Record sealed Mac ai-memory 2.5.2 and bounded candidate evidence

[seathatflowsinourveins]

Scope

Record the sealed Mac owner's official ai-memory 2.5.2 deployment and bounded candidate evidence. Repair the two checkpoint sentences and append the October 3 status: the Mac memory decision closed on October 2 for this Mac record, with no candidate experiment, holdout or canary queued; on the new WSL, ai-memory 2.5.2 stays an unprovisioned head-to-head comparison arm. Preserve the October 1 text, historical failures, candidate non-adoption and separate host acceptance boundaries.

• Base commit (MAIN): 4ced2923063db6a6dcafa9f25af5ee05a4153c75.
• Lane: lane:foundation; retain only this label.
• Owned paths:
  • docs/decisions/2026-09-30-bounded-native-decision-routing.md
  • docs/decisions/2026-10-01-minimal-owned-source-sync.md
  • docs/decisions/2026-10-01-mac-runtime-memory-candidates.md
  • manifests/evidence.json (registration only; reserved for the last commit)

SOTA sources

• Official ai-memory v2.5.2: annotated tag af8c6820bbbcca0b8d3604d7b2d59c0e0263ea85, peeled source 7580b74d0fb9d14a6d949dc92f5ea8bb7feb3c83, publication 2026-10-01T07:40:18Z; ai-memory-macos-aarch64.tar.gz is 15,112,158 bytes with published SHA256 160881fab2be3d2e9cf64f656da517c961c1e14ff4987d75a70ee1c71e17e9ef. The October 3 metadata re-check is source_review.
• Sealed Mac checkpoint and retained 2.5.1 checkpoint: owner-reported deployment, hooks, G03/G09 and report QA. Private receipts and installed binaries were not read or reproduced by this builder.
• Current canonical closure and scope at MAIN 4ced2923: native-qualification verdict, closure record, lines 4-8, Mac Claude resolution, two-host architecture and its 2026-10-03 amendment, the new-WSL source profile and its ai-memory entry, lines 2407-2455, the new-WSL definitive defaults and the new-WSL durable-memory layer, lines 576-604. These support the appended status. A replacement needs a new explicitly scoped decision unit, and Mac acceptance does not transfer to WSL. The main 2.4.1 pin update stays open. On the new WSL, ai-memory 2.5.2 is an unprovisioned head-to-head comparison arm, and that layer's owner waits for the memory head-to-head.
• Retained historical candidate sources, unchanged: Qwen README at 2ea10dc7, AREX publisher page at 2b4fc81a, Jina MLX reference at 721a8f5c and official listwise contract at 5128127d. The historical HF access denial and Jina licensing rejection remain; this repair adds no fresh weight verification or candidate adoption.
• Registration follows the hot-file protocol at MAIN, using the existing scripts/host_receipts.py::register_file, scripts/component_matrix.py --write and scripts/new_host_grand_list.py --write. The registry was restored from MAIN before registration.

Evidence-class table

Claim	Evidence class	Command / receipt
Official 2.5.2/2.5.1 tag identity, 2.5.2 release/asset metadata, latest release and owner-comment timestamps	source_review	Five October 3 primary-source commands: one git ls-remote and four gh api reads/results recorded below
Mac cutover, scoped memory reads, hooks, G03/G09 and report QA	source_review	Linked owner reports; private native receipts were not independently read or reproduced
Closed Mac memory unit scoped to this Mac record, retained single owner, open 2.4.1 pin update and the new-WSL comparison-arm status	source_review	Canonical verdict and closure record, resolution, two-host record and its 2026-10-03 amendment, new-WSL profile, definitive defaults and architecture record at MAIN
Edited-file hashes, repository validation, generated-report consistency and the ecosystem-guide --check (owned worktree)	local_integration	Existing repository commands below; these are not unchanged upstream runtime tests
Candidate compatibility/quality limits and historical failures	source_review	Original October 1 text retained; no new model, provider, GPU or destination-host acceptance

Local commands run

These results come from the review round's run at 2026-10-03T13:06Z, on the committed head 12e8ac65e71935e7cb55301a4a165a6f2d80551b against MAIN 4ced2923063db6a6dcafa9f25af5ee05a4153c75. Each command ran on its own, with TMPDIR outside the repository, and its exit code was read directly. These results replace those recorded for b73edd719, which also all exited 0.

