Product preview

Ask the Archive

This archive is designed for a new era of digital media: public records that can be searched, questioned, challenged, and explored through an evidence-bounded AI assistant. The assistant is not a judge, prosecutor, or narrator of unsupported claims. It is a guide to the record.

This is a placeholder product page. No live chatbot script or API integration is deployed here yet.

What the assistant will do

It will help readers move through the archive without losing evidence discipline.

  • Answer from approved public-site content and approved publication_review summaries only.
  • Explain evidence labels and claim posture.
  • Point to timelines, contradictions, court records, and branch reviews.
  • Identify what is missing when a record is incomplete.
  • Link readers to right-of-reply and corrections pages when the question is sensitive.

What it will not do

Boundaries are part of the product

Evidence labels

The assistant must always label claim posture

CONFIRMED

Directly supported by approved public content or approved publication-review summaries.

CORROBORATED

Supported by more than one approved source, with limits disclosed.

ALLEGED

Asserted in a petition, complaint, filing, source note, or similar record, but not adjudicated or independently proven.

HYPOTHESIS

A testable theory or bridge lead, not evidence.

PROCEDURAL

A filing, docket event, motion, or case-management step.

NOT ADJUDICATED

The reviewed packet does not contain a merits determination.

DISCONFIRMED

Reliable approved sources contradict the claim.

Example questions

Representative questions the assistant should handle

Suggested questions / FAQ

Good starting prompts for readers

What does the Tarrant County petition allege?

The assistant should summarize the petition as allegations only, label the posture ALLEGED / PROCEDURAL / NOT ADJUDICATED, and link to the public court record page.

What is confirmed about the 1995 Texas Petroleum / Omimex / Sabacol records?

The assistant should answer only from approved evidence pages and should separate confirmed record points from unresolved gaps.

What remains unresolved in the telecom bridge hypothesis?

The assistant should explain that the telecom item remains a SEC-comparable lead and not primary FCC proof.

What did the Omimex / ECMC bankruptcy ingest recover?

The assistant should summarize the procedural bankruptcy facts and the source-backed ECMC dispute without inferring causation or liability.

What did the bankruptcy ingest not find?

The assistant should say that no source-backed Naresh Kumar Vashisht reference was found in the reviewed bankruptcy originals.

How can an affected party submit a correction or reply?

The assistant should point to the right-of-reply and corrections pages and avoid making any factual conclusion about the submission itself.

Sensitive-claims protocol

When the record is incomplete, the assistant should say so

The assistant should refuse unsupported accusations, redirect users to the relevant page, and include correction and reply links in sensitive answers.

Examples of safe phrasing: “The record says...”, “The petition alleges...”, “What is missing is...”, “This remains a hypothesis...”, “See the timeline / contradiction card / publication_review summary.”

Right of reply Corrections

Future implementation options

Planned retrieval-augmented architecture

A future version of Ask the Archive should use only approved public pages, sitemap content, evidence tables, timeline entries, contradiction cards, and publication_review summaries. The assistant should retrieve snippets from those approved sources, label each claim posture, and present links to the underlying pages.

The intended architecture is:

static public site → indexed approved documents → retrieval layer → evidence-bounded assistant UI

No third-party chatbot script is added on this page yet, and no API keys are exposed.