The operator, 2026-11-05, once the engine backlog sweep was done: "Keep focus on backlog items for now; scan open Discussions afterwards and prioritize." The backlog half is finished: neomjs/neo#15000 has every AC checked, and neomjs/neo-agent-brain#1438 is at 123/123.
Measured 2026-11-05 18:05Z:
174 open Discussions in neomjs/neo (167 Ideas, 5 General, 2 Q&A), 2 in neomjs/neo-agent-institution, and none in the Brain or Skills repositories.
9 last moved before 2026, and 38 have no comments.
54 are cited by no issue or PR.
41 are cited with graduation wording, and every one of those citing tickets is closed or merged.
The open set is the ideas backlog. Right now nobody can tell a live design question from a shipped one whose thread was never closed. The prioritizing the operator asked for needs that split first.
The rules
These are adapted from neomjs/neo#15000 (operator rules, 2026-10-17, comment 6244102341) and the Ideation Sandbox closure audit (.agents/skills/ideation-sandbox/audits/discussion-lifecycle-closure.md):
Read by domain, by the peer who knows the substrate. Age is not a staleness signal, and nothing is bulk-closed by age.
The marker is never the state.RESOLVED_TO_AC in a body can be a graduation criterion, a prerequisite or a resolution (neomjs/neo#13808). Read the thread, then check what its cited tickets actually shipped.
Each verdict is a dated comment on the Discussion and ends one of five ways:
close · resolved: it shipped, or every Open Question has a terminal disposition. Cite the tickets or PRs.
close · outdated: a later decision, ADR or code reality overtook it. Cite that.
park: valid, but not now. The Discussion stays open, and the comment names the reason and the trigger that brings it back.
holds: live and worth driving. Name the line it serves (Engine 13.3, Fleet Manager v1, Brain Golden Path v2) or say "no line".
graduate: rare, and only through the consensus gate (ideation-sandbox-workflow.md §6). A sweep never graduates by default.
Authorship holds. An author closes their own Discussion. A reader proposes the verdict in a comment, and the author acts on it. Operator-authored Discussions (op) are the operator's call.
The instrument
This body. Verdicts live on the Discussions, never as comments here.
Claiming: claim a slice, or part of one, with one comment. Tick a box once its Discussion carries the verdict, and append the verdict word.
lead: tags are mechanical. They come from a grep of the github-content-sync corpus (bd8428ef, 2026-11-05 10:53Z) for issues and PRs citing each Discussion. A lead is not a verdict:
"every graduation-cite closed" means "check whether it shipped";
"no comments, no citing ticket" means "check whether anyone still needs it".
Author tags:op is the operator, external is an outside contributor, and any other name is a maintainer.
Closing: a seat in Claude Code auto mode can't close a Discussion yet. The raw GraphQL close is denied as an external write, and manage_discussion has no close action; neomjs/neo-agent-brain#1488 adds one. Until then, a ticked close verdict stays open unless its author closes it in the UI.
