Post-1.0 Roadmap · 2026

Sibyl

Collective·Intelligence

In a field where memory systems can't reliably beat cat *.txt on accuracy, Sibyl's edge isn't a bigger number — it's making auth-scoped, graph-native team memory a database guarantee instead of an application-layer hope: the one layer the frontier labs won't build and the benchmarks say no one has solved.

v1.1Prove It
v1.2Coalesce It — safely
v1.3Lead It
RELEASE FLOOR  v1.0.2 shipped RETRIEVAL  96.96% strict R@5 CONTEXT-PACK  p95 84 ms · 0 leaks
The Thesis

The honest edge is not a bigger benchmark number.

The AI-memory field in mid-2026 has a credibility problem. Purpose-built memory systems routinely fail to beat a plain full-context baseline on accuracy: on LoCoMo a no-memory baseline scores ~73% — matching or beating "sophisticated" systems — and every headline leaderboard number is self-reported on a different harness. When the two most-cited systems (Mem0 and Zep) audited each other, both scores moved by 15–25 points. The honest, defensible wins here are cost, latency, token-efficiency, and isolation integrity — not a leaderboard.

Two structural facts make Sibyl's position stronger than it looks.

Fact 01

The substrate is necessary — but it was never the moat.

A unified engine matters: Mem0 dropped external graph stores, Cognee's isolation works on only three backends, Papr stitches four databases together. But SurrealDB's own Spectron ($23M, Feb 2026) now makes the unified-substrate, typed-memory, forgetting-as-a-verb argument on the very engine Sibyl runs on. You can't out-database the database vendor — so Sibyl competes above the substrate: the governed, self-hostable memory product.

Fact 02

Team memory is unsolved — and the literature says so.

Almost all "multi-agent memory" is just per-agent namespacing. The 2026 governance literature names four failure modes no system clears (leakage, stale propagation, contradiction persistence, provenance collapse); GateMem shows no method achieves utility, access control, and forgetting together; and no public benchmark measures team memory at all. Sibyl already ships the two hardest prerequisites: a real auth runtime and a unified graph.

The bet is a three-release arc: Prove It — close the table-stakes eval gaps honestly, at PR-gating cadence, and lay the team-memory substrate. Coalesce It — safely — ship live, provably-scoped, reversible team-memory coalescence, and build the team benchmark internally. Lead It — publish the benchmark the field lacks and push the frontier.


Verified Starting Point

Where we actually stand — receipts, not vibes.

Every claim below is grounded in verified code reality and a pinned run, not the frozen pre-1.0 planning docs. The honest gaps are listed as plainly as the wins.

96.96%
Strict Recall@5 on LongMemEval-S
retrieval recall · no LLM in path · run 26304777971
98.90%
Recall@10 on the same pinned run
retrieval ceiling, not QA accuracy
84ms
Context-pack p95 latency
160/160 packs · zero leaks
0
Cross-scope leaks under test
physical namespace-per-org isolation
Shipped
Substrate is healthy. The 2026-05-28 audit (C1–H10 + mediums) is fully remediated: transactions wrap destructive cascades; the per-org query lock is now a per-org connection pool; a shared ranker and shared RRF serve both retrieval surfaces; the empty-API-key-scope hole is closed. W13 (Graphiti / FalkorDB / PostgreSQL removal), W14 (Epic → task-tree), OIDC, Argon2id, Helm, and Ansible all shipped.
Partial
Forgetting / decay is uneven — and blind. The hybrid path applies temporal decay by default (apply_temporal=True, 365-day half-life); the context/recall path does not (temporal_target=None). priority_decay archives low-importance entities reversibly — but it reads usage fields (last_recalled_at / last_used_at) that nothing writes (verified 2026-07-03), so forgetting is effectively age-only: "old = dead" instead of "unused = dead." No citation contract, no recall stamping, no usage signal anywhere in the loop.
Unwired
Two compounding loops exist as structure, not as loops. Verbatim raw captures + lineage edges + a versioned extractor (v0.12) mean extraction improvements could replay over the whole historical corpus — but no re-extraction path exists. Tasks and memories share one graph, so memory utility could be grounded in task outcomes — but no outcome signal is derived. Both properties are unique in the field; neither compounds yet.
Disabled
Team memory: every ingredient exists, all off/unmanaged/offline. MemoryScope.TEAM is hard-disabled in three places. The teams / team_members tables and RBAC resolution exist but there's no team-management route or CLI. Two coalescence engines exist — online retrieval/dedup.py (HNSW + cosine) and offline migrate/merge.py — and the sibyl-consolidation/ merge tarballs are a real dogfood of exactly this.
Weak automation
Eval infra: strong methodology, weak automation. Honest hit-vs-strict recall, per-question physical isolation, and gate + manifest provenance are in place. But evals don't gate PRs (manual dispatch); the nightly runs on a mock LLM; there's no time-series regression, no end-to-end QA-accuracy number, no scored LongMemEval-V2; the large_corpus_rehearsal is 57 records; zero competitor baselines run.

