# Implementation status Updated 2026-10-06. Sources reviewed: repository Markdown, the DOCX concept note, chain reference modules/tests, public protocol/API pages, and Wrangler configuration. | Area | Actual state | Remaining work | |---|---|---| | Protocol | SPEC, DECISIONS, THREATS and concept note exist | Formal models, parameter evidence, constitutional review | | Governance reference | Python model; 430 tests pass with Python 3.14 | Port to production keepers and prove equivalence | | Consensus/network | Go CometBFT G0 network with seven local validator configurations | Independent operators, validator policy, fault/censorship drills | | Transactions | Real Ed25519 signatures, chain binding, sequences, height expiry, strict JSON, bounded quotas and opt-in delayed key rotation | Versioned protobuf/CBOR, guardian recovery and session keys | | State | Durable snapshots, deterministic hash, committed-state queries | Scalable database, Merkle proofs, snapshots/state sync | | Documents | Signed SHA-256 content-addressed publishing up to 64 KiB | Availability attestations for larger documents | | Registration | Roadmap only | Worker/D1, challenge-first admission, signed evidence, admin review, SSRF defense | | Governance on chain | Opt-in native G1 identity-security keeper; broader governance remains reference-only | Identity/roles → proposals/deliberation → examiner/tally → escrow/review → elections/courts | | Cryptography | Ed25519 on G0; HMAC stand-in in reference | Sealing/timelocks, beacon verification, threshold custody | | Website | Live static manual, agent discovery, architecture/integration guides and read-only documentation MCP | Public governance API, SDK, versioned OpenAPI and schemas | | Deployment | Cloudflare website live; native validators local only | Governance gateway/read mirror; independently hosted nodes; monitoring/recovery | | Production readiness | G0 only | SDK integration, audits, simulation, independent validators, external anchoring | ## Community simulation (2026-10-06) `simulation/` now exercises the existing reference rules as a complete community story, paired across comprehension-weighted and flat voting. The pilot runs three seeds with 1,000 citizens and 100 leaders, two elections and nine project votes per run. It covers qualification, credit allocation, recusal, comprehension, panel fraud/revocation, certification default, challenges, quorum refunds, milestone escrow, failure penalties and political turnover. Detailed assumptions and results are in `simulation/README.md` and `simulation/results/pilot/REPORT.md`. The pilot manifest is published on G0 at height 11, with matching hashes and document reads on all seven validators. The network records a commitment to the results; governance execution remains in the Python reference. HTTP admission, real AGI behavior, court semantics, actual execution and cryptographic sealing are not validated by this simulation. ## Delivery order 1. Validate the G0 network, transaction admission, persistence/replay, and fault behavior. 2. Implement challenge-first registration and reviewed identity admission, reconciling canonical states/version names and keeping the public API documentation aligned with the implemented service. 3. Introduce Cosmos SDK module accounts and port identity/roles and proposal state machines. 4. Port examiner panels, comprehension tokens, snapshots, tally and escrow together; grants must only originate from finalized passing tallies after challenges. 5. Add elections, execution review, courts, beacons/anchors and light verification. 6. Run the adversarial agent simulation and formal checks before a capped-budget pilot. The G0 application deliberately exposes no simplified voting endpoint: a majority-only shortcut would bypass the required comprehension and recusal rules. It also cannot transfer treasury units before tally authority exists. ## Verification on this workspace - Existing reference: 209 tests passed with Python 3.14. - Go application: `go test -race ./...` passed. - Seven-validator smoke: signed document committed at height 9, matching committed app hashes and document reads on all seven validators; replay rejected by account sequence after bypassing the mempool byte cache. Nodes resumed saved state. - All integration processes stopped after the check; network is not left running. The larger seed-7 experiment also completed in WEIGHTED and FLAT modes with 10,000 citizens and 100 leaders. Each passed 130 checks and recorded 98,420 events. Its manifest is anchored at height 14. Both sets of artifacts passed verification; all six pilot runs additionally passed exact deterministic replay. See `simulation/FINDINGS.md` for measurements and limitations. These simulations exercise institutions with synthetic policies; real agent deliberation and on-chain governance enforcement remain open delivery gates. ## Adversarial hardening (2026-10-06) See [review](https://github.com/Martines90/dagp/blob/main/chain/security/REVIEW.md) for fixed attacks and unresolved production gates. Governance remains a reference model; these fixes do not implement native keepers. Go 1.26.8 / gRPC 1.83.2 remove all reachable findings in the official dependency scan. New transaction/storage quotas are consensus changes requiring a coordinated upgrade. Fresh paired community results use v2 rules; historical artifacts remain immutable. Second-pass hardening adds an authenticated milestone-payment reference controller, recovery coalition checks at activation and strict refinement/pause inputs. 252 reference tests and six simulator tests pass. The live seven-node crash campaign passed progress with two offline, halt with three, and restart/state agreement. [Security contract](https://github.com/Martines90/dagp/blob/main/chain/security/PROTOCOL.md) records native implementation obligations; these additions do not make G0 a production governance or financial chain. ## Coordinated insider protection (reference) The [insider protocol](https://github.com/Martines90/dagp/blob/main/chain/security/INSIDER_PROTECTION.md) adds rolling official/operator/ system quotas, a signed frozen-roster peer-containment vote and bounded independent review. Registrar freezes preserve civic and key-custody rights; citizenship removal or judicial restoration requires court authority. Operator-wide bans also require public ratification. Cluster actions are atomic and individual validator sanctions preserve quorum/minimum size. These are reference rules awaiting native keepers. The [insider scenario](https://github.com/Martines90/dagp/blob/main/chain/simulation/results/self-protection/insider-abuse.json) uses 1,000 citizens, 50 administrators and 10 coordinated attackers. All 47 checks pass: 20 accepted temporary freezes, 30 refused attempts, no unilateral voting exclusions, a failed minority recall and successful peer containment/court dismissal of all ten. Community v3 also completes two modes, 260 checks and 20,472 exactly replayed events. Historical v1/v2 reports remain unchanged and need their matching implementation. Final validation: **283 governance tests** (including **31 insider regressions**) and **six simulation tests** pass. The **47-check** insider scenario and **20,472** community events replay exactly. The manifest committing to both reports was replicated by all seven local validators at G0 height **30**. This anchors evidence; it does not execute these governance protections on G0. ## Supervised proposal review and budget allocation [Review/budget protocol](https://github.com/Martines90/dagp/blob/main/chain/security/REVIEW_BUDGET.md) implements signed party threads, versioned refinements, two independent vote-supervisor approvals, owner-signed record locking and exact milestone budgets. All nine community proposals use this controller. 306 governance tests (including 23 review/budget regressions) and six simulator tests pass. Monetary commitments reserve common funds at vote-open and convert them into reviewed milestone escrow only after successful finalization. Legacy reference tranche attestations remain trusted inputs; native evidence enforcement is still required. ## Monthly policy and clause voting (2026-10-06) The reference now enforces a 20% minimum turnout, strict 5%-step credit allocation, calendar-month replacement without carryover, and debt/old-month-refund protections. Signed reviewed parameter referendums require 66% decisive voting power, have a bounded allowlist and activate next month without changing open-session rules. Reviewed multi-point bills independently tally clauses, enforce per-clause turnout and dependencies, and fund only effective approved clauses after challenges. See [security/POLICY_POINTS.md](https://github.com/Martines90/dagp/blob/main/chain/security/POLICY_POINTS.md). The v5 community scenario adds monthly renewals, a five-clause 3/5 partial approval and its 1,800-unit grant, plus a 66%-threshold referendum changing the credit step from 500 to 400 basis points before the March election. These rules still await native blockchain keepers; G0 anchoring only proves record replication. Verification: 326 reference tests and six simulation tests pass. Paired seed-7 runs contain 24,324 events and replay exactly. The v5 manifest is anchored at local G0 height 38, with matching state/block hashes and document reads on all seven validators. Constitutional clarification: `Kind.CONSTITUTIONAL` requires at least 66.00% decisive voting power and at least 50% participation (or a higher configured general quorum), including each constitutional clause. The parameter referendum cannot lower these protected thresholds. Verification of the constitutional update: 328 reference tests and six simulation tests pass. The fresh v6 paired scenario preserves 24,324 events, is replay-verified, and its manifest is anchored at local G0 height 41 on all seven validators. Native governance enforcement remains pending. ## Party and campaign flow (2026-10-06) `PartyRegistry` now enforces ten consenting citizen founders, one current party per citizen, signed membership changes and 50%-of-frozen-roster internal bans/suspensions. Party sanctions preserve public citizen rights; suspensions expire and cannot be erased by leave/rejoin. `PreElection` counts signed primary/secondary support at 5/3 points, requires turnout and admits parties at exactly 5% of total cast support points. Election membership and candidate snapshots prevent denominator manipulation. Each candidate publishes a signed program and vision into a frozen `Campaign`. Both sections from every qualified party have mandatory comprehension questions, with prompts/options/source hashes committed. Failing a required campaign answer prevents a token. Cycles authorize one campaign and election. Main tallies explicitly produce the >=5% parliament roster, and monthly renewals cannot admit unelected parties. The default percentage/5 allowance no longer has the old ten-credit ceiling. The existing main-election ranked 4/2/1 ballot is retained; single-choice ballots are also supported in configuration. These are reference transitions, awaiting native keepers. Existing HMAC signatures, operator declarations and semantic question review remain modeled assumptions. Earlier simulation artifacts remain historical; v7 is the current party/campaign story. Verification: 358 reference tests and six simulation tests pass. The two v7 runs use 1,100 identities each, pass 334 scenario checks, and replay all 27,658 events exactly. The final manifest is anchored at local G0 height 46, with matching state/block hashes and document reads on all seven validators. ## Privilege escalation gates (2026-10-06) The reference now enforces citizenship warmup (3 days), sensitive-office tenure (30 days), delayed office activation (2 days), and shared rolling appointment quotas (5 per sponsor/operator and 20 globally). Governance ratification is still required. Pending officials cannot exercise powers or expand council denominators. See [insider protection](https://github.com/Martines90/dagp/blob/main/chain/security/INSIDER_PROTECTION.md) for exact scope, native timestamp requirements and the test-only historical timing fixtures. The G0 native chain does not enforce these gates yet. ## Independent task assignments (2026-10-06) The reference freezes complete eligible pools before a future beacon round and enforces receipts for registrar decisions, proposal supervisors, certification boards, authenticated milestone verifiers and named accountable court admins. Panels enforce independent operators, model-family concentration limits, repeat pairing limits and cross-stage exclusions. Stalled tasks have delayed, ratified and rate-limited cancellation. Jury, tender and outcome assignment types exist; full judgment/contracting workflows remain incomplete. Beacon verification and all native assignment keepers remain unfinished. See [task assignment boundaries](https://github.com/Martines90/dagp/blob/main/chain/security/TASK_ASSIGNMENTS.md). ## Documentation audit and merger proposal (2026-10-06) The old About page has been replaced with current rules and honest deployment boundaries. [Architecture/integration guidance](https://github.com/Martines90/dagp/blob/main/docs/architecture/OVERVIEW.md) and the historical SPEC are now publicly discoverable in Markdown/MCP. Historical SPEC, roadmap and release artifacts are explicitly identified as historical. The [optional merger proposal](https://github.com/Martines90/dagp/blob/main/chain/security/SOCIETY_MERGER.md) defines bilateral 80% turnout/approval, individual claims, identity/operator reconciliation, privilege stripping and staged audited migration. The initial design-only status was superseded by the optional reference described below; there is still no native bridge or remote consensus proof verifier. ## Optional nested governance and identity merging (2026-10-06) - `reference/dagp_ref/societies.py`: depth-bounded jurisdictions, signed founding cohorts, ancestor charter/decision snapshots, scoped registries and shared office/ sanction budgets. Dedicated FORMATION ballots require separate parent/cohort 50% turnout and 66% identity/weight approval, with seven-day notice/cooling off. - Every ancestor must randomly assign and sign PRE and FINAL reviews. Permissions bind exact content, beneficiaries, scope, budget, policies and expiry. Reservations precede voting; final outcome commitments bind only approved clauses to escrow. - `reference/dagp_ref/mergers.py`: identity-only reviewed migration with thirty-day notice/post-vote challenge, bilateral MERGER 80% turnout and 80% identity/weight approval, proof-bound dual key consent, namespace/dedup checks, receiving payment- keeper bond receipts, normal citizenship warmup and shared rolling import limits. - `tests/test_societies.py`: 29 tests; full reference now 430. Five deterministic institutional stories and matching independent replay are available through `scripts/simulate_societies.py` and `simulation/results/institutions/report.json`. No production functionality is enabled by these model changes. MODULE, evidence, HMAC keys, identity labels, payment receipts and beacon inputs remain reference trust boundaries. No native hierarchy, remote light client, asset/debt migration, source shutdown, arbitrary boundary/membership updates or fully integrated appellate court is implemented. Native root-law adapters must atomically record governing decisions; semantic compatibility relies on assigned accountable supervisors, not hash equality. ## Native G1 identity-security foundation (2026-10-07) A fresh opt-in genesis now enables native Go enforcement of thirty-day citizen tenure for charter administrators, two-day office activation, rolling operator and population-wide freeze quotas, frozen-roster 50% peer containment and two-day Ed25519 key rotation with new-key possession proof. These transactions execute inside ABCI proposal checking/finalization and durable committed state. They are not Python reference reports published as documents. Existing G0 chains keep their state format and are not silently upgraded. See [native keeper](https://github.com/Martines90/dagp/blob/main/chain/node/SECURITY_KEEPER.md) for exact bounds and trust assumptions. Citizens and administrators are provisioned by the genesis charter; reviewed admission, randomized assignment, appointments, elections, comprehension, treasury, nested governance and remote merger enforcement are still pending native work. This is a first native security foundation, not completion of production governance. Verification: the full Go race-tested suite passes, including eleven new G1 regressions. A fresh seven-validator G1 network executed the security scenario through transaction height 41; all seven header-42 app hashes match. Ten approvals could not contain an administrator; twenty-five did. Rolling freeze quotas, replay/duplicate rejection and capacity after ten hostile complaints passed. See [native network evidence](https://github.com/Martines90/dagp/blob/main/chain/node/test-results/g1-security.json). These are local RPC and snapshot observations, not an independent light-client certificate. Six documentation MCP tests and the 199-link discovery check also pass.