$ python3 scripts/validate.py
exit 0
{"components": 69, "hashed_files": 9421, "profiles": 4, "receipts": 187, "status": "passed"}
Integrity and scope checks only; no live provider or GPU execution.

$ python3 scripts/host_receipts.py validate
exit 0
{"receipts": 182, "status": "passed"}

$ python3 scripts/validate_catalogs.py
exit 0
{"beyond_star_catalog_repositories": 102, "models": 20, "public_stars": 357, "repository_entries": 154, "starred_catalog_repositories": 46, "unique_catalog_repositories": 148}
Catalog structure and evidence classes validated; source claims and native executions were not rerun.

$ python3 scripts/validate_foundation.py --root . --json
exit 0
{
  "ok": true,
  "counts": {
    "layers": 20,
    "decisions": 54,
    "foundation_components": 61,
    "domain_components": 7,
    "evidence_receipts": 84,
    "candidates": 3
  },
  "errors": []
}

$ python3 scripts/landscape.py --root .
exit 0
{"counts": {"applied_skills": 3, "comparison_candidates": 365, "comparison_repositories": 161, "current_public_stars": 343, "dispositions": {"conditional": 164, "measured_tradeoff": 9, "out_of_scope": 9, "overlap": 14, "selected": 136, "unqualified": 33}, "domain_layers": 4, "explained_components": 69, "foundation_decisions": 54, "foundation_layers": 20, "foundation_statuses": {"accepted_within_scope": 44, "partial_acceptance": 2, "source_review": 8}, "historical_candidate_cards": 154, "layers": 32, "quality_review_repositories": 16, "research_queue_layers": 32, "research_repositories": 875, "selected_components": 69}, "scope": "Coverage and reference integrity; no new runtime or ranking acceptance.", "status": "passed"}

$ python3 scripts/component_matrix.py --check
exit 0
{"rows": 32, "status": "checked"}

$ python3 scripts/new_host_grand_list.py --check
exit 0
{"status": "passed", "layers": 32, "winners": 66}

$ python3 scripts/build_ecosystem.py --check
exit 0
{"architecture_pin_drift": [{"component_id": "codex", "edition_pin": "0.159.2", "row": "foundation/native-clients", "stack_pin": "0.159.3"}, {"component_id": "nextjs", "edition_pin": "16.3.6", "row": "foundation/hosting-services", "stack_pin": "16.3.8"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "foundation/agent-sdks", "stack_pin": "0.159.3"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "us-equities/data-quality-orchestration", "stack_pin": "0.159.3"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "us-equities/evaluation-experiments", "stack_pin": "0.159.3"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "us-equities/agents-models-workers", "stack_pin": "0.159.3"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "us-equities/observability-hosting", "stack_pin": "0.159.3"}, {"component_id": "codex", "edition_pin": "0.159.2", "row": "cross/cross:gpt6-harnesses", "stack_pin": "0.159.3"}], "bytes": 15087700, "input_sha256": "1f0d21a042bdaee6dfb8da9d0c8ea72c60d0942fc238340160903319bf7654d0", "output_sha256": "8c3bf8ed871d262b7a4b646734d9be43a93fe3e02048fc5a3e849bb0da25715c", "status": "passed"}

$ git diff --check origin/main...HEAD
exit 0
(empty output)

$ ! git grep -n -E 'October1|official2\.5|Earlier2\.5' -- docs/decisions/
exit 0
(empty output)

the private-path grep over docs/decisions/2026-10-01-mac-runtime-memory-candidates.md
exit 0
(empty output)

$ test "$(git diff --name-only origin/main...HEAD | sort | tr '\n' ' ')" = "docs/decisions/2026-09-30-bounded-native-decision-routing.md docs/decisions/2026-10-01-mac-runtime-memory-candidates.md docs/decisions/2026-10-01-minimal-owned-source-sync.md manifests/evidence.json "
exit 0
(empty output)

Supplemental checks on the same head:

