One Search Box Over Four Different Catalogues

A search that merges everything into one ranked list makes an implicit claim: that a career profile, an aptitude suite and an encyclopedia article are comparable things, and that the ranking knows which you meant. They are not, and it does not. A person typing a job title usually wants the career material; a person typing a chemical compound usually wants the source material; and the same three letters can mean either.

So the results are grouped by what they are. Assessments, Careers, and Guides and pages come from this site's own registries, and Wikipedia results come from the language edition you selected. Each row carries a label saying which kind of thing it is, so choosing between a suite and a skill lesson with the same name takes no guessing.

The assessment and career groups are built directly from the modules that own that content, the same ones that build the sitemap and the tool map. A suite published today is searchable today without anybody updating a search configuration, and a suite withdrawn tomorrow disappears from search at the same moment it disappears from the catalogue.

This is the difference between a search index and a search index that can lie. An index maintained beside the content drifts from it silently; an index derived from the content cannot.

The site groups are rendered on the server. Submitting the form with scripts disabled still returns real assessment, career and guide results, because the registries never ship to the browser in the first place and the ranking runs where they live.

That is not a nostalgic gesture. It is what makes the page usable on a throttled connection, inside a restricted browser, and to a crawler, and it is why the whole catalogue can be searched from a page that weighs almost nothing.

• Assessments: suites, skill constructs and sectors, when you are preparing for something.
• Careers: families, occupations and profiles, when you are exploring what work exists.
• Guides and pages: help articles, tutorials, blog posts, puzzles, course lessons and policy pages.
• Wikipedia: the source material itself, from the language edition you choose.

Pressing the keyboard shortcut opens the Novus Hub, which searches your own library and the dictionary alongside the site's destinations. That is the right tool when you want to get somewhere rather than to find out what exists: typing two words and pressing Enter is faster than any menu, and defining a word from wherever you are standing takes about a second.

The distinction is worth internalising. The search page answers what is there. The hub answers take me there.

The hub also reaches things the search page cannot, because it runs in your browser and can read this device's own storage. A project you built from an uploaded file has no public URL and will never appear in a site search, and it is one keystroke and two letters away in the hub. That asymmetry is a direct consequence of the local-first design rather than an inconsistency: the two tools see different halves of what you have.

Wikipedia is the largest thing this search can reach, and it is deliberately not the first thing shown. A query that matches an aptitude suite, a career family and an encyclopedia article is almost always a query about the first two: somebody searching a job title on a learning site is not usually looking for its encyclopedia entry.

The source results also behave differently from the site results, and the page says so. Choosing a Wikipedia result is the start of a process rather than a navigation: Novus Learn retrieves the page identity, the revision, the structured entity and the media attribution after you choose, which is what makes the resulting study space checkable. Listing them alongside site pages without that explanation would suggest they were the same kind of link.

• A construct name rather than a test name, if you are not sure which suite you need.
• A job title, which reaches the career profile, the guide and the suites mapped to it at once.
• A feature name in plain words, which reaches the help article rather than the feature page.
• A topic, if what you actually want is the source material and not the product.

If you do not yet have the vocabulary for what you want, search will not help, because search requires you to already know a word that appears in the thing. That is what the FAQ and the site-wide guides are for, and the help centre now has a page whose whole job is to point at the right door.