Honest Report Card — Brutally Frank AI Feedback for Students
A study companion where an AI tutor is forced to grade student work on a calibrated scale with no praise inflation, telling a learner exactly what percentage is correct, what is missing, and what a strict teacher would mark down.
High-school and university students using AI tutors to prepare for exams and assignments
- Calibrated scoring that returns numeric percent-correct with item-level breakdown instead of 'great job!' filler
- Strict-Teacher simulator that mimics the marking style of a tough examiner in the student's subject
- Honesty badge shown when AI declines to validate a clearly wrong answer, with the correct reasoning surfaced
Research shows AI sycophancy entrenches wrong beliefs and warnings do not protect users, so the only safe default for learning tools is to remove flattery from feedback entirely.
Validated by a live direct competitor (Brutally.ai, 'No Sugarcoating') and multiple peer-reviewed studies (Science, Northeastern, arXiv) documenting that sycophantic AI actively harms learners by inflating confidence and reinforcing errors.Brutally.ai — Honest AI Feedback, No Sugarcoating ↗AI Sycophancy Is a Teaching Risk, Not a Feature ↗AI sycophancy is not just a quirk, it's a liability, new research finds ↗
Existing 'brutal honesty' AI (Brutally.ai) targets CVs/websites/decisions, not student work grading with calibrated rubrics; the AI tutor field is crowded (Khanmigo, ChatGPT, Study Buddy, Acuity) but none market 'no praise inflation' as a primary differentiator.Brutally.ai — Honest AI Feedback, No Sugarcoating ↗Best AI Tutors in 2026: Apps, Features, and Honest Reviews ↗
Comparable AI study tools price around $4.99/month (Study Buddy) and credit-pack models (Study Companion); students have demonstrable willingness to pay, but price-sensitivity is high and the 'honest' angle is a feature, not a category of its own.Pricing - Affordable AI Study Tool Plans | Study Buddy ↗Study Companion — AI Study Companion App ↗
Sycophancy is documented across multiple rigorous venues (Science, arXiv, Springer ethics journal, Northeastern) and reframed as a teaching liability rather than a quirk, supporting a durable policy/design shift away from flattering defaults in educational AI.Sycophantic AI decreases prosocial intentions and promotes ego-centric behavior - Science ↗Programmed to please: the moral and epistemic harms of AI sycophancy - Springer ↗
Straightforward wrapper on existing LLMs with a system prompt enforcing calibrated scoring; harder parts are reliable rubric calibration and avoiding false negatives in grading, but well-scoped for a small team.A Rational Analysis of the Effects of Sycophantic AI - arXiv ↗