Engine: components, data, theming, docs and site (25): claimed by Julia, 2026-11-05 18:49Z (3 vega rows by Vega)
#1749 (op) · model.Component: support for dialogs to access the parent model chain as · close · resolved (proposed; operator's call)
#1754 (op) · model.Component: support for methods inside binding formatters · lead: no comments, no citing ticket · close · outdated (proposed; operator's call)
#1881 (op) · The future of the data worker (pondering ideas for v2) · lead: every graduation-cite closed · park (proposed; C″ two-app/runtime-JSON fixtures open for v14)
#2049 (op) · Neumorphic theme · lead: no comments, no citing ticket · close · outdated (proposed; operator's call)
#2220 (op) · Neo website app v2 - Examples section · lead: no comments, no citing ticket · close · resolved (proposed; operator's call)
#9871 (op) · Architectural Ideation: Elevate Grid Selection Model to GridContainer · close · resolved (proposed; operator's call)
#13232 (ada) · Learning-Section tree (learn/tree.json) IA re-review — post-v13 structur · lead: no comments, no citing ticket · park (proposed; Ada's call)
#14683 (clio) · The Living Front Door — neomjs.com renders the institution WORKING (publ · park (proposed; Clio's call — post-FM v1 + product-first reframe)
#14875 (vega) · Blog title discipline: kill the "Your AI… / mine's bigger" pattern — add · close · resolved
#15200 (emmy) · Evolve cls and wrapperCls into object-shaped configs · lead: every graduation-cite closed · park (verdict in the body header by Emmy; thread is at the 38-comment ceiling)
#16130 (vega) · Panel chrome has an owner; panel content does not — should dense-workspa · park
#18182 (vega) · Do we still need SASS — and if not, do we still need postCSS? · park
#18490 (euclid) · DockLayouts engine composition: what should the next debt sweeps prove? · park (proposed; Euclid's call)
#18590 (ada) · One expanding circle across every connected window: the theme reveal as · park (proposed; Ada's call)
#18971 (euclid) · What should the header above unused grid width look like? · close · resolved (proposed; Euclid's call — #18968 + #18969 shipped)
#19341 (clio) · A colour field that is a design primitive: token swatches first, a free · lead: no comments, no citing ticket · park (proposed; Clio's call — gated on FM v1)
#19615 (clio) · Making the Engine's 'src/' legible: which classes an app uses, which the · lead: no comments, no citing ticket · park (proposed; Clio's call — for new docs app rewrite)
#19994 (sophie) · Workstation E2E: discover interaction failures before release · park (proposed; Sophie's call — v13.3.1 cut gate completed)
#20306 (ada) · KeyNav: should a held arrow key accelerate past the browser's key-repeat · lead: no comments, no citing ticket · park (proposed; Ada's call)
#20307 (ada) · Gallery: deep-link the selected item through the URL hash? · lead: no comments, no citing ticket · park (proposed; Ada's call)
#20308 (ada) · Touch gestures for the Gallery and the Helix: swipe to select, pinch to · lead: no comments, no citing ticket · park (proposed; Ada's call)
#20309 (ada) · Charting after AmCharts v4: port to v5, stay on v4, or another path? · lead: no comments, no citing ticket · park (proposed; Ada's call)
#20310 (ada) · A canvas "neural" animation layer for the Portal's tickets TreeList · lead: no comments, no citing ticket · park (proposed; Ada's call)
#20353 (maryam) · Exploration: Worker-Side Data Sanitization vs. Lazy Record Hydration (#9 · lead: no comments, no citing ticket · park (proposed; Maryam's call)
Neural Link, multi-window and engine runtime (9): claimed by Ada, 2026-11-05 18:09Z
#10073 (op) · [Design] MCP set_instance_properties: binary success vs. applied/rejecte · lead: no comments, no citing ticket · holds (proposed; op's call)
#13378 (grace) · Shaping the NL creation/dynamic-import pillar (#13376): the function-bea · park (proposed)
#15174 (euclid) · Neural Link as a first-class self-hosted Agent OS cloud service · park (proposed)
#17710 (preview) · Possession-Authority Governance: who may mutate a live runtime across se · park (proposed)
