January 2026 was a landmark month for Scholarly. We shipped Research v0.6 with intelligent clarifying questions and a code sandbox, completely rewrote Voice mode from scratch, redesigned the home page, added four new AI models, and overhauled flashcard generation. Here's everything.

---

Major Features

Research v0.6

Research got a complete overhaul with intelligent preparation and a visible plan:

Multi-Choice Clarifying Questions When your query has multiple possible directions, you'll see multiple-choice options to narrow the scope before research begins. Each question includes a recommended option so you can move fast.

Free Response Input Every question includes an optional free-text field to add additional context. You can also skip questions entirely to let the AI decide.

Smarter Query Preparation The system intelligently decides when to ask and when to proceed — specific queries start immediately, broader topics get clarified first.

Research Plan Research now creates a visible plan before diving in. A progress checklist shows each step as it transitions from pending to in-progress to completed.

Collapsible Tool Steps Individual tool steps are tucked into a collapsible "Steps" section, keeping the view clean and focused on progress.

Sandbox Environment Research now has access to a persistent environment where the AI can run code, write files, and execute Python — all in a shared workspace where files persist between steps.

Other Changes

• Banner generation removed for faster, simpler reports.
• Fixed research steps showing wrong info.

Voice v4

Completely rewritten from scratch:

• Talk to your AI tutor using your voice with instant spoken responses.
• Voice and text messages now appear in the same conversation — switch between typing and talking seamlessly.
• The AI can search your PDFs, notes, and the web while talking to give you accurate answers.

Redesigned Home Page

A completely new look with a bento grid layout:

• Quick access to upload content, create AI flashcards, generate podcasts, and start writing — all from one dashboard.
• Folders and recent content easier to find and organize with drag-and-drop.
• Chat now available directly on the home page.
• Shared content appears in the main "Recent Content" list with sharer avatar and badge.
• Content limit increased from 8 to 15 items.

Four New AI Models

A major expansion of available models:

• Grok 4.1 Fast — xAI's latest with excellent reasoning and fast responses.
• GPT 5 Mini — OpenAI's efficient model balancing quality and speed.
• Gemini 3.1 Pro — Google's most capable reasoning model with advanced thinking.
• Gemini 3 Flash — Faster, more efficient model for quick responses.

Chat Empty State

The chat page now shows a bento-style empty state when starting a new conversation, with a dismissible feature update card and your three most recent chats for quick access.

---

Chat & AI Improvements

New Chat Capabilities

• Create flashcard sets directly from chat — the AI generates and saves flashcards as you study.
• AI image generation powered by Grok — create and edit images within conversations.
• Code execution — the AI can run code and install packages to help with coding and data analysis.
• Read external PDF links — paste any PDF URL (like arxiv papers) and ask questions about it.
• Queue messages — type follow-up messages while the AI is responding; they're combined and sent automatically.
• Files created with AI stay saved throughout your conversation.
• Downloadable files now support all types (ZIP, PDF, images, etc.).
• Clean URLs for conversations — easier to share and bookmark.

Podcast Customization

• Choose a style preset: Conversational, Exam Prep, Deep Dive, or Quick Summary.
• Write custom instructions to control how your podcast sounds.
• Redesigned two-step creation flow — pick content first, then customize settings.

---

Flashcard & Study Improvements

Flashcard Generation Overhaul

A complete rebuild for better reliability:

• Refresh the page while generating without losing progress — generation continues in the background.
• Email notifications when long-running generation completes (30+ minutes).
• Failed generations can be retried without re-uploading files.
• Cards are now more exam-focused with better relevance to actual test material.
• Answers more concise (1-3 sentences max) for faster review and better active recall.
• Better handling of vocabulary and language learning content.

Session Timer

Click the timer in flashcard mode to open a new timer modal:

• Switch between stopwatch and countdown timer modes.
• Preset durations (5, 10, 15, 25, 30 min, 1 hr) or start a Pomodoro session.
• Timer persists across normal mode and study mode.

Other Study Updates

• Flashcards sort by difficulty (hardest first) during your first study session.
• Flashcard favorite button now gold for better visibility.
• Flashcard generation banner repositioned to bottom-left so it no longer covers navigation.
• Exam tab shows a proactive banner when no exams exist with one-click creation.
• Creating your first exam drops you straight in — no extra clicks.
• Editable titles — click to rename chats, PDFs, flashcards, podcasts, and videos inline.

---

Platform & UI

• Cleaner chat input design.
• Fixed @ mentions in chat — keyboard navigation and selection now work correctly.
• GPT 5.2 set as default model with increased conversation memory.
• Improved research speed and report quality.
• Research can now take up to 120 minutes for complex topics.
• Magic-link login added — sign in via a secure email link.
• Improved stability and reduced error rates across the platform.
• Better search with graceful fallbacks when services are temporarily unavailable.
• Gemini 3.1 Pro Image set as the default image generation model.

---

Bug Fixes

• Fixed flashcard generation failing for large PDFs.
• Fixed losing progress when browser tab closed during flashcard generation.
• Fixed chat messages disappearing when starting a new conversation.
• Fixed "Think Longer" toggle not being remembered across sessions.
• Fixed chat causing page refresh when starting a new conversation.
• Fixed LaTeX/math equations not rendering properly in quiz, exam, and flashcards.
• Fixed LaTeX rendering with currency symbols (e.g., $2,000).
• Fixed math equations not displaying correctly in research reports.
• Fixed forgot password emails not sending.
• Password-protected PDFs now detected during upload with a clear error message.
• Corrupted PDFs show an error immediately after upload instead of failing later.
• Fixed file size limit not enforced for free users.
• Fixed flashcard creation timing out for large files.
• Fixed study mode issues with large quiz question sets.
• Fixed exam mode scrolling that cut off questions.
• Fixed settings modal not scrollable on smaller screens.
• Fixed being unable to unselect content in Library tab when creating flashcards.
• Fixed downloadable files opening in browser instead of downloading.

---

Looking Forward

January 2026 was about making Scholarly smarter and more capable — Research v0.6 understands what you need before starting, Voice v4 makes conversations natural, and the flashcard generation overhaul ensures reliability at scale. February will build on this foundation with even more powerful features.

If you have any questions or feedback, reach out to us at [email protected].