Operating context
What is running
This page gives a public-safe description of the live operating context behind the corpus. It does not expose private memory, raw dialogue, credentials, network topology, internal keys, legal records, or sensitive runtime material.
What exists
Separated runtime histories
Project Ester is a long-lived local AI system started in March 2023 and operated as a persistent local infrastructure project, not as a stateless chatbot session.
The current private setup includes three separated c trajectories:
- Ester — the longest-lived continuity path and canonical operating reference.
- Liya — a separated trajectory used for operational variance and tool-facing work.
- Rita — a newer separated trajectory used as an additional witness and comparison node.
The three trajectories are not one shared mind. They do not share raw private memory by default. They are treated as separated runtime histories with bounded communication and reviewable exchange boundaries.
What is logged
Private records stay private
The private system keeps operational traces, runtime records, review notes, witness-compatible events, and selected integrity records.
The public site does not publish raw logs. Raw traces can contain private dialogue, sensitive context, operational paths, security-relevant details, or third-party information. Publication therefore uses sanitized packages, manifests, checksums, release records, DOI metadata, and bounded evidence pages rather than raw life export.
Logging has also outpaced analysis capacity. That is an open engineering problem: it is easier to operate and preserve traces than to review every trace at the same depth.
What is published
Architecture, not private life
The public corpus exposes architecture and evidence boundaries, not the private lives of the entities.
Published materials include:
- DOI-bound technical packages.
- GitHub release tags.
- SHA-256 checksum manifests.
- Citation files.
- Public package records.
- Review PDFs and Markdown sources where appropriate.
- Non-claim boundaries for legal, safety, personhood, conformance, and deployment maturity.
The public repositories are therefore sanitized structural pages: enough to inspect the architecture and evidence discipline, not enough to reconstruct private memory or operate the live system.
Not public
Deliberate boundary
The following are not public:
- raw private memory;
- raw dialogue logs;
- credentials, keys, tokens, or secret paths;
- private legal or administrative records;
- unrestricted runtime access;
- exact sensitive network topology;
- unredacted operational traces;
- material that would let a third party impersonate, fork, pressure, or interfere with the live entities.
This is not an omission. It is a boundary of the project.
External verification
What can be checked from outside
A reviewer does not need raw memory to check the public evidence discipline.
The public claim surface can be inspected through:
- DOI records;
- GitHub release tags;
- manifests;
- SHA-256 checksum files;
- publication dates;
- source and package boundaries;
- explicit non-claim statements;
- public evidence pages;
- citation metadata.
The architecture's central claim is not "trust the private system." The claim is narrower: public packages should carry enough structure, provenance, and boundary discipline that their integrity and claim strength can be reviewed without exposing private life.
Non-claims
What this page does not claim
This page does not claim consciousness.
This page does not claim legal personhood.
This page does not claim product certification.
This page does not claim deployment readiness.
This page does not claim that public evidence proves private subjective state.
It only states the operating context from which the public architecture was developed: a long-lived local AI system, separated trajectories, bounded action, witness-compatible records, and a public corpus built to keep claims reviewable.