AskTribune Terminal (Beta)

AskTribune Terminal (Beta) is an Australian political and policy intelligence tool. You ask questions about the public record — parliamentary Hansard, ministerial media releases, and other primary sources — in plain language, and get answers grounded in the record and cited, so you can check them. It's made by Tribune Analytics, an Australian company.

In open beta, AskTribune draws on the Commonwealth public record from 1 March 2026 onwards:

• House of Representatives Hansard — speeches, questions, and debates.
• Senate Hansard — speeches, questions, and debates.
• Prime Ministerial media — press releases, transcripts, and statements.
• Ministerial media — releases and statements across the ministry, with rolling week-to-date briefings on every portfolio.

The roadmap adds parliamentary committee submissions and primary data sources such as the Australian Bureau of Statistics (ABS) and the Australian Electoral Commission (AEC). The retrieval framework is built to absorb new record types without changing how you ask.

Records are collected directly from official primary sources on a regular schedule and added to the corpus throughout the day. New Hansard sittings, media releases, and ministerial statements typically appear in the corpus the same day they're published, or the following morning — depending on when each source releases. Late corrections and proof-to-final updates are reconciled automatically so the working record stays accurate.

The corpus is also growing backwards. The open beta started with records from 1 March 2026 onwards, and historical material is being added progressively so the same retrieval and Note framework can cover a deeper window over time.

Every answer is built from the public record, not the model's memory, and the records behind it stay visible so you can check the work. The same question returns comparable results regardless of which party or politician is involved — nothing in the system favours a side. AskTribune reports what the record shows; it doesn't assign motives or take positions.

Our Civic Use Principles describe the constitutional posture in full.

AskTribune is built around retrieval and connection rather than model memory. As each new record comes in, the system processes it into a structured catalogue — who spoke, when, in what setting, on what topic, and how it sits alongside other records — and stores those connections next to the source text. When you ask a question, the system finds the relevant records using both their content and their relationships, assembles a working set, and asks the model to answer from that set with citations back to source.

The split is deliberate: the model's job is to read and reason; the retrieval framework's job is to make sure it's reading the right material. We don't publish the specifics of how the catalogue is built — that's the part of the system we keep refining — but the principle is straightforward and stable: every answer should trace back to records you can open and check.

An AI model is the kind of software that powers tools like ChatGPT — trained on enormous quantities of text so it can read, write, summarise, and reason in natural language. (You may also hear these called large language models, or LLMs.) AI models are very good at handling language, but on their own they have no live access to a specific, current set of records — they answer from what they happened to learn during training. AskTribune Terminal (Beta) pairs AI models with a dedicated retrieval framework so the language work happens over the right material.

Retrieval-augmented generation is a way of using an AI model that gives it the relevant source material at the moment you ask the question, instead of relying on what it learned during training. The model reads the supplied material — in AskTribune's case, Australian public records — and answers from there, with citations back to source. RAG is what turns a general-purpose model into a system that can answer reliably and verifiably about a specific, current corpus. Most of AskTribune's engineering effort goes into making the retrieval step as accurate and well-connected as possible.

In a graph-based system, a node is a discrete thing — a record, a speaker, a portfolio, a piece of legislation, a place, a date — and an edge is a relationship between two nodes (a minister speaks on a bill; a bill is debated on a sitting day; a sitting day sits in a portfolio). Modelling the public record as a network of nodes and edges lets a question travel between connected ideas rather than depending only on text similarity. As we transform AskTribune's retrieval substrate toward a graph, nodes and edges become the underlying grammar of how records, people, and themes are linked.

Computational political science applies computer-science methods — large datasets, retrieval, structured modelling, network analysis — to political and policy questions. AskTribune sits in this tradition: we use computation to make the public record searchable, connected, and useful at speed, without taking human judgment out of policy or politics. The goal is to give analysts, journalists, advocates, and citizens better tools to work with — not to automate the thinking, and not to replace the reading.

AskTribune Terminal is a live open beta, and the framework underneath it keeps evolving. The directions we're actively building include:

• Spatial reasoning. Linking records to Australian places — electorates, regions, sites, jurisdictions — so a question about a policy area or a piece of news automatically picks up local context: who represents the area, what's been said about it in Parliament, what primary data tells you about it. Geography becomes a first-class dimension of retrieval, not an afterthought.
• Richer Notes. Notes are the structured briefings the system writes for each sitting day and across portfolios. We're expanding the Note format to cover more angles — committee work, week-to-date summaries by minister and shadow minister, and cross-portfolio threads — and tightening the editorial pass so the briefings stay tight, faithful, and easy to verify.
• Graph-based retrieval and concept hopping. Today's retrieval grounds answers in semantic search over records and Notes. We're transforming the substrate underneath into a graph where records, speakers, themes, bills, places, and dates are explicit nodes with explicit relationships. The payoff is better concept hopping — when you ask about one issue, the system can traverse the connections around it instead of relying on text similarity alone.

A general AI answers from its training data, which can be out of date, unsourced, or simply wrong — and you can't see where the answer came from. AskTribune retrieves the actual Australian records relevant to your question and cites them, so every answer traces back to a primary source you can open and verify. You get ground truth, not a confident guess.

Modes let you choose the kind of work you want — and each puts the right model on the task:

• Answer — quick, focused replies grounded in our daily Notes. Powered by Claude Haiku 4.5.
• Research — wider retrieval across records and metadata, suitable for desk work. Powered by Claude Sonnet 4.6.
• Analysis — deep synthesis across multiple records and themes. Powered by Claude Opus 4.6.

Chat is the conversational surface in the centre of the Terminal — you ask in plain language, AskTribune retrieves and answers. Notes are the structured briefings the system writes from each sitting day and each portfolio: an executive summary plus the records and reasoning behind it. Open a Note to load its records into your session for deeper Research or Analysis, or read it on its own to catch up on a day.

You start with a free 7-day trial — 30 credits, Answer-mode only, no card required.

After that you can:

• Pay as you go — $10 AUD gets you 100 credits, valid for three months. Buy more whenever you need them.
• Subscribe — Tribune Monthly ($19.99/month) or Tribune Quarterly ($50/quarter). Both include 40 credits each week, unused credits roll while the subscription is active, and subscribers unlock larger top-up packs with bonus credits.

You only pay for what you use. Pre-query estimates show the credit cost before you commit. Billing is AUD only. Final tax treatment depends on your circumstances — check with your accountant.

Bring Your Own Model is part of how the Terminal is built: because Tribune's value is in retrieval and connection over the public record — not in owning the model — the framework is designed so you can plug in your own AI model and your own compute instead of Tribune's. The shape of the answer doesn't change; the source of inference does. BYOM lets enterprise users control cost, keep inference on their own infrastructure, and meet data-residency requirements.

What you create in Terminal is yours — including anything you generate with an AI model. The source records are public records provided for fair use.

We use third-party AI providers to deliver answers, which means your queries are processed through those providers under their terms. Our Terms of Use and Civic Use Principles cover data handling, anonymity, and disclosure in full.

Anyone who needs to understand the Australian public record accurately and quickly — policy and government-relations professionals, journalists, researchers, advocates, and engaged citizens. Built for professionals. Open to everyone.