 GiB, with smaller sizes retained for comparison. A larger allocation alone does not prevent ASICs or establish a GPU advantage.

A CUDA memory probe and independent host sample check are included as source. Those hardware runs were pending in the original release; the later server entry above records the completed baseline and 4 GiB probe checks. No hashrate or energy-efficiency result is published.

The first protocol proposal.

The initial whitepaper proposes post-quantum authorization, capped issuance, a transaction-fee burn and open GPU participation. The project name remains provisional. There is no live asset, token sale or promised launch date.

Measure before committing.

Run the GPU workload on actual NVIDIA hardware, publish reproducible measurements and evaluate verifier costs and shortcuts. Build peer networking only around an explicit, reviewed consensus specification.