Discuss a pilot

TRACKING INFRASTRUCTURE FOR TEXT THAT CHANGES

The evidence.
The context.
The moment.

Datasets Frontier is tracking infrastructure for text libraries that change over time. It is designed to index every edition and bring the exact source text forward, word for word, with its clause, edition and date; each result carries its source reference, and the extraction that captures the source text uses no LLM.

Discuss a pilot See how it works

This page describes how the platform is designed to work.

A steel ring with a glowing data core orbits above a night city of documents on the water
Source-linked resultsEdition-aware contextDesigned to work alongside your systems

01 / MAKE THE EVIDENCE VISIBLE

Quoted word for word, or marked as not verified

See the original page, the exact wording and the edition in force, side by side. If a match can't be verified, it stays marked unverified.

EVIDENCE VIEWOriginal and result, side by side
Sample data
01

Original document

S13C3 Metal pipe systems

(3) A pipe system comprised entirely of metal (excluding pipe seals or the like) must be wrapped but must not be lagged or enclosed in thermal insulation over the length of its penetration of a wall, floor or ceiling unless the lagging or thermal insulation fulfils the requirements of S13C7. (4) The gap between a metal pipe and the wall, floor or ceiling it penetrates must be fire-stopped in accordance with S13C7.

Document and edition
NCC 2025 Volume One (Building Code of Australia)
Clause and page
Specification 13, S13C3(4), page 209
02

Source-linked result

EXACT EXCERPT · VERIFIED (EXAMPLE ONLY)
“The gap between a metal pipe and the wall, floor or ceiling it penetrates must be fire-stopped in accordance with S13C7.”
Point in time
In force from 1 May 2026 · Victoria
Verification
Matched to source page and clause
Changes and reason
Not listed as a major change in NCC 2025 · Reason not documented

If the source can't be found, the result says so. Verification is being validated with pilot partners; results are drafts for a qualified person to review.

Read the verification standard

The National Construction Code 2025 was provided by the Australian Building Codes Board under the CC BY 4.0 licence .

02 / ONE INFRASTRUCTURE, MANY LIBRARIES

Any text library that changes over time

Building codes and standards are the first library. The same template fits other text libraries.

01

Standards and building codes

The first library: standards and building codes, tracked edition by edition.

Bring
Standards, codes and edition history.
Review
Exact source text with its clause, edition and date.
Evaluate
Faithful extraction across every edition in the library.
Scope this use case
02

Legal PDF libraries

A library of about 1,000 legal PDF files can be tracked the same way, document by document and version by version.

Bring
Legal PDF files and their version history.
Review
Exact source text with its document, section and version.
Evaluate
Faithful extraction across revisions and amendments.
Scope this use case
03

Policies, specifications and contracts

Any text library that moves through editions fits the same template, inside the systems you already use.

Bring
Policies, specifications and contract documents.
Review
What changed, when it changed and where it applies.
Evaluate
Coverage and tracking across the library.
See the coach and services

The template is reproducible, so a new data set is designed to cost much less to stand up for the next library.

03 / COACH AND SERVICES

Understand the tracking, then design the workflow

The coach uses an LLM to explain how the tracking works and the key points an edition-over-time end point needs; it may read the files you choose to give it. Its answers are written by an LLM, so check any text it quotes against the source; the extraction that captures the source text uses no LLM. Around the infrastructure sit these services.

  1. Succinct reports

    Reports with LLM-assisted structuring, where the AI structures the report and the report is designed to keep the cited text verbatim from the source.

  2. JavaScript workflows

    Workflows and packages for audits and reporting, built to run beside the systems you already use.

  3. Bespoke and data services

    Bespoke workflows, plus data clean-up, recovery and integration for asset management software.

04 / ANSWERS FOR YOUR EVALUATION

What your security and procurement teams need to know

Request our security overview

05 / DEFINE A USEFUL FIRST STEP

Start with one workflow and agree what success looks like

A pilot discussion can define the sources, users and outcome that matter to your organisation.

  1. 01
    Define the task

    What needs to improve, and who needs the result?

  2. 02
    Agree the evidence and boundaries

    Which sources may be used, and which systems may be read or changed?

  3. 03
    Set an evaluation

    Agree measurable acceptance criteria and how findings will be reviewed.

To discuss a pilot, email [email protected].

What happens to what you send is in our privacy policy.

THE ORGANISATION BEHIND THE PLATFORM

Datasets Frontier

Read our source-verification approach