What's shipped
Changelog
Newest first. Every entry here is something real, already built and verified — not a preview. Looking for what's next instead? See the roadmap.
Added real numeric methods and inferential statistics — ill-conditioned linear systems, root-finding, ODE integration, plus real t-tests, correlation, regression, ANOVA, and chi-square with real p-values.
Generate a study plan, flashcards, or a practice exam from just a topic now — no document required. Great for test prep (SAT, GRE, MCAT) or starting a subject from zero.
Added Deep Research: a standalone tool that synthesizes a cited report from several real web sources for an open question, separate from full paper writing.
The green "Verified" badge is now real and server-checked — computed independently of the model's own wording, so it only appears on genuinely computed results.
Focus Mode is now enforced by the server, not just requested of the model — a direct-answer dump gets caught and regenerated as real coaching before you ever see it.
Reviewing a flagged weak area now opens exactly the right flashcards or missed exam questions, instead of a generic, unfiltered review panel.
Research papers now get their citations checked against the real fetched text of each source — flagged honestly as grounded, ungrounded, or never actually fetched.
Added real linear algebra (determinants, inverses, eigenvalues, null spaces) and a proof-checking tool that verifies every real algebraic sub-step and critiques logical structure.
Added cross-document synthesis (ask a question that spans several of your readings) and in-document highlighting: select a passage and get an explanation attached right there.
Added Conversation Practice: real turn-based spoken roleplay in your target language, with real multi-language text-to-speech and gentle in-context corrections.
Added real code-correctness checking — runs your own code against real tests in a sandbox and reports exactly what happened. It never writes or fixes your code for you.
Added MLA and Chicago paper styles, alongside the existing IEEE and APA7 — real compiled LaTeX documents, not just formatted citations.
Added real chemistry computation: balancing equations with real linear algebra, stoichiometry with real atomic weights, gas-law and pH problems solved exactly.
Added production-mode flashcards — shown the definition, type the term yourself, graded with a diacritic-aware comparison, instead of just recognizing an answer.
Added AI-generated interactive artifacts: diagrams, charts, slideshows, and quiz games built to your actual material.
Added document and Anki deck export, plus a personal calendar with an auto-updating ICS feed you can subscribe to from any calendar app.
Added voice: real speech-to-text dictation and text-to-speech playback for chat replies.
Added a home dashboard — recent documents, upcoming reviews, and your real progress at a glance, replacing an empty chat as the landing view.
Added "Learn Mode": step-by-step math you have to attempt yourself before Newton reveals the next step, plus comprehension checkpoints during explanations.
Added a baseline safety net that recognizes clear, direct statements of self-harm intent and responds with real crisis resources instead of continuing the tutoring conversation.
Added self-service Focus Mode — a student-controlled toggle for stricter, more Socratic coaching on math and writing.
Added full research paper writing: a plan-then-approve workflow that compiles a real, cited IEEE or APA7 paper to PDF.
Added FSRS-scheduled spaced-repetition flashcards and adaptive practice exams generated from your own documents.
Added the Newton Notepad: take live lecture notes, highlight a confusing passage, and get an explanation grounded in your own notes and documents.
Added real symbolic math — exact solving, differentiating, integrating, simplifying, and factoring, with the derivation shown step by step.
Newton's tutor loop shipped: real tool-calling, a persistent document library, and memory that carries across sessions instead of resetting every chat.