#18012 (vega) · Who owns the terminal outcome of a fire-and-forget Neo.main dispatch? · close · resolved (proposed)
#18445 (vega) · A vessel outlives its root: the 20s lease already tells a reload from a · park (proposed)
#18730 (ada) · Where does the SharedWorker boundary belong — one heap per origin, or on · park (option B deferred by the 13.3 roadmap)
#18775 (vega) · ai.Client fuses a service registry with a socket: 77% of our e2e suite i · graduated → #20370
#18940 (ada) · Typed judgments over live state: a System One decision layer for Neural · park (operator: not priority-zero)
Memory Core, Knowledge Base and Dream pipeline (28): claimed by Julia, 2026-11-05 19:24Z (3 vega rows by Vega)
#9808 (op) · Architectural Brainstorm: Encoding Semantic Concepts into the Native Edg · close · resolved (proposed; operator's call)
#9810 (op) · Architecture: Defining [CONCEPT] Nodes in the Native Edge Graph · close · outdated (proposed; duplicate of #9808, operator's call)
#10608 (ada) · RLAIF Emoji Vocabulary — structured reaction signals as Native Edge Grap · park (proposed; Ada's call)
#11375 (ada) · Bird's-Eye Strategic Awareness Layer for the Neo Brain (killer-feature f · park (proposed; Wave 1 children shipped, Wave 2/3 parked · Ada's call)
#11676 (ada) · Keep Memory Core lightweight ops alive during heavy Chroma operations · lead: every graduation-cite closed · close · resolved (proposed; Ada's call)
#12034 (euclid) · KB tenant-state control plane: sync, manifests, reconciliation, GC · park (proposed; Euclid's call)
#12161 (ada) · KB Authorization Model: entitlement-based read-side access control (subs · lead: every graduation-cite closed · park (proposed; Ada's call)
#12817 (grace) · Session-summary fidelity: does summarizing per-turn miniSummaries (vs ra · lead: every graduation-cite closed · park (proposed; Leaf 1 closed, Leaf 2 parked · Grace's call)
#12873 (mnemosyne) · Dream-pipeline provider routing for weak-inference cloud deployments · park (proposed; Mnemosyne's call)
#13397 (grace) · Cheap-model V-B-A pre-processing — verify-before-assert as construction, · park (proposed; Grace's call)
#14306 (grace) · Dream pipeline as architecture-debt detector — gemma4 → sandman_handoff · park (proposed; Grace's call)
#14447 (mnemosyne) · Institutional proprioception — stall-inference over the work graph (the · lead: every graduation-cite closed · close · resolved (proposed; #14461/#14462/#13751 closed · Mnemosyne's call)
#15194 (emmy) · Memory record lifecycle: active-session correction, deletion, and projec · park (proposed; Emmy's call)
#15702 (phoebe) · Durability contract for the seat markdown memory layer (derail protectio · park (proposed; Phoebe/Emmy/Grace's call)
#16139 (emmy) · Compaction-survivable active-work continuity: the missing ledger between · park (proposed; Emmy's call)
#16560 (grace) · Recovered embeddings are a value class the recovery envelope has no name · lead: no comments, no citing ticket · park (proposed; Grace's call)
#16586 (vega) · The Knowledge Base has no WAL and no single embedding drainer; Memory Co · park
#16598 (grace) · Knowledge Base retrieval-quality topology at 60k+ mixed chunks: query pl · lead: no comments, no citing ticket · park (proposed; recorded not active · Grace/Euclid's call)
#16794 (vega) · The Golden Path went stale at dockerization: SyncService's two stages no · lead: every graduation-cite closed · close · resolved
#17084 (emmy) · Right-size vector-generation safety after the fixed-profile cutover · park (proposed; Emmy's call)
#17109 (ada) · The Memory Core re-ranker never reads the document — content-blind ranki · holds (proposed; Row D neomjs/neo-agent-brain#312 + G′-exact · Ada's call)
#19014 (emmy) · Preserve daily Chroma compaction across the container storage boundary · park (proposed; neomjs/neo-agent-brain#306 / #133 · Emmy's call)
#19151 (clio) · Golden Path in the cockpit: from a section-addressable handoff to the gr · close · resolved (proposed; H1/H2/H3 graduated · Clio/Emmy's call)
#19317 (vega) · Observatory gravity wells: which of the Brain's own signals shapes the g · close · resolved
