Methodology

How the Continuity Map Works

This project keeps a public map of AI continuity, memory, and stateful-agent systems. This page explains how entries get on the map, how they are categorized and re-checked, and what to do if something here is wrong. For the entries themselves, see Signals from the Field and the Research library.

What gets included

An entry earns a place on the map when it meets all three tests:

It is public. There is a primary source anyone can read — a paper, a repository, a dataset, a product page, or a published essay. Private or invite-only claims are not mapped until there is something to point at.
It touches continuity. It bears on how AI systems carry state forward across sessions, tools, or models — memory, retrieval, long context, agent state, provenance, or evaluation of any of these.
It can be verified. The claim can be checked against its own source. Where a claim is the author’s own framing rather than an independent result, the entry says so.

Source types

Every entry is tagged with one source type. The counts below are read live from the map’s single data source.

Lab08: An organization or research group building or studying continuity and memory systems.
Paper04: A research publication — preprint or peer-reviewed — making a testable or positioning claim.
Tool03: Installable software, a library, or a package that implements memory or continuity.
Benchmark09: An evaluation dataset, suite, or scoring harness for memory or continuity properties.
System05: A named, deployed system or architecture, usually with a public write-up or evaluation.
Essay04: Field commentary — an argument or opinion piece, read as argument rather than independent research.

Continuity roles

The map reads adjacent systems by the job they perform, not by the marketing word they choose. Memory stores. Retrieval finds. Agents act. Runtime executes. Continuity governs what carries forward. A source can matter to the field without claiming to be the whole continuity layer.

Status vocabulary

Status is a reading of how a source stands today, not a ranking or a score. It changes as the public record changes.

Status measures field maturity — how established a source is. It is a different axis from claim strength: how much a specific claim has been shown. Claim strength is what the A–E evidence ladder measures. A mature source can still carry a weak claim, and a new one a well-evidenced result.

ActiveCurrently maintained and publicly moving — recent releases, commits, or writing.
EmergingNew or early — a real artifact exists, but the track record is short.
MatureEstablished and widely referenced; changes slowly.
UnclearReal enough to note, but the public record is thin or unverified.
WatchlistTracked for context or language overlap, not asserted as a peer system.

Update cadence

The target is weekly source review; the changelog shows what was actually done. A target is not a claim that this cadence has been maintained.

Dated source checks and notable corrections are recorded when completed.

Re-verification is also what moves a source through its lifecycle. Some entries have lifecycle notes in the source data. The terms below describe the review framework, not a claim that every page displays a lifecycle label. This is separate from the map-status above, which reads how a source stands in the field. The lifecycle state is the evidential standing of the specific result: whether it has been reproduced, disputed, superseded, or withdrawn. Nothing is deleted when it changes — superseded and retracted sources stay on the record with the change noted, so a correction is visible rather than silent.

Three scales run through this site and they are not interchangeable. Map status (Active, Emerging, Mature, Unclear, Watchlist) measures how settled a part of the field is. Lifecycle (below) records what has happened to a source. The evidence ladder on the Thesis (A–E) measures how strong this project’s own claims are. A source can be Mature in a young field and still rest on a claim this project has not recorded — so a map-status token is never used as an evidence verdict.

Current: At its last check, nothing about the result is contested or stale.
Awaiting-replication: A reported result for which this map has not established independent reproduction. This is a limit of our review, not a claim that no replication exists.
Disputed: A headline claim has been publicly challenged or corrected; treated as self-reported until independently reproduced.
Superseded: A newer source has replaced it as the reference for its claim. Kept on the record, not deleted.
Retracted: Withdrawn by its authors or venue. Kept on the record with the retraction noted.
Historical: Foundational or of-its-time — still cited for context, no longer the live state of the art.

Correction policy

Entries are written to be corrected. If a profile is wrong, out-of-date, or unfair, the fix is a first-class change: the entry is updated, its last verified date is bumped, and the change is noted. Where a headline result has been publicly disputed or later corrected, the entry links the dispute — one of the open cases gathered at open questions in AI memory — and treats the claim as self-reported until independent reproduction exists.

The public field-map data and glossary data are generated from the same source as these pages. Corrections are welcome from anyone, including the maintainers of the systems profiled here. The standard is the same in both directions: point to a source.

To send a correction, email thecontinuitylayer@gmail.com or use Submit a Signal.

Independence & disclosure

The Continuity Layer is built and maintained by one independent builder who also does commercial work in this field. The System, Governance and Lab pages describe the maintainer's own research prototype; they are project reports, not independent evaluations. The disclosure is published on the homepage and on the Kenotic Labs profile, because its thesis paper shares this site’s name and its work overlaps the maintainer’s.

Where this project states its own results, it says so plainly and separates them from independently verified claims. The aim is a reference the field can rely on — including the parts of the field this project competes with.

The full standard every entry is held to — independence, even-handedness, corrections, and scope — is set out in the project’s public charter, the Constitution.

For who keeps this map and the standards it’s held to, see Who maintains The Continuity Layer.

Cite this page

Last updated: 2026-10-21

Maintained by: The Continuity Layer

How to cite: Cite the specific page and section (e.g., “The Continuity Layer, How the Continuity Map Works”), not the site as a general source. Attribute to The Continuity Layer.

Canonical URL: https://thecontinuitylayer.com/how-the-map-works

Suggested citation: The Continuity Layer. “How the Continuity Map Works.” https://thecontinuitylayer.com/how-the-map-works (last updated 2026-10-21).

Continuity layer — this site’s working definition: A continuity layer is the part of an AI system that carries forward governed state across sessions, tools, models, and hosts: what is currently true, what changed, why it changed, by whose authority, and which evidence supports it. It sits above memory and retrieval, and below agent execution, turning stored history into usable, auditable continuity; and it can correct, supersede, revoke, expire, and forget what it carries, with evidence that it did.