Tutor AI
Designed to explain, guide and ask helpful questions rather than simply provide answers. AI can still be wrong, so schoolwork should be checked when accuracy matters.
What the AI in Kindos is designed to do, what it should not replace, and where families should keep human judgment in the loop.
Kindos uses AI for specific experiences while keeping parents, children and other family members in control of important decisions.
Designed to explain, guide and ask helpful questions rather than simply provide answers. AI can still be wrong, so schoolwork should be checked when accuracy matters.
Helps with everyday family organization and supported actions. Users should confirm important dates, addresses, safety information and other consequential details.
Can interpret intentionally submitted images such as homework or chore proof. Image understanding can be imperfect and may require human review.
Can host or support family games with age-aware guardrails. It is entertainment—not a source for high-stakes factual or safety decisions.
Kindos uses leading third-party foundation models for language and vision capabilities. On top of those models, Kindos is rolling out a proprietary orchestration layer built specifically for family use. That Kindos-owned layer manages which AI experience is being used, what authorized family or learning context is appropriate, tutoring behavior, privacy and safety rules, model routing, fallback behavior, and response handling.
Kindos does not claim to have trained the underlying foundation models. The proprietary part is the Kindos-built application layer around them: family-aware orchestration, tutoring logic, permissions, safeguards, memory workflows, vision workflows, and product-specific behavior. This architecture is being introduced in stages and tested before wider rollout.
Do not rely on Kindos AI for emergencies, diagnosis or treatment, legal advice, financial decisions, school discipline decisions, or other situations where a qualified person or authoritative source should make the decision.
If an AI response feels wrong, confusing or inappropriate, stop using that output, verify the information and use available parent controls or support channels. AI is a tool inside Kindos—not the authority in your family.
Kindos treats foundation models as one component inside a controlled product system. The public experience is intentionally simple; routing, privacy boundaries, verification, fallbacks and learning logic stay behind the interface.
Public web sessions start fresh. Homework is resized, cropped and optionally redacted in the browser before processing when possible.
Tutor combines intent, learner state, strategy selection, misconception/prerequisite knowledge and reveal controls instead of relying on one generic prompt.
Supported math is checked with deterministic logic so the model can focus on teaching rather than being the only judge of correctness.
Homework Focus and visual teaching use confidence thresholds and fall back safely when a problem or handwriting cannot be read reliably.
Core AI features are versioned with fallback paths; a single provider or feature failure should not make the entire Tutor unusable.
Kindos separates guided success from independent understanding. Independent checks and teach-back can confirm when the learner is actually ready.
Current public release metadata: Kindos release manifest · platform health.