Our Place in the Field

You can't out-database the database. So we compete above it.

Most rivals bolt memory onto the application layer; the substrate vendors now sell memory too. Sibyl's durable edge isn't the storage engine — it's the governed, self-hostable memory product and the coalesced-team-memory frontier no one has cleared.

System Substrate Isolation Coalesced team memory Principled forgetting Open / self-host
Sibyl unified graph + doc + vector + FTS + auth, one engine physical namespace-per-org building substrate ready; coalescence is the v1.2 bet partial decay + reversible archival; unifying in v1.1 OSS product MCP + CLI + web, self-hostable
SurrealDB Spectron unified same engine class as Sibyl scope grants DB-enforced, claimed claimed newest, least-demonstrated forget-verb first-class, claimed preview invite-only · commercial TBD
Neo4j agent-memory graph+vector no document / time-series model app-layer shared brain mature, with provenance noneprovenance, no decay GPLv3 + Aura
Zep / Graphiti temporal KG over Neo4j / Falkor app-layer group graphs shared, coarse bi-temporal edge invalidation — best-in-class Graphiti OSS Zep CE deprecated
Mem0 split (BYO) vector + optional graph app-layer namespacing minimalsearch-time re-rank Apache-2.0 layer bring your own backends
Letta / MemGPT runtime Postgres + pgvector app-layer shared blocks real coalescence sleep-time background consolidation Apache-2.0
Cognee stitched relational + vector + graph partial 3 of N backends no aspirational Apache-2.0

Every headline benchmark number in this space is self-reported on a non-standardized harness. Marks: green = truthful strength · yellow = partial / claimed / app-layer · grey = absent or unproven.

The sharpest competitor is our own substrate vendor. SurrealDB's Spectron makes nearly Sibyl's entire differentiator list — on the engine Sibyl depends on. The lesson isn't to out-database the database; it's that the substrate was never the moat. Sibyl's moat is the governed, self-hostable memory product above it — the memory loop as legible verbs, a real auth runtime, task coordination as a first-class memory citizen, and coalesced team memory the benchmarks say no one has solved. Spectron is a preview engine; Sibyl is a complete, installable product on the frontier the labs refuse to build.


How We Measure · The Honest Instrument

Numbers you can't dismiss — because we report the way no one else does.

Every headline in this field is self-reported on a different harness. When the two most-cited systems audited each other, both scores moved 15–25 points. Our moat isn't a leaderboard slot — it's a measurement discipline.

