Guides

These guides cover common scenarios for using Ichnus, along with prompt techniques and honesty features that make responses more reliable. Each guide is short and focused on one workflow.

Education

• Course design: use Ichnus to draft lesson plans, rubrics, and study guides, then review and adapt them to your context.
• Student support: generate practice questions and worked examples in most subjects. Ask for hints instead of full answers to support practice.
• Accessibility: request plain-language summaries of dense readings, and use Corus to summarize long textbooks in one context.
• Guardrails: ask the model to avoid giving advice in sensitive areas and to flag uncertainty instead of guessing.

The education guide includes ready-to-copy prompts and suggestions for classroom policies on AI use.

Nonprofits

• Grant writing: outline proposals, turn notes into drafts, and revise for clarity while keeping your voice.
• Communication: adapt one message into a newsletter, a social post, and a donor update.
• Data-heavy reporting: attach spreadsheets or long reports and ask questions against the full text.
• Volunteer coordination: draft schedules, role descriptions, and onboarding checklists.

Nonprofits can also apply for program service allocations for API access at cost. See Rate limits.

Research

• Literature review: attach papers and use Corus to compare methods, results, and limitations across documents.
• Experiment planning: use Ichnus as a sounding board for study design and statistics, then verify with primary sources.
• Reproducible notes: keep your prompts and the model versions you used so your process can be recreated.
• Citation care: never trust the model for citation details. Verify every reference against the source.

Prompt tips

• Give the model a role and a goal: “You are a copy editor. Tighten this paragraph without changing meaning.”
• Set constraints up front: length, audience, tone, and what to avoid.
• Ask for structure: “Answer with a decision, the reasons, and a next step.”
• When the answer matters, ask the model to separate facts from its own inference.
• Iterate. The first response is a starting point, not a final product.

Honesty features

Ichnus includes features that make uncertainty visible:

• Ask the model to mark uncertain claims, and it will flag statements it is not confident about.
• Ask “What are you unsure about here?” for a second-pass self-check.
• The model can decline a question with “I do not know” instead of guessing; this behavior is encouraged by the OversightQA eval.

For deeper guidance on any scenario, the Community forum has threads contributed by educators, nonprofit staff, and researchers.

All documentation →