The live tool is not up yet. The classification logic exists inside the full Ora system and runs every time the orchestrator routes a question to a framework; the standalone browser version exposes that routing decision as a single-purpose service. This page describes what the tool does and the methodology behind it.
What it does
The Classifier takes a free-form description of the question or problem the visitor wants to think about and returns three things. First, the recommended primary framework, with a short explanation of why that framework fits the shape of the question. Second, alternative frameworks ranked by closeness of fit, with a sentence on each about what dimension of the question they would emphasise instead. Third, the complete prompt — the recommended framework's instructions concatenated with the visitor's question — formatted and ready to paste into the commercial AI of the visitor's choice.
The third output is the operational point. The Classifier is not just a recommendation engine; it produces an artifact the visitor can use immediately, in whichever AI they already have a relationship with. The framework instructions plus the question, composed into a single prompt, give the AI the analytical scaffolding the question needs.
Why this exists
Ora's framework library carries more than forty analytical frameworks organised across five super-clusters: Argument and Reasoning, Causation, Decision and Future, Position and Strategy, Synthesis and Generation. Each framework is calibrated for a specific kind of question and produces a structured output specific to that kind. Knowing which framework to apply to a given question is itself a meta-skill — and a non-trivial one, because the same surface question can have very different analytical shapes underneath depending on what the visitor is actually trying to do.
A question that looks like "is this argument any good" might be a coherence audit (does the reasoning hold up internally), a frame audit (what is the framing doing), an interest analysis (who benefits from this argument's acceptance), or a hypothesis evaluation (does the evidence actually support the claim) depending on what the visitor cares about. The right framework for each of those is different. Without the Classifier, the visitor has to learn the territory taxonomy and the framework library before they can deploy a framework usefully; with it, they can describe the question in their own words and get routed correctly on the first attempt.
How it works, operationally
The visitor pastes or types the question they want to think about — as briefly or as fully as they want; the tool reads the shape of the question, not just the surface phrasing. The Classifier runs a territory-mapping pass over the framework registry: which super-cluster does the question live in, which framework within that super-cluster matches the question's depth and specificity, what disambiguation questions might still be live. It returns the recommendation with reasoning, the ranked alternatives, and the composed prompt.
The visitor takes the composed prompt and pastes it into whatever AI they use. The output they get back is the framework's structured analysis applied to their question — same operation Ora would run if the visitor had the full system installed, with the trade-off that this single-shot version cannot draw on a persistent vault, cannot hook into adversarial review, and cannot escalate to a combined framework when the question is genuinely cross-cutting. For many questions, the single-shot version is enough. For the ones where it is not, the Classifier output is at least a clean orientation toward what the deeper version would do.
The underlying methodology
The Classifier runs a territory-mapping operation over the Ora framework registry — the same operation Ora's internal dispatch uses to decide which framework to invoke when a question lands in the orchestrator. The registry is organised first by super-cluster (the broad kind of cognitive operation), then by territory within the cluster (the specific class of question), then by mode within the territory (the depth and specificity of treatment). Each mode carries an explicit "when to use this" specification with plain-language triggers, boundary conditions, and the kinds of question it does not fit.
Classification matches the visitor's question against those triggers, considers the alternative routes the question could take, and surfaces the trade-offs rather than collapsing them. When two frameworks are both plausible, the Classifier names both and explains the dimension each would emphasise; the visitor decides which dimension matters more for their purpose. When the question is cross-cutting, the Classifier flags it as cross-cutting and recommends the primary framework with a hook to the secondary one.
What the tool will not do
The Classifier does not run the framework on the question. It routes the question to the right framework and hands the visitor the prompt; running the framework is the visitor's next step, in the AI of their choice. This is deliberate. The Classifier's job is the routing decision; running the framework is a separate operation that benefits from the visitor seeing the full output rather than getting a summary buried inside the routing tool.
The Classifier is also not a problem-solving substitute. If the visitor's question is well-defined and they already know which framework to apply, the Classifier adds latency they do not need. The tool is for the case where the visitor has a question and is not yet sure how to operationalise it as analytical work.
Related
- The framework library — browse the full collection the Classifier routes among.
- The methodology — for the architectural treatment of how the territory taxonomy and the framework registry compose.
- Propaganda Analyzer — for the specific case of structural analysis of a media artifact, when that is already known to be the right framework.
- Download Ora — for the full orchestrator, when single-shot prompts stop being enough.