What we report — pinned to a receipt

  • LongMemEval-S QA accuracy (official GPT-4o judge), reported separately from our 96.96% retrieval R@5 — never conflated.
  • BEAM-1M / 10M — the honest way to substantiate million-token claims: long-context-alone vs. +memory delta.
  • FAMA & ForgetEval — forgetting quality that penalizes obsolete-memory reuse. Almost no one reports this.
  • MemoryAgentBench — selective forgetting + test-time learning, where most systems collapse.
  • LongMemEval-V2 — agentic + multimodal, harder than V1; the current bar is ~72–75%.
  • Cost / latency / tokens at a fixed accuracy budget — the frontier, not raw savings against a strawman.

What we refuse to report — and why

  • Single-needle NIAH / passkey — saturated; a green heatmap is now an anti-signal (NoLiMa, HELMET, RULER).
  • DMR — retired; plain full-context already scores ~98%.
  • Bare LoCoMo — ~6.4% wrong gold answers, judges accept ~63% of wrong answers, and full-context beats "sophisticated" systems. Cite it critically or not at all.
  • Retrieval recall dressed as QA accuracy — the category error we refuse to make.
  • Cross-vendor LoCoMo rankings — non-comparable by construction.
The guard rails every number carries

A no-memory control (subtract parametric knowledge, report the memory lift); the judge model and exact prompt disclosed; N ≥ 5 runs with confidence intervals; parallel — not sequential — retrieval for latency; and competitor configs published so they can object. One canonical run, never rounded up.


The Frontier · Where We Aim

Team memory is the field's unsolved problem — and we're built to solve it.

Single-agent memory has quietly converged on a mature stack: bi-temporal graphs, hybrid retrieval, tiered consolidation. The moment the field turns to shared memory, it falls apart. That is the opening.

Solved · commoditized
  • Hybrid retrieval — BM25 + HNSW + RRF + rerank
  • Bi-temporal validity (valid vs transaction time)
  • OS-style memory tiering & paging
In progress · unsettled
  • Memory-aware ranking (time · salience · provenance)
  • Sleep-time / background consolidation
  • Outcome-conditioned forgetting
Open · we lead here
  • Team / shared-memory governance
  • Coalescence & convergent merge
  • Provenance & isolation at scale

We don't compete where it's solved. We lead where it's open.

The 2026 governance literature has already named the way shared memory breaks. Any system that lets multiple agents or people write into one memory faces the same four failure modes — and current systems trip every one of them.

Failure 01

Unauthorized leakage

Data crosses a tenant or sub-tenant boundary it was never scoped to.

Failure 02

Stale propagation

Outdated facts keep circulating long after they stopped being true.

Failure 03

Contradiction persistence

Conflicting facts coexist unresolved, and the system can't say which holds.

Failure 04

Provenance collapse

Who asserted what, and from where, becomes unrecoverable.

"No method simultaneously achieves strong utility, robust access control, and reliable forgetting." — GateMem (arXiv:2606.18829), benchmarking memory governance across multi-principal agents · 2026

The field's own prescribed cure is four systems primitives — scoped retrieval, temporal supersession, provenance tracking, and policy-governed propagation — and its own benchmarks show nobody has assembled them into something that holds. Sibyl already ships the two hardest prerequisites the academic prototypes bolt on after the fact: a real multi-tenant auth runtime (namespace-per-org) and a unified graph. The primitives aren't a research project for us; they're the substrate we already stand on.

The thesis, crystallized

A bi-temporal, provenance-carrying, scope-governed unified graph — where compatible knowledge merges by convergence, conflicting knowledge supersedes by valid-time invalidation (never destructive overwrite), forgetting is outcome-conditioned, and a background dream cycle consolidates while keeping every summary linked to its lossless evidence.

No published system does all of this together. Governed coalescence over a bi-temporal graph — with forgetting tied to whether a memory actually helped, which only a task-aware system can measure — is the position Sibyl is uniquely shaped to own. The three-release arc below is how we earn the right to claim it.

bi-temporal · valid vs transaction time temporal supersession scoped retrieval policy-governed propagation provenance-union retrospective permission checks convergent (CRDT-style) merge outcome-conditioned forgetting active (verifiable) forgetting sleep-time consolidation

Testing It At Scale · The Benchmark The Field Lacks

If team memory is unmeasured, we build the measure.

