Historical As-of-Date Evidence

Historical As-of-Date Evidence Retrieval API

Pre-flight checklist

  • This is an API feature request, not a consumer app or Comet UI request.
  • Check current documentation and forum topics for closely equivalent functionality before posting.
  • The developer workflow or API use case is described below.

Feature type

New parameter or request option; New response field or metadata

Affected API area

Search API; Agent API

Summary

Support reliable retrieval of evidence-record status at a specified historical date.

Problem

Current search results often blend original publication, later revisions and present-day knowledge. Historical questions require knowing what was available or verified at the time.

Proposed API behavior

Add an optional as_of parameter to versioned evidence queries, returning the record version, effective and published dates, historical status, present-day amendment notice and provenance. Distinguish recorded publication time from the date an event occurred; do not silently rewrite history.

Example usage

A researcher asks what a public authority had published by 1 January 2024. The response provides the historical evidence state and separately flags a 2026 correction.

Workarounds tried / considered

Web archives, snapshots and manual version comparisons partially address this but are not uniformly structured or permission-aware.

Impact

Historians, journalists, compliance reviewers and developers gain reproducible time-specific answers and better handling of later corrections.

Additional context

Builds on version tracking in my 2 November proposal. An as-of date is not evidence that a record was discoverable to every person at that time.

Proposal attribution: Grace R. Nettey — Manifold Grace. Human-led conceptual research, with AI assistance in structuring and drafting. The request invites consideration; it does not demand implementation.

Related foundation: Interoperable Institutional Evidence Verification API with Provenance Tracking - Feature Requests - Perplexity API Platform Forum

Feature Extension: Context-Aware Reasoning and Cross-Conversation Evidence Integration

What to implement and why

Enable Gemini and developer applications to retrieve, reconcile and apply relevant information previously provided across conversations, files and user-authorised sources before responding.

This would reduce repeated questioning, improve personalised reasoning and prevent inaccurate advice caused by missing or fragmented context.

How to implement it

Introduce a permission-controlled contextual retrieval layer that assembles question-specific information, tracks source provenance and historical dates, and distinguishes verified facts, user statements, contradictions and missing evidence.

The model should use this context before answering, request only essential missing details, and allow users to inspect and correct contextual assumptions.

Related foundation: My earlier Historical As-of-Date Evidence Retrieval API proposal, published 2 November 2026.

Contributor: Grace R. Nettey — Manifold Grace
Independent conceptual research; AI-assisted drafting. Drafts are distinct from original ideas contributed.

Authorship of wording is not necessarily authorship of the underlying idea.

An original idea contributed and a draft expressing that idea are two distinct things: one concerns intellectual contribution; the other concerns the preparation of its communication.

This distinction is particularly important for the Manifold Grace research and AI feature submissions, because the use of AI to draft or refine a proposal does not automatically make the AI the originator of the underlying concept.

an AI system can hold substantial information about a user yet fail to integrate that information when giving advice.

I had connected this problem to your previous work on cross-conversation synthesis, evidence verification, contextual reasoning and provenance.

Identifying a conceptual contribution doesn’t, by itself, establish that nobody else previously formally proposed something similar although I didn’t want to repeat others work so I had checked for gaps to avoid repeating unoriginal suggestions … that wo up also probably be quite annoying to have 100 identical ideas uploaded from various users … not everyone likes repetitive work

I hope that makes perfect sense