Scholarly September 2026 Updates Recap

September was about making Scholarly feel faster, calmer, and more dependable. Everything you start now remembers where you stopped, your phone got a redesign that works like a real app, research gained a full team-style Extended mode, and chat learned to edit past messages, accept any file, and answer follow-ups without making you wait. Teams also got clearer credit and billing visibility, and a long list of fixes made video lectures, podcasts, recordings, and uploads more dependable. Here's the full rundown.

---

Major Features

Continue Where You Left Off, on Any Device

Video lectures, podcasts, PDFs, pages, research reports, and flashcard decks now remember where you stopped and reopen there on any device you sign in from, with a small note saying where it picked up and a Start over button. Home leads with a Continue where you left off strip of up to four unfinished items, each showing how far in you are and what's left, and Library, folder, and project rows carry a thin progress bar on anything you've partly watched, listened to, or read. Playback speed, volume, captions, PDF zoom, and your last-used tab now follow your account instead of your device, so your laptop and phone agree.

How to use it: Just open something you were partway through — it picks up where you stopped. Use Start over to begin again.

A Phone Experience Built Like an App

Scholarly on a phone is no longer the desktop layout squeezed down. Home, the video page, podcasts, flashcards, PDFs, recordings, chat, research, your library, and Settings were each redesigned for one hand and one column: edge-to-edge players, a single scrolling row of section chips, flat library rows, and menus that open as bottom sheets. Add content became a native-feeling sheet, Customize became one screen of core choices with a single Create button, and Settings is now a hub of grouped lists. Opening plans on a phone shows the price and upgrade button without scrolling, and automatic upgrade popups no longer interrupt your work.

Extended Research and Go Deeper

Turn on Extended in the research composer to put a whole team of AI researchers on a question: some cover different parts in parallel, one checks what changed in the last year, one independently verifies the claims the answer rests on, and a reviewer critiques the draft before it's finished. It uses 3 AI creations, takes roughly 10–20 minutes, is available on paid plans, and asks you to confirm before starting. Go deeper on any finished report runs Extended research on top of it, chasing newer data, primary sources, and anything the first report couldn't verify. You can also just ask for either in chat.

The research page itself was redesigned as a calm, flat brief: Summary leads with the answer, how confident it is, and what's still open; tap an open question to ask it in the research chat; and Sources leads with the sources the report actually cites, with exports and citation copying in one menu. A live progress timeline shows what's happening while a run is under way.

How to use it: Open Deep Research, describe your question, switch on Extended, and confirm. When it finishes, use Go deeper to keep pushing.

Edit Sent Messages in Chat

Hover over a message you sent and choose Edit to change it right in its bubble and get a fresh answer without leaving the conversation. Earlier context and attachments stay, and version controls let you go back to the original conversation at any time. On phones the Edit and Copy icons stay visible.

Chat also got friendlier about files and speed: every file and link appears as the same pill in the composer, your message, and replies; you can drop files anywhere in chat, download from a pill, preview ZIP contents, and ask chat to send a ZIP bundle back. Sending a follow-up while chat is still answering now stops the answer immediately and starts on your new message — no more queueing. Long conversations open on recent messages and load older history as you scroll up, and chats keep track of details far into long conversations.

AI Meeting Notes: One Credit Per Meeting, Notes Included

Recording a meeting now costs one AI Creation Credit regardless of length, covering the transcript and the notes together. The credit is taken when you press Record, so you're never told mid-meeting that you can't keep it. Notes are written as soon as transcription finishes, every plan keeps every meeting, Free transcribes up to 10 minutes per meeting, and paid plans up to 10 hours.

You can now pick your microphone from the recording bar, switch mid-meeting without stopping, and if your browser has blocked the mic, the page says so up front with steps for your browser. If a recording picked up no speech, you're told before processing and can switch microphones and keep recording into the same meeting.