No public benchmark scores whether memories written by many agents coalesce into one correct, attributable, isolated graph. CoalesceBench is that benchmark — pluggable through a standard Insert/Query API so every existing memory system can be run against it, and co-evolving with Sibyl's own namespace-per-org isolation and provenance edges.

Task 01

Cross-agent coalescence

A fact written by agent A must answer agent B's later query; the hidden-profile variant needs facts uniquely held by ≥2 agents. Does the graph merge, or silo per writer?

Integration Gap = team-memory acc − best single-agent acc. A siloed store scores ≈ 0.
Task 02

Conflict reconciliation

Contradictory writes — temporal and authority-based. Return the superseded-correct answer, or surface both with provenance when it's genuinely unresolvable.

Reconciliation accuracy · supersession precision / recall.
Task 03

Provenance & attribution

"Who established fact F, and through what derivation chain?" — including multi-hop derivations. Essentially untested in public today.

Writer-identity accuracy · chain reconstruction @ depth-d.
Task 04

Isolation under load

Two tenants with deliberately similar facts under concurrent writes; scoped queries must never leak, even under adversarial paraphrase and embedding collision.

Cross-tenant leakage = 0 (a cap, not a tradeoff) · utility retained.
Task 05

Consolidation fidelity

Overlapping multi-agent experiences dedupe and merge without double-counting; deletions propagate to every reader (no stale propagation).

Merge precision / recall · post-deletion non-recovery across agents.
Scoring

A radar, not a scalar

Report the five axes as a vector; always include a no-memory and a single-agent control so "team lift" is explicit. Any nonzero cross-tenant leakage caps the composite — isolation is a safety property.

Frozen harness · fixed reader · N ≥ 5 + CIs · public baselines.

Design Principles

Eight rules that govern sequencing and scope.

Principle 01 · governs everything

Trust and reversibility come first.

One private-memory leak or one irreversible bad merge makes team memory radioactive. Every coalescence action must be scoped-by-construction, attributable, previewable, and undoable. This principle governs the entire sequencing of the arc.

Principle 02

Lead with cost / latency / isolation.

Not benchmark size — the field's only honest edge and Sibyl's structural strength.

Principle 03

Honest benchmarks stay the moat.

One canonical run, never round up, and the retrieval-vs-QA-accuracy caveat always travels with the number.

Principle 04

Eval the write path, not just the read path.

Memory systems accumulate hallucinations during extraction/update that QA-only judging hides (HaluMem).

Principle 05

Gates carry budgets, not slogans.

Every gate names numeric thresholds: recall, latency, leak, error, cost.

Principle 06

Do not chase discredited benchmarks.

LoCoMo is saturated, has ~6.4% wrong gold answers, and is harness-dependent. Prioritize LongMemEval-V2 and Sibyl's own team benchmark.

Principle 07

Turn the platform threat into distribution.

Be the MCP-native, OKF-speaking backend behind others' memory tools rather than competing with client-side file memory. Structured substrate, curated projection: the graph is the source of truth; filesystem-native agents get it as curated files.

Principle 08

Close loops before tuning them.

Linear work improves the system additively; feedback loops compound. Sequence loop-closing (usage signal, re-extraction, outcome grounding) ahead of parameter tuning — never tune a decay function whose input signal does not exist yet.

I
v1.1

Prove It

Make every public claim boring and complete. Move evals from manual to PR-gating, finish the eval story, close the usage feedback loop, and enable the team-memory substrate. No headline claim ships without a receipt.