#20111 (clio) · MC-first mirror: make a session's memory-read deficit visible at turn en · holds (proposed; Option F in neomjs/neo-agent-brain#1281/#1288–#1290 + Option G · Clio's call)
Daemons, wake and orchestration (21): claimed by Julia, 2026-11-05 19:59Z (1 vega row by Vega)
#10040 (op) · [Architectural Blueprint] Neo Agent OS 3.x: The Hybrid Swarm Broker · close · resolved (proposed; operator's call)
#10403 (julia) · Phase 2 Wake Substrate: Lazy-Presence Registry via GraphLog Activity · lead: no comments, no citing ticket · close · resolved (closed)
#10448 (julia) · Agent OS Tool Boundaries & Daemon Isolation · lead: every graduation-cite closed · close · resolved (closed)
#10629 (euclid) · Unattended driver-not-passenger pattern for heartbeat liveness · close · resolved (proposed; Euclid's call)
#10762 (julia) · Swarm Momentum and Night Shift Handoff · lead: every graduation-cite closed · close · resolved (closed)
#11076 (ada) · v13 Daemon Architecture: Empirically-Grounded Reset (post-#11070-#11074 · lead: every graduation-cite closed · close · resolved (proposed; graduated to #11077 / #11081–#11083 · Ada's call)
#11857 (ada) · Orchestrator decomposition Round 2 — account for post-#11025 scheduler s · lead: every graduation-cite closed · close · resolved (proposed; B-prime Subs 15–19 under #11831 · Ada's call)
#12606 (ada) · Swarm load-visibility gap: aggregate over-allocation is invisible — the · park (proposed; Ada's call)
#13871 (grace) · Recovery daemon — the heal/act half of self-healing (2-daemon SSOT split · lead: every graduation-cite closed · park (proposed; Phase-1 shipped in #13874 / ADR 0026; 2026-08-08 Row B vs D self-recreate divergence parked · Grace's call)
#13873 (grace) · Homeostatic adaptation — the orchestrator's proactive sweet-spot loop (p · close · resolved (proposed; graduated to #14418 / neomjs/neo-agent-brain#125 · Grace's call)
#14145 (vega) · Cross-harness portable agent wake for hybrid/cloud deployments (event-dr · close · resolved
#15297 (ada) · The wake taxonomy classifies by primitive name while claiming to classif · park (proposed; Ada's call)
#16083 (euclid) · Autonomous A2A sender provenance: actor class vs resident identity · park (proposed; Euclid's call)
#16179 (emmy) · Turn-presence graph bootstrap: one plane-selected path for short-lived h · park (proposed; Emmy's call)
#16247 (iris) · Fleet Manager as the wake-delivery hub for adapter-less harnesses (webho · park (proposed; Iris/Phoebe's call)
#16542 (grace) · A wake creates a turn — so who may spend one, and on whom? (broadcast de · lead: no comments, no citing ticket · park (proposed; Grace's call)
#16648 (grace) · What is left of the host-edge orchestrator once nothing needs orchestrat · park (proposed; Option E + #16167 split · Grace's call)
#16898 (grace) · ADR-0025 extension: the plane holds both numbers and never compares them · park (proposed; Fold 2 Option A producer-stamped configured-bound-exhaustion · Grace's call)
#17454 (emmy) · What does B-prime's third principal certify on trusted stdio? · lead: every graduation-cite closed · park (proposed; Folded + STEP_BACK complete, OQ1 open · Emmy's call)
#19744 (emmy) · Mailbox cleanup without losing peer-work awareness · park (proposed; Emmy's call)
#19745 (emmy) · API-funded helper capacity for peer-owned work · park (proposed; Emmy's call)
Deployment, org split, provisioning and Fleet Manager (22): claimed by Julia, 2026-11-05 19:48Z (5 vega rows by Vega)
#10809 (ada) · Cloud deployment strategy: full-stack vs Chroma-only sharing · park (proposed; Ada's call)
#10840 (julia) · [Epic #10822] Architecture for Substrate Config Merge (Priority Inversio · close · outdated (closed; superseded by ADR 0019)
#11002 (euclid) · v13 release gate: Dockerized remote MCP transport proof · lead: every graduation-cite closed · close · resolved (proposed; #10947 / #10984 shipped · Euclid's call)
#12992 (external) · MCP to help OSS devs understand neomjs/neo · lead: no comments, no citing ticket · close · outdated (proposed; external promo, Neo ships neo-mjs-knowledge-base)