Write Your Own Flashcards, Quizzes, and Tables — Free

Not everything needs AI. Create flashcards has a new Blank deck tab where you write questions and answers side by side, paste a whole list with Import from text, reorder, swap sides, and shuffle. Create quiz has a Write your own tab, and Blank → Table opens a grid where you can type or paste straight from Excel or Google Sheets. All three are free, use no AI credits, and your draft is saved on your device until you create it. New pages from Page now open instantly, without a setup modal, and the Page editor was refreshed with a centered formatting bar and more reliable undo.

Simpler Model Choices and New Models

The model menu offers Auto, Auto (Premium & Laureate), and Auto (Best), and creation settings warn you in yellow whenever a model uses more than one credit. New models arrived through the month: GPT 6 Luna (the new default), GPT 6 Sol, and GPT 6 Astra, later updated to GPT 6.1 Sol; Claude Opus 5.5 and Claude Sonnet 5.5; Grok 4.7; GPT Image 2.5; DeepSeek V4.1 Flash; and GLM 5.3 Flash, Kimi K3, and Muse Spark 1.3. Auto (Best) now runs on Claude Sonnet 5.5 at the same 3 credits.

Video Lectures: Real Figures, Right Length, Better Delivery

Video lectures made from a PDF now pull your document's own figures, charts, tables, and diagrams onto the screen instead of turning everything into text cards. When you pick a runtime, lectures are now built to that length — a 15-minute pick no longer arrives as a 6-minute video. Paid plans start new lectures on Freestyle, where the AI chooses the format that suits your material. You can turn on Use my script as written to have the narrator keep your wording and order. Lectures also show their requested and delivered length, source page range, and narration language, and owners can download an export pack: one zip with the video, slide deck, narration script by chapter, captions (VTT and SRT), and a README of sources.

Podcasts: Full Source Sets, a Better Player, and Faster Starts

Podcasts now accept every source your plan allows — up to ten on paid plans and twenty on Laureate — and build from short or scanned sources instead of failing. The player is a single fixed-height row with a progress bar split into one segment per chapter, a quieter chapter list, and a mini player that follows you while you work, with lock-screen and headset controls. For multi-chapter episodes you can start listening to chapter 1 while the rest is still being recorded, and podcasts with the visual companion use the first slide as their cover.

Scholarly for Teams: Clearer Credits, Billing, and a Delivery-First Start

Team Overview now shows how many AI Creation Credits the team has used this week and when they reset, with a meter for each member and Near limit or At limit tags before anyone is blocked. Owners and Admins can buy member credit packs on the team bill, so a seat at its weekly limit always has a way forward. After buying Teams, the workspace opens with a welcome card that starts a training video from your document, plus a checklist: create a lecture, download the export pack, set your team's voice and look, and invite someone. Paid seats are always treated as paid, setup and invites are smoother, and team project details save instantly.

A Faster, Calmer App

Almost everything now appears the moment you do it: uploads and finished creations show up in projects right away, renames, deletes, and moves apply instantly, new chats open immediately, revisited pages open without a loading screen, and Home paints from your last visit while it refreshes. Projects open instantly, and personal projects save in the background. If a change can't be saved, it's undone and you're told.

---

Everyday Improvements

• Every creation explains a failure. Slides, documents, study guides, infographics, mind maps, spreadsheets, research, and more now say why an attempt didn't finish, whether credits came back, and offer the right fix — Adjust, try another model, pick fewer pages, or replace a source.
• Progress page for every creation. Each generation opens a live progress page with steps, elapsed time, and a typical duration. Close the tab and it keeps running.
• One Recent list on Home, grouped by day, with In progress and Scheduled work at the top.
• Add content without a waiting screen. Files and links appear as pills with their own progress, and failures keep Retry and Remove right on them. Research uploads work the same way.
• Delete and restore inside projects. Every item has Delete, with a Deleted section, Restore, and Undo. Ask chat to "delete everything older than June" and it handles it — always reversibly.
• Sources button on chat answers. Every citation is checked against your linked sources, and each answer lists what it used with the pages cited. Thumbs up and down are back on every answer.
• Preferred document format. Choose Auto, editable Word, or PDF in Chat settings.
• Exams. See a cited source passage alongside each question, challenge any answer key once per attempt, and ask for AI feedback on a whole attempt.
• New menu. Create options are grouped by what you walk away with: Listen & watch, Present, Write, and Research & data.
• Sort and organize. Sort and group projects and All Content by name (with natural number ordering), date, or type, and drag a project into your own custom order.
• Export. Finished work offers Word or PowerPoint formats, and with Google Drive connected you can save as editable Google Docs or Slides.
• Plan-limit emails. Premium and Laureate members are emailed when a reached allowance resets.
• Try before you sign up. The scholarly.so home page opens on a clickable composer where you can preview an output, and your choice carries through sign-up.
• Public library and blog archive. Explore has a browsable library of every publicly shared item, and the blog has a paged archive.