W1

End-to-end QA-accuracy lane

Add a reader pass over Sibyl's retrieved LongMemEval-S sessions plus the official judge; publish QA accuracy alongside the 96.96% R@5 retrieval number. Closes the miscategorization gap where tables read retrieval recall as QA accuracy.

qa-accuracy-gate publishable QA number from a pinned run; fails if QA accuracy drops > 1.0 pp vs last committed.
W2

Eval automation + regression-over-time

Scheduled real-key runs (stratified nightly, full weekly); a committed time-series ledger; and a deterministic local-embedding variant so quality gates run on PRs without OpenAI cost or nondeterminism.

eval-regression-gate · blocks PRs strict recall@5 ≥ (last committed − 0.5 pp); context-pack p95 ≤ 1000 ms; leak_count = 0.
W3

Cost / latency / token accounting

Co-report tokens per query, embedding calls, p50/p95, and a dollar estimate against a full-context baseline, in every live artifact.

cost-latency-gate per-query cost and p95 recorded; p95 within budget; cost regression flagged.
W4

Write-path integrity (HaluMem-style)

Measure whether extraction and consolidation inject or reduce hallucinations, and gate it. On-trend, under-shipped by competitors, consistent with the honest posture. Plus two cheap hygiene guards: a self-feeding guard (the dream cycle must never digest its own reflection output) and a no-op gate on background extraction (minimum signal or write nothing).

write-path-integrity-gate extraction/consolidation hallucination rate ≤ threshold on a seeded fixture; self-feed fixture produces zero self-referential writes.
W5

LongMemEval-V2 (published)

Stand up the official-full harness and publish the LAFS-Gain number with a citable receipt. The external leaderboard is empty and the benchmark is agent-shaped — it matches Sibyl's coding/agent use case, so an early credible entry is high-leverage.

longmemeval-v2-gate scored run with committed receipt; regression bound on LAFS Gain.
W6 · highest leverage

Usage feedback loop (citation contract)

Write the usage fields the decay machinery was designed to read. Context packs already carry stable item IDs; the skill packs ship a citation contract (a cited_ids field on task complete / reflect, or sibyl cite); a server hook records idempotent usage events, stamping last_recalled_at (exposure) and last_used_at (citation). Usage drives both survival and consolidation priority — forgetting becomes "unused = dead," and every later layer (decay tuning, distillation, TeamMemBench utility) consumes this signal.

usage-loop-gate usage events flow end-to-end from a live session; a cited entity measurably outlives its uncited twin through decay; recall stamping on 100% of pack builds.
W7

Forgetting: uniform + benchmarked

Apply temporal decay across the context/recall path (not just hybrid); confirm consolidation scheduling; tune priority_decay and benchmark it (FadeMem-style storage-reduction %, recall impact, write-path integrity). Turns the self-named "honest gap" into a measured feature. Hard-depends on W6 — tuning decay without a usage signal is tuning noise.

W8

Team-memory foundation (substrate)

Enable the team scope across its three gates; ship team-management routes + CLI; wire team → memory-space; implement the SHARE/promote action with provenance and attribution retained. Substrate, not coalescence yet.

team-scope-trust-gate private / delegated / project memory provably cannot surface in a team pack; promotion is attributed and preview-shown; leak fixtures = 0.
W9

OKF export — doubles as the memory changelog

sibyl export --format okf — one Markdown + YAML file per entity, edges as Markdown links, the labeled-property-graph preserved losslessly via OKF-legal extension frontmatter so the bundle is valid OKF for other tools yet round-trips back. ~2–4 days, disproportionate payoff: "your memory stays yours, export in one command." And the git-diffability is the feature: a scheduled export committed to a branch makes every consolidation/decay/promotion cycle a reviewable diff — "what did the agent learn this week" becomes git log, and memory gains a rollback story.

W10

Doc & claim truth-up

Land the doc-staleness reconciliation; keep benchmark-claim discipline; keep the AI-memory-landscape doc accurate (retrieval-vs-QA framing, forgetting/decay reality).

