One question. One line. One why. In seconds.

Quick Solve is the fastest answer-only mode on SpinWheelGo: type or photograph a single problem and AI returns the final answer in one line plus a 2-3 sentence explanation. Worked Examples is the opposite — it generates a full step-by-step walkthrough with strategy, numbered steps, rationale, and a check-yourself prompt. Use Quick Solve when you just need the answer fast (homework check, trivia, lab calc). Use Worked Examples when you need to teach or learn the method. There's a one-tap "Need full steps? → Worked Examples" button on every Quick Solve result that hands the same prompt over.

Yes. Drag-drop or pick any JPEG / PNG / WebP image up to 5 MB. The browser resizes it to 1024 pixels on the longest edge before upload (so a 12 MP phone shot becomes a few hundred KB), strips EXIF metadata for privacy (no GPS, no device serial), and sends it to Google Vertex AI Gemini in multimodal mode. The model handles both clean printed text and messy human handwriting on lined paper, including math notation like the radical sign, fractions, exponents, and chemical formulas. For best results: write large, use blue or black ink, keep the photo well-lit and roughly straight.

Tap the calculator icon in the textarea to open an 8-tab symbol picker with 192 symbols. Tabs: Common (+, −, ×, ÷, ≤, ≥, √), Math (powers, fractions, ∞, ∑, ∫, π, e, set ops), Greek (α through ω, Δ, Σ), Chem/Phys (→, ⇌, H₂O, CO₂, NaCl), Units (SI: m, kg, s, Hz, V, Ω, mol, pH, °C), Stats (σ, μ, x̄, ŷ, χ², ρ, p̂, df, z, t, F, α, β), Logic (∧, ∨, ¬, →, ↔, ∀, ∃, ⊢, ⊨, ∅, ∈), Music (♩, ♪, ♫, ♬, ♭, ♮, ♯, 𝄞, 𝄢, dynamics: p, mp, mf, f, ff, cresc, rit). Tap a symbol to insert at your cursor; tap the calculator icon again to close.

Completely free. No signup, no email, no credit card. The server is stateless: nothing is written to the database — not your prompt, not your image, not the answer. The only history is the last 5 successful solves in your browser's localStorage on this device only, purely so you can re-look-up something you solved a minute ago. Clear it with one tap. Per-IP rate limits protect the AI quota: 15 text solves per minute and 10 image solves per hour — real students never hit either cap.

Quick Solve is tuned for single-question prompts where the user mostly wants the final answer with a short why. It does very well on: arithmetic, fractions, percentages, square roots, single-variable algebra, geometry area / perimeter / volume, trigonometry identities, logarithms; chemistry conversions (moles, molar mass, percent composition), single-step balanced equations, pH; physics (Newton's laws, kinematics, simple Ohm's law); statistics (mean, median, mode, variance, std dev, z-score, single chi-square); logic (truth tables for ≤ 3 vars, set operations, validity of short arguments); and general trivia. For multi-step proofs, full essay structure, or a teaching-grade walkthrough, the Worked Examples generator is the right tool — Quick Solve has a one-tap hand-off to it.

Yes — mobile-first. The modal goes full-bleed on screens smaller than 640 px so the textarea, symbol picker, and image upload all stay reachable with one thumb. Tap targets meet WCAG 2.5.5 (minimum 44 × 44 px). The symbol picker is an inline expanding panel below the textarea (not a popover) so it never gets clipped by the modal's body scroll on small screens. The image upload uses the device's native file picker — on iPhone or Android that lets you pick from photo library OR open the camera to snap a fresh shot of homework.

The AI prompt is engineered for accuracy-first: the model is instructed to be RIGHT first and brief second, with explicit rules about units, balanced equations, and "don't restate the question". The server validates the response shape on every call (answer ≤ 80 chars, explanation ≤ 200 chars, single domain enum) and runs one corrective retry on parse failure before returning a soft error. Output is hard-capped at 150 tokens to stop runaway responses. We still recommend a quick read-through for high-stakes use — AI is a strong assistant, not infallible. If the model can't solve something safely or unambiguously, it returns a friendly explanation with no answer rather than a guess.