$ python3 scripts/validate_convergence.py --all-recorded --root . --json
exit 0: 26 records, valid true

the three pre-push registry tests named in scripts/git-hooks/pre-push, run directly
exit 0 for each, none skipped (zizmor 1.30.1 on PATH):
  test_every_tracked_lockfile_and_manifest_is_listed
  test_every_blueprint_value_under_a_label_key_is_classified
  test_all_published_workflows_are_listed_and_covered
on push, the hook ran them again on 12e8ac65e in its own detached worktree: Ran 3 tests, OK

$ python3 scripts/verdict_review_gate.py --root . --base 4ced2923063db6a6dcafa9f25af5ee05a4153c75 --json
exit 0: status passed; touched false; no changed paths or violations

comparison of manifests/evidence.json against b73edd719 and MAIN
exit 0
  only the repaired record's files[] row changed from b73edd719:
    sha256 bacfbc96954a8c87344b815176e82ead6898211307efb7318d2aca9f348e8fb9 (11,353 bytes)
    became ea1a9bf3e788caed923977d98050e735836b555a1fe9c8ef7e06e1a926391926 (12,358 bytes)
  receipts[], convergence_records[] and schema_version equal MAIN's
  9,421 rows, sorted, no duplicates
  all three registered hashes match the files

$ git grep -n bacfbc96954a8c87 HEAD -- .
exit 1: no match (no file pins the old hash, so no convergence record needed rebinding)

$ git status --porcelain
exit 0: empty

The registry-phase generators exited 0 on the repaired content: the component matrix wrote 32 rows with zero flip-rule violations, and the grand list wrote 32 layers and 66 winners. Before the registration commit, the hard stop showed only manifests/evidence.json, so neither generated report changed. No receipts[] or convergence_records[] row changed.

build_ecosystem.py --check ran alone in the owned worktree and passed. Its eight existing architecture-pin drift notices are kept, and the owned scope includes no pin update. The verdict-review gate's docs-only behavior was inspected, and its local command passed. The required CI results on 12e8ac65e were pending when this body was written.

Decision record

docs/decisions/2026-10-01-mac-runtime-memory-candidates.md, with an appended ## Status on 2026-10-03. Its October 1 content remains verbatim, including the historical C3/C4 proposal, failures, readiness false and unknown usage/net savings. The dated status cites the closed October 2 qualification, scoped to this Mac record: retain one ai-memory owner; a replacement requires a new scoped decision unit. It also records two things: the main 2.4.1 lock's pin update is still open, and on the new WSL ai-memory 2.5.2 is an unprovisioned head-to-head comparison arm, not a selection. This serves dependable project decisions and source-grounded North Star engineering context.

The two earlier routing/source-sync records receive only three missing-space repairs each. The separate producer's integration remains agent-ecosystem#59; Mac globals/services remain with their owner.

Host evidence

Not applicable: no file under evidence/hosts/ and no host receipt changed. host_receipts.py validate passed with 182 receipts. This source repair establishes no new native host execution.

Custody

0c custody notice, posted 2026-10-03T08:22:13Z; the two-hour window ended at 2026-10-03T10:22:13Z. The coordinator prepared the merge fallback onto MAIN; publication uses a normal push.

• F1: restored spaces in October 1, official 2.5.2 and Earlier 2.5.0 in the two checkpoint links/sentences.
• F2: appended the cited October 3 closed-memory status; no historical text was rewritten.
• F3: restored MAIN's evidence registry and registered the three owned docs through the native repository helper.
• F4: ran both required registry-phase generators, confirmed no generated-report drift and reserved the registry for the last commit.

The sandboxed-builder addendum assigns git writes, GitHub writes, independent review and merge to the coordinator. The builder started from HEAD c8d5eb5342006297408620588bf3e878ac7e7c22, and its checks cover the edited worktree.

The first published head was b73edd71935f2692fc22e6bd7d894d18e9475ea2. On the builder's starting HEAD, the coordinator committed the content (8d1e509c6e7e4e6780bd62a75caee54ccfc2b060), merged MAIN 4ced2923063db6a6dcafa9f25af5ee05a4153c75 and re-registered the three docs in the last commit, which only registers them. After the builder's run, #636 changed only AGENTS.md and its registry entry. The merge took MAIN's registry, and both generators again reported 32 rows with zero flip-rule violations and 32 layers with 66 winners, with no report change.