Exit

Published QA-accuracy and LongMemEval-V2 numbers with citable receipts; a regression gate blocks PRs on quality drops; the cost/latency curve is published; the usage loop is closed (cited-vs-uncited decay divergence demonstrated); forgetting is uniform, usage-aware, and benchmarked; the team scope is enabled with proven isolation, a way to create/populate teams, and a promote/SHARE path; OKF export ships with the scheduled memory-changelog mode.

II
v1.2

Coalesce It — safely

Turn the offline/online merge primitives into a live, provably-scoped, reversible, provenance-preserving team-memory coalescence engine — with a first-class data model and deterministic conflict states. Cross-user entity resolution is the field's unsolved problem; the differentiator is doing it without ever leaking scope and always reversibly. Contributors are writers, not just humans: the wedge is one operator running agent swarms — the same coalescence problem with zero privacy risk, dogfoodable on day one — and human teams land on the same records.

W1 · build first

Coalescence data model & reversibility

Before merging anything, define the model: canonical entity vs contributor aliases vs contributor assertions; an attribution schema; a conflict lifecycle with deterministic states (open / merged / superseded / contested); split/undo (every merge reversible); revocation semantics; redaction/anonymization transforms; and a human review UX for contested merges.

W2

Live coalescence engine

Unify online retrieval/dedup.py and offline migrate/merge.py into a live cross-contributor entity-resolution + relationship-redirection engine scoped to a team space, emitting the W1 records. Provenance-preserving — contributor assertions are never destroyed.

W3

Concurrent multi-writer consistency

Additive-safe vs conflict-checked write semantics (Letta-style memory_insert / memory_replace); bi-temporal edge invalidation for contradictions (Zep-style validity windows — invalidate, don't delete). Deterministic conflict states, not silent merges.

W4

Eval team memory at scale

Wire multi_user.py into a CI gate and define the load matrix explicitly: N orgs × users/org, corpus size, QPS, concurrent writes, with cache/pool under stress (more than surreal_graph_client_cache_size = 64 hot orgs). Replace the 57-record rehearsal with adoption-grade corpora; resolve the filtered-HNSW recall@k=0.0 finding at scale.

scale-load-gate at the defined matrix: p95 ≤ budget, error rate = 0, leak_count = 0, recall ≥ floor.
team-isolation-under-load-gate concurrent multi-tenant load + revocation-under-load produces zero cross-tenant / cross-scope leakage.
W5

TeamMemBench (internal)

Build the internal benchmark for cross-user entity resolution, concurrent multi-writer consistency, and "helped AND stayed scoped" (utility + access control + forgetting, GateMem-style). Dogfood seed: the eternia/macbook merge. Internal-only this release — publication waits for v1.3; synthetic-only risks "benchmark theater." The utility axis is outcome-grounded: tasks and memories share the graph, so "helped" means cited during tasks that completed — the ground-truth utility metric no public memory benchmark has.

W6

Distillation pass — the per-project handbook

Mechanical merge keeps N records; distillation compresses knowledge. A dream-cycle stage maintains a distilled per-project handbook + wake summary (built with synthesis_plan/draft/verify), regenerated when the graph changes enough. Wake bundles serve the distilled artifact instead of raw top-k; the v1.1 usage signal tells the distiller what earned prominence; the write-path-integrity gate applies to the distiller too.

W7

Materialized memory-as-files

sibyl export --project + a session hook materialize the handbook + recent context pack into .sibyl/memory/ read-only files. Filesystem-native agents grep at zero marginal latency and survive server outages; citations still route usage back over the API. Structured substrate, curated projection — Sibyl is the governed source of truth behind the flat-file UX the frontier converged on.

W8

Retroactive re-extraction loop

Verbatim raws + lineage + a versioned extractor mean every extraction improvement can replay over the whole historical corpus — memory quality compounds retroactively, a property no competitor has. sibyl admin re-extract --since-extractor-version re-derives, diffs against current, and scores the delta with the v1.1 regression harness before promoting; supersession edges keep the old derivation for rollback.

