Prompts as versioned, auditable building blocks.

In AIScholar, a prompt is never a loose blob of text. It is a template with placeholders, filled by named snippets — project-scoped text blocks that each carry multiple indexed versions (v0, v1, v2…). A template references a snippet by name, like {{framing}}-style tags, and each snippet version is bound to one or more levels of a design factor.

At execution time the platform resolves each condition by injecting the snippet version bound to that condition's factor levels. This is the only channel through which a factor level becomes concrete wording — which means the mapping from design to text is explicit, inspectable, and impossible to fudge silently. Single-version snippets act as fixed, shared text across all conditions, so boilerplate stays consistent everywhere.

1. 1

   Write or generate snippets

   Create named text blocks in a card-based snippet manager, or bulk-edit versions in a spreadsheet-style grid. Each version of a multi-version snippet realizes one level of a factor.

2. 2

   Bind versions to factor levels

   Attach each snippet version to the (factor, level) pairs it realizes. The platform validates coverage: every level of every factor a template varies must have a bound version.

3. 3

   Compose templates

   Build one or more prompt templates that reference snippets by name. Templates that reference a factor name directly are flagged invalid and blocked from save and run — text must flow through snippets.

4. 4

   Preview the variation matrix

   A combinatorial matrix shows every fully-resolved prompt your design will produce, with snippet text highlighted so you can see exactly what varies between conditions. Select or deselect instantiations before running anything.

5. 5

   Version on every edit

   Editing a saved template creates a new version rather than overwriting. Trials record which template version they used, so results never point at text that has since changed.

• Snippet library

  Card-based manager for all project snippets: names, descriptions, versions, and the factor levels each version realizes.

• Variation grid

  Spreadsheet-style bulk editing of snippet versions across factors and levels — the fastest way to author a full design's worth of wording.

• Template versioning

  Every template edit is a new version; every trial records the version it ran. Nothing is lost, nothing is ambiguous.

• Live preview with highlighting

  Template previews highlight snippet insertions in blue and surface invalid factor-name placeholders in red, so structural errors are visible at a glance.

• AI prompt wizard

  Generates a complete template-and-snippet set from your design using a creator→verifier loop that catches unbound snippets, uncovered factors, missing level versions, and unknown variables before you see the result. Residual issues are surfaced as explicit warnings.

• Prompt variants

  Auto-generate semantically equivalent phrasings along six perturbation dimensions — paraphrasing, structure, verbosity, output format, instruction specificity, and persona — so wording becomes a measurable random effect rather than a hidden confound.

With templates validated and the variation matrix reviewed, the study is ready to run — see experiment execution for how trials are dispatched at scale.