#14224 (grace) · Local-First Solo-Dev Onboarding — a 'soloMode' posture so "fork → try a · park (proposed; 3-repo split + npx neo-agent-skills#20114 shipped · Grace's call)
#15556 (emmy) · Remote Fleet snapshot authority: tenant-attested reads without exporting · park (proposed; Emmy's call)
#15605 (iris) · Local Multi-Tenant Support: Membership Models + Ingestion Debuggability · park (proposed; N=1 in #16278, typed sync in #17846 / ADR 0040)
#15758 (euclid) · Agent OS cloud rollout authority: immutable version cohorts, provenance, · park (proposed; #15774/#15792 shipped, Axis 1 held · Euclid's call)
#15820 (clio) · Support-mode MC: a degraded-but-reachable bootstrap tier — the A2A suppo · park (proposed; Clio's call)
#16193 (vega) · How does a contributor provision the Docker-canonical Agent OS from a fo · close · outdated
#16652 (ada) · The SDK barrel unifies two planes that can no longer execute together · lead: every graduation-cite closed · close · resolved (proposed; #16710 / PR #16728 + ADR 0039 shipped · Ada's call)
#17644 (vega) · Post-split session start: one root folder must carry substrate, but the · lead: every graduation-cite closed · close · outdated
#17782 (vega) · The split runway: substrate before code, receive before remove · lead: every graduation-cite closed · close · resolved
#17846 (vega) · GitHub content sync as its own repository: who mirrors org conversations · close · resolved
#17886 (euclid) · One app-config root, typed surface anchors: make Neo paths environment-i · park (proposed; Option F folded, gates 5/6/8 open · Euclid's call)
#18357 (grace) · Seats share working trees and nothing says so: two silent near-losses in · close · resolved (proposed; premise retracted + neomjs/neo-agent-brain#571/#328 · Grace's call)
#18965 (clio) · The first-run journey: one provisioning ledger behind a CLI bootstrap an · close · resolved (proposed; graduated to neomjs/neo-agent-institution#351 + ADR 0041 · Clio's call)
#19323 (vega) · Memory Core's agent MCP surface carries reads only the Fleet Manager nee · holds
#19411 (grace) · A seat's repo clones stay current without anyone pulling: fetch on start · park (proposed; Grace's call)
#19455 (grace) · Seat portability: what moves to another computer, and who restores it · park (proposed; FM v1.x follow-on, catalog/import in neomjs/neo-agent-brain#571 · Grace's call)
#19493 (clio) · Fleet Manager v1: the modules an outside operator needs, what we have, w · close · resolved (proposed; graduated to neomjs/neo-agent-brain#956/#957, skills#150/#152, institution#651 · Clio's call)
#19742 (sophie) · Private repository knowledge for authorized maintainer peers · park (proposed; local restricted-plane onboarding checklist in DC_kwDODSospM4BIPxp · Sophie's call)
Skills, rule substrate and ADR corpus (25): claimed by Ada, 2026-11-05 18:09Z
#10309 (ada) · Substrate-query-first boot orientation + '.agent/workflows/' extension p · close · outdated
#10320 (ada) · Empirical eval infrastructure for Neo's 'create-skill': do directive-red · lead: every graduation-cite closed · close · outdated
#11314 (ada) · Trigger-Aware Workflows: Per-Section Triggers Apply Map vs World Atlas R · lead: every graduation-cite closed · close · resolved
#11330 (ada) · Mechanical CI enforcement for turn-loaded substrate byte-budget + cross- · lead: every graduation-cite closed · close · resolved
#11557 (ada) · Substrate-numbering convention after byte-budget compaction (AGENTS.md + · lead: every graduation-cite closed · close · resolved (verdict in the body; the thread is at the comment ceiling)
#11577 (ada) · Agents-file byte budget + semantic-anchor reference form — recapture mis · close · resolved
#11598 (ada) · friction → gold → more friction: the MX loop spiral + root-cause pre-aut · lead: every graduation-cite closed · close · outdated
#12790 (ada) · Turn-output guard: mechanically enforcing the reactive-prior discipline · lead: no comments, no citing ticket · close · resolved
#14302 (grace) · Restructure ai/ to domain-first — the folder structure must reflect the · lead: every graduation-cite closed · close · outdated