Exit

Reversible coalescence with deterministic conflict states and attribution, proven on the agent-swarm dogfood; isolation proven under concurrent load and revocation; internal TeamMemBench passing with a documented scale envelope and an outcome-grounded utility metric; distilled handbook + files projection serving wake bundles; one re-extraction replay executed end-to-end with a scored, gated delta.

III
v1.3

Lead It

Publish the benchmark the field lacks, and push the frontier — only from a position of proven isolation and honest numbers.

Publish

TeamMemBench (dataset + leaderboard)

Only with a defensible dataset (real-team-log or a rigorously-justified hybrid) and at least one external comparator. Category-defining if done right; theater if rushed.

Frontier

Frontier retrieval

MAGMA-style multi-graph disentanglement (seam decided in v1.2); an optional cross-encoder reranker path for diffuse-evidence questions; belief-revision semantics (Kumiho/AGM) if observed conflicts justify it; procedural/skill memory; outcome-conditioned memory valuation — decay driven by task-outcome lift, the strongest form of "unused = dead."

Reach

Platform reach

Be the MCP-native, OKF-speaking backend behind Claude's /memories and other agents; SurrealDB live queries → reactive UI; SurrealDB Cloud managed multi-tenant; the Haven lighthouse integration; the adjacent Rust high-throughput runtime.

Interchange

Interchange & round-trip

An OKF importer (needs LLM edge-inference from prose links); watch the W3C "DataBook" RDF/SPARQL profile as a better typed round-trip target than plain OKF. Keep graph.json as the lossless internal archive; OKF is the portable public projection.


Sequencing Rationale

Why evals come before the feature.

Team-memory coalescence is isolation-correctness-under-merge: it cannot ship credibly without the harness (v1.1) to prove isolation and quality, and the eval gaps are table-stakes the field judges on. v1.1 also front-loads the honest-benchmark reputation before the v1.2 bet. v1.1's team substrate (scope enable + team CRUD + promote) is the minimum for v1.2's engine to stand on. Within v1.2, the reversibility/attribution model (W1) precedes the merge engine (W2) so nothing irreversible ships. The public benchmark waits for v1.3 because a bad dataset is worse than none.


Decisions Still to Lock

Four forks that shape the arc.

Open · team roles

Team role model

Reuse project roles for teams, or define a distinct team-role set?

Open · structure

Memory spaces

Hierarchical, tag-based, or both?

Open · gates v1.3

TeamMemBench dataset

Real-team-logs (privacy-heavy), synthetic, or hybrid — this gates whether v1.3 publication is viable.

Open · boundary

Cross-org sharing

In scope for this arc, or explicitly out — org stays the hard boundary and team memory is within-org only?


Built On The Literature

The work this vision stands on.

Every load-bearing claim traces to a source. The 2026 preprints are new and often single-source — we treat their exact numbers as provisional and their structural findings as solid.

Team / shared memory

  • Governed Shared Memory (ArgusFleet) — arXiv:2606.24535
  • GateMem — arXiv:2606.18829
  • Collaborative Memory — arXiv:2505.18279
  • HiddenBench — arXiv:2505.11556

Temporal graph & retrieval

  • Zep / Graphiti temporal KG — arXiv:2501.13956
  • Microsoft GraphRAG — arXiv:2404.16130
  • MemGPT · sleep-time — arXiv:2310.08560 · 2504.13171

Forgetting & consolidation

  • Memory Worth (when to forget) — arXiv:2604.12007
  • Memora / FAMA — arXiv:2604.20006
  • ForgetEval — arXiv:2606.15903

Benchmarks & field map

  • LongMemEval · V2 — arXiv:2410.10813 · 2605.12493
  • BEAM (≤10M tokens) — arXiv:2510.27246
  • Agent-memory survey — arXiv:2603.07670