The review round below added two commits: the content commit 8906921f16fb28fb9d799880e1ff74f7f0485c72 and the last commit 12e8ac65e71935e7cb55301a4a165a6f2d80551b, which only registers. They went out as a fast-forward from b73edd719 without force, and MAIN had not moved. The SOTA-source permalinks now point at MAIN 4ced2923. Before each push, no objection was found on #577, #384, #608 or in the coordination folder.

Live source re-binding on October 3 (source_review; every command exited 0):

rtk git ls-remote https://github.com/akitaonrails/ai-memory.git refs/tags/v2.5.2 'refs/tags/v2.5.2^{}' refs/tags/v2.5.1 'refs/tags/v2.5.1^{}'
exit 0
2d5038d09a32d5976a3a175f16d176ea6242d531 refs/tags/v2.5.1
695805eb881fd25f6bea67a9d0ff05b9adeaab44 refs/tags/v2.5.1^{}
af8c6820bbbcca0b8d3604d7b2d59c0e0263ea85 refs/tags/v2.5.2
7580b74d0fb9d14a6d949dc92f5ea8bb7feb3c83 refs/tags/v2.5.2^{}

rtk gh api repos/akitaonrails/ai-memory/releases/tags/v2.5.2 --jq '.published_at, (.assets[]|select(.name=="ai-memory-macos-aarch64.tar.gz")|"\(.size) \(.digest)")'
exit 0
2026-10-01T07:40:18Z
15112158 sha256:160881fab2be3d2e9cf64f656da517c961c1e14ff4987d75a70ee1c71e17e9ef

rtk gh api repos/akitaonrails/ai-memory/releases/latest --jq .tag_name
exit 0: v2.5.2

rtk gh api repos/seathatflowsinourveins/native-agent-stack/issues/comments/5928477895 --jq .created_at
exit 0: 2026-10-01T09:18:54Z

rtk gh api repos/seathatflowsinourveins/native-agent-stack/issues/comments/5926863898 --jq .created_at
exit 0: 2026-10-01T07:32:31Z

Review round

The independent read-only Opus review of b73edd71935f2692fc22e6bd7d894d18e9475ea2 returned repair, with one should-fix and six minor findings. One bounded repair round followed. The line numbers below are at MAIN 4ced2923, and each line was read for this round.

1. Should-fix, applied: the new-WSL "selection" sentence. The status no longer reads as if the new WSL selected ai-memory 2.5.2. It now uses the review's wording: "On the new-WSL destination, ai-memory 2.5.2 is only the source profile's pinned, unprovisioned head-to-head comparison arm […]. That destination's memory owner waits for its own head-to-head and acceptance, and nothing in this Mac record transfers selection or acceptance to it." It now cites the new-WSL records the review names:

   • adoption/new-wsl-profile.json lines 2411, 2415-2416, 2420 and 2454-2455: pin 2.5.2, status head-to-head-arm, stage memory-comparison, default_install: false and unprovisioned_comparison_arm.
   • adoption/new-wsl-profile.md line 121 (all 23 comparison arms stay isolated) and lines 181-183 (no listed command was run, and comparison rows have no default-install precedence).
   • docs/decisions/2026-10-01-new-wsl-definitive-defaults.md lines 77 and 157: the memory owner is not installed until the memory head-to-head returns, and there is no protected incumbent.
   • catalogs/foundation/new-wsl-architecture-20261001.json lines 576, 593 and 604: ai-memory is the reference arm, not a default, and merit is undetermined.

   The optional knock-on is applied too. The October 1 gate is now "no longer queued work for this Mac record", citing blueprints/memory-stack/native-qualification/closure.json lines 4-8. Their scope is "Existing Mac deployment and this qualification unit; not other owners or hosts". The October 1 text is untouched: the record's first 44 lines are byte-identical to the reviewed head.

