Skip to content

About

DarkWow chain research, updates weekly.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Latest commit

 

History

5 Commits

Folders and files

Repository files navigation

DarkWow Chain Research

Weekly, source‑verified research notes on DarkWow — Patrick Mockridge's privacy‑first fork of DarkFi built around composable Object‑Capability (O‑Cap) primitives, Uncle Merkle proof‑of‑work consensus, a zero‑premine emission schedule and per‑block Pedersen supply audits.

Each week's topic is set by Patrick Mockridge, DarkWow's author. The write‑up takes that slice of the chain, reads the primary sources (The DarkWow Book and the linear-master source tree), and documents it with diagrams, tables and small programs that reproduce every number quoted. The goal is a reference that a newcomer can trust because every claim points at the line of code or spec page it came from.

Status: an ongoing weekly research series produced in collaboration with the DarkWow project — topics come from Patrick Mockridge; the analysis, code and any errors are the researcher's own. DarkWow is unaudited software ("use at your own risk" per its README); nothing here is investment advice.

Weekly index

Week Dates Topic Artifacts
01 2026‑09‑21 → 09‑27 Genesis block & architecture — the nine genesis contracts, deterministic ContractIds, why only two are consensus‑critical, and the five structural breaks from DarkFi (O‑Cap vs DAO, the Money split, zero premine + emission, Uncle Merkle vs overlay‑DAG, ZK predicates vs ACLs). Deep dive: the three L1 genesis contracts — Box, Purse and Promissory Note — under zero knowledge: all ten Halo2 circuits, the params‑based wire format, the Book's complexity ceiling, and an owner‑binding gap in the Box/Purse leaf (§3) 11 diagrams · 5 charts · 4 models · 56 tests · 17 findings
02 2026‑09‑28 → 10‑04 O‑Cap governance primitives vs. monolithic DAOs — four bug claims (execution reordering in Box/Purse composition, attestation replay, stale oracle masking, manifest.toml capability aliasing) read against the Book's dao_escrow design and the code at d775e37c. One is not reproducible, one is not applicable, one is confirmed and understated (attestations can be forged with one secret), one is real but wallet‑side — and beneath all four the tx_binding chain is missing its fourth stage, so every proof is portable across transactions (§7); includes two corrections to Week 1 13 diagrams · 8 charts · 5 models · 102 tests · 17 findings

What's in the repository

.
├── README.md                     ← you are here
├── weeks/
│   ├── week-01/
│   │   ├── README.md             ← the Week 1 write‑up (start here)
│   │   └── charts/*.png          ← generated figures embedded in the write‑up
│   └── week-02/
│       ├── README.md             ← the Week 2 write‑up: four bug claims vs. the code
│       └── charts/*.png
├── code/
│   ├── darkwow_research/         ← pure‑Python ports and models
│   │   ├── emission.py           ←   expected_reward() / fixed_pow_decay(), integer‑exact
│   │   ├── uncle_split.py        ←   compute_reward() pin split + mass‑balance invariant
│   │   ├── genesis.py            ←   the nine genesis contracts, counters, crates, roles
│   │   ├── l1_circuits.py        ←   Box / Purse / Promissory Note circuits, tiers, wire formats
│   │   ├── ocap_primitives.py    ←   the six governance primitives, 41 circuits, dao_escrow Book‑vs‑code, composition matrix
│   │   ├── call_tree.py          ←   DarkTree post‑order, slot checks, atomic execution; escrow ClaimV1 and Book‑era TreasurySpend
│   │   ├── binding.py            ←   attestation authority table, forgery path, the tx_binding chain, 19/11/1/1 split
│   │   ├── oracle_freshness.py   ←   oracle push/read model, staleness‑vs‑false‑reject simulation
│   │   └── manifest_caps.py      ←   manifest surface, trust layers, name‑aliasing demo
│   ├── plot_charts.py            ← regenerates every chart and data table
│   └── tests/                    ← pytest suite mirroring the upstream Rust unit tests + the notes' claims
├── data/                         ← CSV outputs (emission, genesis, L1 circuits, wire formats, composition matrix, attestation authority, …)
└── sources/                      ← primary‑source snapshot + provenance (The DarkWow Book, PDF)

Quick reference — numbers you will see quoted

All re‑derived from src/sdk/src/blockchain.rs and src/sdk/src/crypto/contract_id.rs; see Week 1 §2.4 for derivations.

Parameter Value
Block time 120 s (262,980 blocks / year)
Proof of work RandomX; Monero merge‑mining supported
Genesis height 1, timestamp = 0, target = u32::MAX, 1 coinbase + 9 deployment txs
Genesis ContractId poseidon_hash([42, 0, counter]), counters 2 … 10
Consensus‑critical contracts Deployooor (2), NativeToken (4) — the other seven are ecosystem primitives
Initial reward R₀ 1,383,764,049 base units ≈ 13.8376 DRKW (1 DRKW = 10⁸)
Decay continuous, half‑life 1,051,920 blocks ≈ 4 years, no step halvings
Tail emission 79,853,981 base units ≈ 0.7985 DRKW / block, forever (≈ 210 k DRKW / yr)
Tail onset height 4,327,299 ≈ 16.45 years; supply then ≈ 19.78 M DRKW
21 M DRKW a reference supply, crossed ≈ year 22 — not a hard cap
Uncle pins base / 2^depth, depth 1 … 6; subtractive, canonical + Σ pins == base

Reproduce

python3 -m pip install -r code/requirements.txt
( cd code && python3 -m pytest )      # 158 tests
python3 code/plot_charts.py           # ~20 s; rewrites weeks/*/charts and data/

Python ≥ 3.10; the models themselves have no dependencies beyond the standard library — only the plotting script needs matplotlib/numpy.

Sources

  • Code: PatrickMockridge/DarkWow (GitHub mirror of codeberg.org/PatrickM123/darkwow), branch linear-master, AGPL‑3.0‑only. Short excerpts are quoted in the notes for commentary, with file paths.
  • Spec: The DarkWow Book, the project's mdbook documentation, snapshotted as a PDF on 2026‑09‑03 — sources/ has the file, its checksum and how to rebuild a fresh copy from doc/.
  • Original notes (Weeks 1–2): Google Doc.

Conventions

  • One directory per week: weeks/week-NN/README.md plus its charts/. Notes are written to stand alone.
  • Every quantitative claim carries a tag — 📖 Book, 🦀 Source, 🧪 Re‑derived — and re‑derived claims have a test.
  • Discrepancies between spec prose and code are recorded in each week's Findings section rather than silently corrected.
  • Corrections and questions are welcome via issues.

About

DarkWow chain research, updates weekly.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages