Status sync: ROADMAP vs BACK_LOG vs GitHub issues/PRs (Sep 11, 2026)

[umermjd11]

Why this exists

ROADMAP.md had drifted from develop: its most recent status block (Aug 5) still framed P3 as "eight PRs open, merge sequencing is the only blocker." All eight merged weeks ago, plus a dozen follow-on PRs closing out the backlog items their reviews raised. I've just updated ROADMAP.md in place to match current reality (diff: work-package statuses, a CI/CD row and a fair-launch row that existed in code but never in the table, and a refreshed top-of-doc status block). This discussion is the narrative version — what's done, by which PR, what's genuinely still open, and how that lines up against BACK_LOG.md and the live issue/PR state — posted ahead of the Sep-14 testnet-readiness deep dive so that call starts from accurate ground truth.

P3: what's actually merged to develop

The eight core mechanism PRs named in the Aug 5 update are all merged: #51 fee router, #60 tokenomics (mint cap, faucet retirement), #62 scoring median fix, #63 auditor commit-reveal, #64 reward engine, #65 staking + jailing, #66 slashing taxonomy + invariants + dispute scaffold, #68 backlog tracking.

Reviewing those eight generated BL-8 through BL-15 in the backlog, and those are now mostly resolved too: #110 (BL-13, address-keyed encrypted test-data keys), #116 (BL-8, import-deployments six-proxy fix), #117 (slashing-invariants burn/treasury realignment), #129 (BL-1, nonce pending-tag fix), #131/#134/#136 (BL-10/BL-14, claim-pull O(1) reward settlement + gas benchmark), #133 (BL-12, dispute-bond denomination).

P3: what's genuinely still open (not just "PR pending")

Don't read "merged" above as "closed" — issues #37, #38, #39, #41, #42, #43 all stay open for real remaining slices:

• Partial-vs-full slash penalty fraction — still an open DAO-settable number (MECHANISM_DESIGN §9 item 2).
• BL-11 — dispute-redo subgroup selection still shuffles via grindable blockhash, reusing security-review finding H-2. The model owner is both the grinding party and the sole adjudicator, with no deadline. Unaddressed. Worth putting on the table for the security-bar question in Testnet readiness — what needs to be in place before we ship? (prep for week-of-14th deep dive) #132.
• Validator-income simulation across participation scenarios (P3-5.2) hasn't been run.
• Scoring issue P3 scoring: BlockFLow validation (P3-SCR), on-chain median, scenario-driven scoring redesign #39's "scenario-driven scoring redesign" — the median fix landed, the redesign hasn't started.

Three open PRs are actively continuing this work right now, none reviewed yet: #140 (issue #37 — per-model stake floor + concurrent-registration cap), #141 (issue #43 — dispute-bond + reward-pool treasury forwarding), #142 (issue #78 — commitAuditScore gas measurement + networkFeeFloor). All three pass CI.

P4: din-indexer / din-sdk / din-daemon — real, but nothing's merged yet

Covered in full in discussion #139; summary for this cross-reference: none of #20 (SDK), #21 (daemon), #24 (indexer) have merged anything to develop — no dincli/sdk/, no dincli/dind/, no subgraph/ exist there today. PRs #29+#72 (indexer) and #31+#32 (SDK + daemon, both draft) are substantially built and mostly well-tested, but both integration branches are 187 commits stale against develop. The fixes needed to merge are small (a GraphQL schema typo, a non-CliRunner test file, one wallet-test isolation gap). The one real "not started, not just unmerged" gap: the task-level/platform-level contract events the indexer depends on (LocalModelSubmitted, T1AggregationSubmitted, etc.) don't exist anywhere in foundry/src — audited and agreed in July, no tracking PR since.

Practical implication for #132: since none of this is merged, dind role-automation isn't a testnet dependency today — it's P4 scope running in parallel, not a blocker.

Two things that existed in code but weren't anywhere in ROADMAP.md until this update

• CI/CD Phase 1 — proposed in discussion #74, shipped as #103 + #118, both merged. "Phase 1" implies more phases are expected; none are scoped yet.
• Fair-launch distributor contract slice (BL-18) — merged as #77. Issue #75 stays open — eligibility computation is deliberately deferred until the indexer exists.

din-dao — unchanged, still deferred

No change since Abraham's Aug 4 decision: §6 of ROADMAP.md, tracked as BL-16. PR #27 stays open, unmerged, out of active scope.

Backlog reconciliation

BACK_LOG.md is current as of commit 3bde8ee (Sep 10) — of BL-1 through BL-24: 7 are ✅ Resolved with a merged PR each (BL-1, BL-8, BL-10, BL-12, BL-13, BL-14, and BL-18's contract slice). Genuinely open with no PR yet: BL-11 (flagged above — arguably the most testnet-relevant one), BL-19 through BL-22 (daemon hardening — PID-file race, /health can't signal degraded, no log persistence, unbounded nonce-manager cache; all inside the still-unmerged #31/#32), BL-23/BL-24 (DP calibration questions needing a specialist opinion), and the idea-stage items (BL-3 through BL-7, BL-16, BL-17) that haven't been triaged into a phase.

Open vs. closed issues, at a glance

Open: #20 (SDK), #21 (daemon), #23 (DAO, deferred), #24 (indexer), #36 (P3 umbrella), #37/#38/#39/#41/#42/#43 (staking/slashing/scoring/rewards/tokenomics/fees — real remaining slices, not stale), #75 (fair launch, pending indexer), #78 (gas simulation, PR #142 open against it).

Recently closed (Sep 8–10): #40 (auditing), #99 (DP follow-ups F1/F3/F5–F8 — note F4 was explicitly excluded from scope and is not tracked anywhere now that #99 is closed, see BL-24), #114/#115/#120/#121/#122/#123/#124/#125/#126/#127 (the backlog-resolution wave above, plus its own test rewrite).

For the Sep 14 call

A few things from the above worth having on the table in #132:

1. BL-11 (grindable dispute-shuffle) is the one open security-relevant gap sitting directly in the slashing/dispute path — worth an explicit call on whether it's a testnet blocker or an accepted scaffold-stage risk.
2. P4 (indexer/SDK/daemon) isn't merged anywhere, so it's fair to scope testnet without it — narrows what "done" needs to mean for the Sep 14 checklist.
3. Three P3 PRs (feat: per-model stake floor + concurrent-registration cap (closes #37) #140/feat: wire dispute-bond and reward-pool treasury forwarding (closes #43) #141/feat: commitAuditScore gas measurement + networkFeeFloor parameter (closes #78) #142) are open and unreviewed — worth a pass before the call so the checklist reflects where they land, not where they were when opened.

---

Full detail: ROADMAP.md (updated) · BACK_LOG.md · discussion #139 (P4 merge detail) · discussion #132 (testnet readiness)

cc @abrahamnash @robertocarlous @Santiagocetran @Abidoyesimze

[abrahamnash]

Thank you, @umermjd11. @robertocarlous @Santiagocetran @Abidoyesimze — if you have any additional thoughts or see any gaps in the above that are worth surfacing ahead of the Sep 18 discussion, please share them here and we’ll combine them with discussion #132

[umermjd11]

Progress check for the Sep 14 call: are we behind?

Short version: not behind on engineering output — behind on integration.

P3 (scheduled May 1–Aug 30, now tracked by WP completion rather than calendar week): core mechanism work is essentially done and merged — staking, slashing engine, fee router, tokenomics, reward engine, scoring fix, dispute scaffold, CI/CD Phase 1 — all 8 original PRs plus 6 of the BL-10–15 follow-ups are on develop. Two things genuinely haven't moved:

• P3-6.2 (adversarial/collusion/Sybil/stress testing) is still 📋 Planned — the one core WP row that hasn't started, and it's the gate before any external audit.
• Testnet exit criteria still aren't written down — that's exactly what #132 exists to fix.

P4 (scheduled Sep 1–Nov 30 — we're in week 2 of 13): paradoxically ahead on raw code. Daemon core, SDK extraction, preferences, and capability detection (248 passing tests) plus the indexer design+implementation (406 SDK tests, 34-entity/31-handler schema) are all substantially built. But none of it has merged to develop — feat/din-sdk and feat/din-indexer are both 187 commits stale, each with a small, already-identified blocker sitting unaddressed (indexer: a GraphQL schema typo + a broken test file; SDK: a test-isolation gap patching the wrong module). The one structural gap with no owner at all: the indexer's dynamic task-contract events (LocalModelSubmitted, T1AggregationSubmitted, etc.) were agreed in July and still don't exist anywhere in foundry/src.

Net: the team hasn't fallen behind writing code — if anything P4 got a head start while P3 mechanism design was catching up. The real schedule risk is merge debt on two large, staling integration branches, plus P3-6.2 stress testing sitting untouched. Adding this to the three items already flagged above for Sep 14.