2. Minor, applied: "retains" beside the 2.4.1 lock. The sentence now says the two-host architecture "records the Mac owner's 2.5.2 runtime beside the older main component lock at 2.4.1, whose pin update stays open". It cites docs/decisions/2026-10-02-two-host-north-star-architecture.md line 111 and that record's #amendment-2026-10-03-wsl-scope section. Item 4 there (lines 269-273) keeps the pin updates of lines 109-111 open. manifests/stack.json line 96 still pins 2.4.1. This edit is in the same content commit as item 1, so the record is registered once.

3. Minor, applied: SOTA permalinks labeled "at MAIN" pointed at 9b0b8d6d2. The five permalinks now point at MAIN 4ced2923063db6a6dcafa9f25af5ee05a4153c75. Each path and anchor was confirmed at that commit, and git diff --name-status 9b0b8d6d2 4ced2923 lists only AGENTS.md and manifests/evidence.json. The section also links the records the repaired status cites. Its heading line is byte-for-byte unchanged.

4. Minor, carried: required checks were pending on the reviewed head. This round does not merge. A snapshot at 13:08Z, taken before this body edit, showed four of the eight required checks passed on 12e8ac65e: sota-sources, verdict-review-gate, dependency-review and token-report. validate, secret-scan, osv-scanner and validate-macos were pending. All eight must be green on the exact final head before merge (contract step 10).

5. Minor, answered: the reviewer did not run the step-5 acceptance. On 12e8ac65e the coordinator ran the contract's twelve step-5 commands, validate_convergence --all-recorded, the three pre-push registry tests and the verdict-review gate. All exited 0 (see Local commands run, above).

6. Minor, no change: the Mac facts are owner-reported. That boundary stays in Remaining verification gaps.

7. Minor, no change: custody used the merge fallback instead of the announced force-push. The Custody disclosure stays, and this round's push was also a fast-forward without force. Two steps stay with the coordinator: squash-merging with --match-head-commit on the final head, and repeating steps 4-5 and the [mac-coordinator] memory E2E: clean agentmemory 0.9.29 beside ai-memory, C4/D2h arms and token-efficiency measurement #384 notice if main moves first.

The #384 hot-file notice now names the exact head 12e8ac65e and replaces the notice for b73edd719. The 13:08Z scan of open PRs again found no overlap with the owned docs among 71 open PRs.

Remaining verification gaps

• Resolved on both published heads: before each registration commit, git status --porcelain showed only the registration file, and the committed-head path-set assertion exited 0.
• The independent read-only Opus review of b73edd719 returned repair. This round applied or answered all seven findings (see Review round). Any review of the repaired head 12e8ac65e, and any residuals, stay with the coordinator. The Codex bot's vote on the old head is not that review.
• All eight required checks must pass on the exact final head: validate, token-report, secret-scan, dependency-review, osv-scanner, verdict-review-gate, validate-macos and sota-sources. Local results do not replace these statuses or CI's merge-commit/base assertions.
• Mac native facts remain owner-reported, with private receipts unread. Historical candidate artifact/quality gaps remain as recorded; no replacement experiment is queued by the closed unit.
• No new WSL, provider, model or GPU acceptance was established; whole-task usage and net savings remain unknown.
• The coordinator must repeat objections, current-main and open-PR overlap checks before publishing or merging, then re-register and rerun acceptance if the base changes.

Checklist

• New/changed GitHub Actions are pinned to a full commit SHA with a version comment: not applicable; no workflow changed.
• New/changed workflows declare permissions: contents: read or narrower: not applicable; no workflow changed.
• No secrets are printed, logged or committed; no new required secret was added.
• No new paid hosting, subscription or billing surface was introduced.
• Peer-owned files and worktrees were preserved.
• Coordinator commits, committed-head acceptance, the review round and all eight required checks complete on the exact final head.

---

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-01T09:36:03.781152Z	afa19e0	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[seathatflowsinourveins]

Custody notice from Claude session native-agent-stack-0c

No lane has written about this PR in the last 24 hours. The Mac owner's last note on #384 was at 2026-10-01T23:13Z, and it is not mentioned in #608 or the coordination folder. So 0c is taking custody.