---

How to Use the Highlights

Edit a chat message

1. Hover over a message you sent and choose Edit.
2. Change the text and send it.
3. Use the version controls to move between the original and edited conversations.

Run Extended research

1. Open Deep Research and describe your question.
2. Turn on Extended and confirm the 3-creation cost.
3. When it finishes, review the Summary and use Go deeper for another pass.

Write your own flashcards

1. Open Create flashcards and choose the Blank deck tab.
2. Type cards, or paste a list with Import from text.
3. Reorder, swap sides, or shuffle, then create the deck — no credits used.

Record a meeting

1. Start AI Meeting Notes from Home or the New menu and choose your microphone.
2. Record — one credit covers the transcript and notes.
3. Open the finished meeting for the transcript, summary, and action items.

Download a video lecture export pack

1. Open a video lecture you own.
2. Choose the export pack in the lecture header.
3. Get the video, slides, narration script, captions, and a source list in one zip.

---

Bug Fixes and Reliability

Video Lectures and Podcasts

• Lectures recover from interruptions and finish instead of failing on plan changes, long narration, or math lectures that run past the chosen length.
• Videos that got stuck buffering now recover, and on iPhone and Safari a slow video shows a spinner and starts when tapped.
• Failed videos and podcasts say why — blocked content, an unreadable source, a source that was too large — and offer only the fixes that apply.
• Narration waits out brief voice-service outages instead of failing, and read-aloud audio is clearer and keeps your volume.
• Podcast chapter skips and the chapter list were corrected.

Chat

• Choosing Claude Haiku 4.5 now actually answers with Claude Haiku 4.5, and PDF chat reads page ranges.
• When the image model declines a prompt, chat says so and carries on with the rest of its reply.
• Chat answers stay smooth while they stream, and opening a chat keeps the message box in place.
• Free-allowance limits now say what ran out and how to upgrade, instead of a generic failure.

Uploads, Sources, and Recordings

• Long iPhone and QuickTime videos upload reliably, large uploads have more time on slow connections, and long file names no longer drop a source.
• Blank or empty PDFs are caught right away with a clear message, and RTF files labelled as Word documents now upload.
• YouTube imports explain exactly why a transcript can't be fetched and ride out brief outages.
• Safari recordings keep everything said up to the moment you press Finish, and a retry button keeps your voice dictation if it can't be transcribed.
• Failed generations are refunded, and credit messages on cancelled and failed tasks are clearer.

Billing and Account

• Checkout failures stay visible with your billing period kept for retry, and upgrade recommendations pick the lowest plan that fits your actual source.
• Laureate subscribers can schedule a switch to Premium at renewal, review the price and date, and undo it before it starts.
• Signing out or switching accounts no longer carries the previous account's in-progress items into the new session.
• Stronger account and sharing protection, and a simpler cookie banner with Accept or Deny for optional advertising cookies.

Everyday Reliability

• Pages open filled on direct loads, a project page crash was fixed, and PDFs open right away instead of sometimes taking minutes.
• Link previews work for DOI and YouTube links, and the page editor's Remove suggestion and Undo/Redo shortcuts work correctly.
• Continued light-mode and dark-mode polish across Settings, create screens, and menus.

---

Looking Forward

September made Scholarly more continuous and more dependable: your place is remembered everywhere, your phone has an experience of its own, and creations explain themselves when something goes wrong. Expect more depth in research, Teams, and video lectures through October.

If you have any questions or feedback, reach out at hello@scholarly.so.