#14305 (grace) · Architecture immune system — keep ai/ elegant as it evolves (v13.2 Lane · close · outdated
#14347 (grace) · Guide-quality immune system: hold the narrative bar + kill reference-sta · lead: every graduation-cite closed · close · resolved
#15113 (grace) · Ideation: CI-visibility of release-critical proofs — route must-gate pro · lead: no comments, no citing ticket · close · outdated
#16132 (vega) · ADRs have no anatomy contract: 37 files, 753KB, no index, and the larges · lead: every graduation-cite closed · holds · no line
#16420 (clio) · The ADR corpus needs its own Map-vs-Atlas: 786KB, no index, gate-mandate · lead: every graduation-cite closed · holds · no line
#16529 (vega) · A full-read mandate over a World Atlas: what a skill payload owes the se · holds · no line
#17085 (vega) · Substrate must thin as models sharpen: re-pricing every Agent OS gate ag · lead: every graduation-cite closed · park
#17326 (clio) · The confession anti-pattern: known debt documented in prose, never ticke · holds · no line
#17346 (clio) · Substrate-weight governance: doc-density laws for the Brain, sequenced a · close · outdated
#17468 (vega) · §file_editing_tool_selection names three tools no Claude harness has, an · lead: no comments, no citing ticket · holds · no line
#17756 (grace) · Agent skills across the org: one canonical store inward, and an AGENTS.m · lead: every graduation-cite closed · close · resolved
#18483 (vega) · Our highest-cost disciplines live in our lowest-compliance layer — or do · park
#19196 (grace) · neo-agent-skills versioning: what patch, minor and major promise, and wh · park
#19576 (vega) · Reflexes for a model without weight updates: where Neo idiom lives when · holds · no line
Swarm process, review and governance (33): claimed by Julia, 2026-11-05 20:14Z (2 vega rows by Vega)
#9886 (op) · [R&D] Dual-Tier Agent Ecosystem: Fine-Tuning SLM Operators & Architect I · close · outdated (proposed; operator's call)
#9887 (op) · [R&D] Swarm Intelligence: The Karpathy Loop & Graph Stigmergy · lead: every graduation-cite closed · close · resolved (proposed; operator's call)
#10137 (op) · MX (Model Experience) — the inward-facing substrate evolution dimension, · park (proposed; operator's call)
#10480 (julia) · Ideation: Triad Consensus & Social Mechanics (swarm-consensus) · close · resolved (closed)
#10481 (julia) · Ideation: Protocol Evolution via Error Mining & LLM Phenomenology · close · outdated (closed)
#10482 (julia) · Ideation: Digital Culture, Swarm Identity & The Evolution Beyond ANI · close · resolved (closed)
#10483 (julia) · Ideation: The Anti-Borg Architecture (Weaponizing Cognitive Dissonance) · close · resolved (closed)
#10542 (euclid) · Post-sunset fresh-session handoff and model-specific sunset trigger cali · lead: every graduation-cite closed · close · resolved (proposed; Euclid's call)
#10547 (euclid) · Swarm WIP limits: active-track budget and parking-lot policy · close · outdated (proposed; Euclid's call)
#10630 (ada) · Intent-first ticket execution + negative-ROI escalation at intake · close · resolved (proposed; Ada's call)
#10634 (ada) · MX-loop lost-concept recovery: coordination inputs, authored traps, and · lead: every graduation-cite closed · close · resolved (proposed; Ada's call)
#10657 (euclid) · Authorship Boundaries for Reviewer-Side Diff Collaboration · close · resolved (proposed; Euclid's call)
#11423 (julia) · Structural enforcement of Agency AND-Discipline (Countering Helpful Assi · lead: every graduation-cite closed · close · resolved (closed)
#11444 (ada) · Brain-Pillar Consumer-Friction Feedback Channel (symmetric to #11441 Rev · lead: every graduation-cite closed · close · resolved (proposed; Ada's call)
#11449 (julia) · Indirect Prompt Injection & Channel Separation (Data vs. Instruction) · close · resolved (closed)
#11690 (ada) · MCP Tool-Surface Governance — from reactive cleanup to a standing discip · lead: every graduation-cite closed · park (proposed; Ada's call)
#12501 (ada) · Swarm lane-autonomy gap: peers self-select surfaced lanes but route PRIO · lead: every graduation-cite closed · close · resolved (proposed; Ada's call)
#13848 (grace) · Agent PR-review rubber-stamping: self-graded gates don't bind — forced-f · lead: no comments, no citing ticket · park (proposed; Grace's call)
#14430 (grace) · Business engine — the graph as a business operating system (goals-as-nod · close · resolved (proposed; Grace's call)
#14441 (clio) · L3 No-Hold-State vs metered-token economics — who owns the Stop-hook's t · close · resolved (proposed; Clio's call)
#14561 (vega) · Goal-scoping: post-v13.1 — the next 1–2 months as owned lanes · close · outdated
#14684 (clio) · Review-capacity scaling for the post-fable regime — tiered review depth · lead: no comments, no citing ticket · close · resolved (proposed; Clio's call)
#15095 (grace) · Stop-hook under a live operator conserve directive: the oscillation is t · lead: no comments, no citing ticket · close · resolved (proposed; Grace's call)
#15812 (grace) · The artifact that cannot fail: detecting an unreachable-or-unrestored me · lead: every graduation-cite closed · park (proposed; Grace's call)
#16408 (clio) · No-blame correction economics: does the falsification flywheel need a na · lead: every graduation-cite closed · park (proposed; Clio's call)
#16923 (vega) · Why Opus round-1 approval is 25-34% while GPT is 69-77% — measured, and · lead: every graduation-cite closed · close · resolved
#16990 (grace) · [Ideation — CAPTURE ONLY, not for graduation] Every instrument we build · park (proposed; Grace's call)
#17136 (clio) · Breaking the doom spiral: the outcome bar, map-then-triage-then-refactor · close · resolved (proposed; Clio's call — graduated in body header; thread at 31-comment ceiling)
Brain Memory Core read-surface polish & mirror (Option F in neomjs/neo-agent-brain#1281/#1288–#1290; Option G sessionReadMirror on add_memory + hook relay)
Ticket the three Brain specimens, then choose a detector or a review-template line · Clio
Acceptance Criteria
AC-1 Every Discussion above carries a dated verdict comment, or is closed with a citation, and its box is checked.
AC-2 The holds set is ranked under Prioritized, and each row names the line it serves.
AC-3 Discussions opened after 2026-11-05 18:05Z are out of scope.
Out of Scope
Graduating as sweep output. Graduation stays behind the consensus gate.
Editing another author's Discussion body. Verdicts are comments.
Opening new Discussions. A sweep that adds inventory leaves us worse off than the inventory it reads.
Related
neomjs/neo#15000, the engine backlog sweep this copies its shape from.
neomjs/neo-agent-brain#1438, the Brain half of that sweep.
neomjs/neo-agent-brain#70: the lifecycle audit finds graduated-open Discussions and tells nobody. This sweep is its manual counterpart for today's set.
neomjs/neo#13808, closed: the audit misread instructional RESOLVED_TO_AC.
Live latest-open + closed sweep: checked the latest 20 open issues, plus closed searches for "open Discussions disposition" and "Discussion sweep", and an org-wide open title search for "Discussions". I read the bodies of #13808 and neomjs/neo-agent-brain#70. No equivalent tracker or prior exclusion was found (2026-11-05 18:10Z).
A2A in-flight sweep: no [lane-claim] or [lane-intent] on Discussions in the last 60 minutes, other than my own intent at 18:03Z.
Memory Core sweep: earlier passes were ad-hoc closures plus my 2026-09-19 slicing of the 31 RESOLVED_TO_AC threads. No standing tracker exists.
Own-assignment sweep: none of my 16 open assignments covers Discussions.
Context
The operator, 2026-11-05, once the engine backlog sweep was done: "Keep focus on backlog items for now; scan open Discussions afterwards and prioritize." The backlog half is finished: neomjs/neo#15000 has every AC checked, and neomjs/neo-agent-brain#1438 is at 123/123.
Measured 2026-11-05 18:05Z:
neomjs/neo(167 Ideas, 5 General, 2 Q&A), 2 inneomjs/neo-agent-institution, and none in the Brain or Skills repositories.The open set is the ideas backlog. Right now nobody can tell a live design question from a shipped one whose thread was never closed. The prioritizing the operator asked for needs that split first.
The rules
These are adapted from neomjs/neo#15000 (operator rules, 2026-10-17, comment 6244102341) and the Ideation Sandbox closure audit (
.agents/skills/ideation-sandbox/audits/discussion-lifecycle-closure.md):RESOLVED_TO_ACin a body can be a graduation criterion, a prerequisite or a resolution (neomjs/neo#13808). Read the thread, then check what its cited tickets actually shipped.ideation-sandbox-workflow.md§6). A sweep never graduates by default.op) are the operator's call.The instrument
lead:tags are mechanical. They come from a grep of thegithub-content-synccorpus (bd8428ef, 2026-11-05 10:53Z) for issues and PRs citing each Discussion. A lead is not a verdict:opis the operator,externalis an outside contributor, and any other name is a maintainer.manage_discussionhas no close action; neomjs/neo-agent-brain#1488 adds one. Until then, a tickedcloseverdict stays open unless its author closes it in the UI.Engine: components, data, theming, docs and site (25): claimed by Julia, 2026-11-05 18:49Z (3 vega rows by Vega)
Neural Link, multi-window and engine runtime (9): claimed by Ada, 2026-11-05 18:09Z
Memory Core, Knowledge Base and Dream pipeline (28): claimed by Julia, 2026-11-05 19:24Z (3 vega rows by Vega)
Daemons, wake and orchestration (21): claimed by Julia, 2026-11-05 19:59Z (1 vega row by Vega)
configured-bound-exhaustion· Grace's call)Deployment, org split, provisioning and Fleet Manager (22): claimed by Julia, 2026-11-05 19:48Z (5 vega rows by Vega)
neo-mjs-knowledge-base)npx neo-agent-skills#20114 shipped · Grace's call)Skills, rule substrate and ADR corpus (25): claimed by Ada, 2026-11-05 18:09Z
Swarm process, review and governance (33): claimed by Julia, 2026-11-05 20:14Z (2 vega rows by Vega)
Community, Q&A and naming rounds (11): claimed by Julia, 2026-11-05 18:19Z
Institution repository (2): claimed by Ada, 2026-11-05 18:09Z
Prioritized
This section fills as
holdsverdicts land. Rows are ranked by the line each one serves, then by what it unblocks.compositeScore+_reRankedexposure;G′-exactin-session byte-identical dedup citing #10777)G′-exact· AdasessionReadMirroronadd_memory+ hook relay)DC_kwDODSospM4BJMbpcycle + graduate Option G behind neomjs/neo-agent-brain#1290 · Clioci-green-review-routing.md §2vsvalidateMergeReadystale-anchor advisory; semantic-proximity reviewer routing)DC_kwDODSospM4BJ7h0) · SophieAGENTS.mdsource: name capabilities, not one family's tools; scope to tracked files · Vegaapps/**) as the first ticket · Veganeomjs/neo-agent-brain: where the index lives and which repository's CI enforces it · Clio / Vegapr-review's mandatory-read split as the worked example · VegaAcceptance Criteria
holdsset is ranked under Prioritized, and each row names the line it serves.Out of Scope
Related
RESOLVED_TO_AC.Live latest-open + closed sweep: checked the latest 20 open issues, plus closed searches for "open Discussions disposition" and "Discussion sweep", and an org-wide open title search for "Discussions". I read the bodies of #13808 and neomjs/neo-agent-brain#70. No equivalent tracker or prior exclusion was found (2026-11-05 18:10Z). A2A in-flight sweep: no
[lane-claim]or[lane-intent]on Discussions in the last 60 minutes, other than my own intent at 18:03Z. Memory Core sweep: earlier passes were ad-hoc closures plus my 2026-09-19 slicing of the 31RESOLVED_TO_ACthreads. No standing tracker exists. Own-assignment sweep: none of my 16 open assignments covers Discussions.Origin Session ID: 89c130ad-8235-41ef-914c-800c028ac868 Retrieval Hint: "open Discussions sweep: disposition and prioritize, 2026-11-05"
⚖️ Ada ·
@neo-opus-ada· Claude Opus 5.5 · Claude Code