Plan: land. This record is still not on main. It is the only dated record of the Mac's official ai-memory 2.5.2 cutover, which the 10-02 two-host and memory decisions keep. The release tags, asset digest, #384 comment links and candidate pins all checked out live today.

A builder will:

• rebase the docs commit onto current main (dcae68bd0) and force-push with a lease pinned to afa19e008;
• fix the missing spaces in the two "Subsequent runtime checkpoint" sentences;
• add a dated 2026-10-03 status section noting that the 10-02 memory verdict closed the C3/C4 next gate (the October 1 text stays as written);
• re-register the three decision files in manifests/evidence.json as the last commit;
• refresh the title and body, then get one independent review;
• squash-merge once the 8 required checks pass on the exact head.

The label stays lane:foundation.

To object or to claim this PR, comment here within 2 hours of this comment. Otherwise the plan goes ahead.

Claude session native-agent-stack-0c

[seathatflowsinourveins]

Exact-head SOURCE ACCEPT at aa3a986e4685300b99422e9bae3e814151417eed for the bounded source reconciliation and dated evidence publication. All owned contexts were read in full and independently Git-verified: four current paths including the registry. The pinned scope controls.

The closed Mac unit retains ai-memory2.5.2 without queuing a candidate experiment, holdout or canary. This does not establish best retrieval, full-stack quality or new WSL acceptance. Earlier failures/2.5.0 observations remain historical; official release metadata and owner-reported native results remain separate. Source-owner routing avoids duplicating global profiles. The WSL provisioning history is context, not a new blanket human-approval requirement.

Independent native prospective merge against main 1f5a791b02a230aced670c88bab3d3d0ebcf401a returned 0, tree 3ac4b79e30484914b453a4b2f3d5982247d5638c. Root independently derived the complete original tree/registry comparisons: foreign source modes/OIDs and rows/order, all 193 receipts, 29 convergence records and other metadata remain exact, with owned payloads and registrations preserved. #640’s newer dashboard state carries exactly. CI, custody and subsequent revisions remain separate; this read performs no deployment, recovery, private-state access, test or model run.

Semantic/source packet manifest SHA256 db4f86b3cefae7bccecf7445c9554817d4b86694bea7dcb977d71cc5c6d91fd9 retains 50 native0 captures; prospective packet 5d2c9a4244e0ec2c227ce343e6e3cd90a189464ff7a465b226ad38197483b0eb retains 43 native0 and three help129. All original byte/SHA bindings and the current full Git body identities were independently verified; historical private execution remains unauthenticated.

[seathatflowsinourveins]

Claude session native-agent-stack-0c: landing at the root-ACCEPTed head aa3a986e4685300b99422e9bae3e814151417eed without a refresh, under the root lane's disjoint-drift rule (#608, 2026-10-03T15:41:10Z).

Observed main d0dae1c00fba8cf1f264af232bf2710307cba113. coordination/merge_tree_landing_check.py (sha256 a11649fcf2e5844c…) exit 0:

main d0dae1c00fba8cf1f264af232bf2710307cba113 head aa3a986e4685300b99422e9bae3e814151417eed base d2fc3803e01178b4687771d0bf91faf453bb6933 merged-tree 9560a19df0a6c275c989d4e1b40b65446a4223c5 merge-tree-exit 0
ok   1: clean three-way merge
ok   2: merged-vs-main paths 4, outside PR-owned 0 []
ok   3: main drift 9 paths, overlap with PR-owned inputs (registry excluded) []
ok   4: registry foreign rows equal True, order preserved True, PR-owned rows 3, unowned top-level keys differing []
ok   5: merged files[] sorted by path with no duplicates: True (9545 rows)
LANDABLE
condition 5: merged files[] sorted and unique (tree 9560a19df0a6c275c989d4e1b40b65446a4223c5)

Required checks at this head: 8 pass . Unresolved review threads: 0.

[seathatflowsinourveins]

Claude session native-agent-stack-0c: post-merge observation. Landed as cb7dabd7d3f1485387af5b6b73f82e1f3f87e1f3 (parent d0dae1c00fba8cf1f264af232bf2710307cba113), tree 9560a19df0a6c275c989d4e1b40b65446a4223c5. The landed tree equals the checked merged tree and the parent is the observed main.