I'm excited to publish public beta newsletter #4 sharing our most recent Readwise Reader updates 🙂

As a reminder, I write this newsletter approximately every four weeks covering features we've just shipped, bugs we've recently fixed, and what we intend to work on next. I also share tips & tricks to help you get the most out of Reader. If you prefer to read these in Reader, you can subscribe to the private RSS feed linked here.

I'm two weeks later than usual in sending this newsletter because Tristan & I went to London in April for business and then to Ireland for our biannual team offsite. We've found it really important to get everyone together in person twice a year to make sure we're all aligned, working well together, and shipping quickly. And of course it's fun! Speaking of team, we're about to begin recruiting for two new roles which I'll mention in the farewell: Customer Success & Technical Writer and Senior Growth Engineer. If you or someone you know might be a good fit, we'd love to hear from you!

Onto the product updates:

The theme of the last month continues to be smashing bugs, fixing random UX glitches, and refactoring infrastructure. There's still so much for us to do here, but we can finally see the light at the end of the tunnel. Particularly as it relates to the long-awaited performance refactor being led by Adam that will dramatically speed up the subjective responsiveness of the mobile apps (particularly on slow devices or accounts with lots of documents).

We've also got some fun feature updates:

• TTS on EPUBs & Emails – You can now use text-to-speech to listen to EPUBs and emails once again. (Right now, this is iOS only while we figure out Android.)
• First-Class Summarization – You can now quickly & easily replace summary metadata with a GPT-3.5 response (and optionally supply your OpenAI token to automatically summarize new documents going forward!).
• Links Tab – You can now view and interact with all the forward links and backlinks contained in a document: the start of Reader as a knowledge graph.
• PDF Upgrades – We've been quietly elevating the PDF experience through a series of smallish enhancements with lots more on the way!

I'll go through each of these and then share a bunch of small improvements, a huge smattering of bug fixes, and the new features we're currently working on. Let's go!

TTS on EPUBs & Emails 🔊📚📧

If you previously tried to listen to an EPUB or email only to be notified that that feature was temporarily disabled, I'm happy to report this barrier is removed. You can now listen to EPUBs and emails using text-to-speech once more!

As mentioned above, this only works on iOS at the moment because there are some platform-specific issues with how Android handles audio streaming that we're still grappling with. It's turned out to be a surprisingly gnarly technical challenge.

As part of this TTS upgrade, we've also enabled a v0.1 of text-to-speech in non-English languages, but we're not formally announcing multiple language support yet because we have to fix how languages are automatically determined and/or manually set by the user before this feature realizes its full potential. Also a friendly reminder that last update we shared you can now seek arbitrary listening starting points when using TTS, which pairs nicely with EPUBs.

Props to Mitch and Artem for executing this refactor.

First-Class Summarization 🤏🥇

After months of Ghostreader in the wild, we can clearly observe that document-level summarization is the go-to use case of GPT applied to reading. In a recent interview led by Patrick Collison of Stripe, Sam Altman agreed.

YouTube of this conversation here.

For this reason, one of the first enhancements we wanted to make as part of a broader Ghostreader v2 initiative was elavating summarization to a first-class feature. Now when you hover a summary in the right sidebar, our little ghost will appear giving you the option to replace the default summary (typically creator-supplied) with a GPT-3.5 summary. So much sleeker than clicking into a context menu to invoke Ghostreader into a document note!

After you tap, Ghostreader will animate while waiting for the GPT response.

This also works on mobile if you open the right sidebar.

Tap the i button or slide right-to-left to open the right sidebar.

During internal testing, we immediately found these summaries to be so useful that we knew others, like us, would want them automatically generated for everything. Unfortunately, the quick napkin math suggested that the projected token costs would put us out of business. We expect this to change over time, but until then, we've added a bring-your-own-API-key option.

You can find this by clicking your initials in the bottom left > Integrations

Once you add your OpenAI API key on the Integrations page, all your new documents going forward will automatically be summarized GPT-3.5.

Props to Hannes for taking over Ghostreader from Simon and shipping this first milestone. We have more Ghostreader v2 improvements to come including an easier way to see these summaries on mobile, updates of the existing prompts to newer GPT models, and better treatment of highlight-level responses.

Links Tab 🔗👀

When we originally started building Reader, we made the unconventional decision to reserve space in the UI for a feature called Links that we hadn't yet built. We could have simply not designed that tab in until later, but we felt it important to remind both ourselves and our beta testers that Reader is, at its core, a knowledge graph. Each Library is a unique, intertwingled web of documents, tags, highlights, and annotations. It's just not so obvious yet.

Fortunately for us, Mitch rolled up to the Toronto office in March looking for a hard problem to work on and decided to build the first iteration of Links, setting the stage for a full-blown Reader-as-knowledge graph future.

Every document now has an activated Links tab in the right sidebar. This tab contains a list of all backlinks (documents in your Library that mention the document you're viewing) and forward links (documents in your Library mentioned by the document you're viewing) for the current document.

This YouTube video only has backlinks.

For example, you can see the Patrick Collison and Sam Altman conversation I screenshotted above was mentioned by three other documents saved to my personal Library: Patrick O'Shaughnessy in one of my Twitter list digests, one of Packy McCormick's newsletters, and one of Jason Crawford's newsletters. I can click any of those cards to jump straight to the mention of this interview.

That's Backlinks. Forward links are more nuanced best explained through an example.

In the document (The Umami Theory of Value) above, you can see that the only forward link mentioned therein that I've also saved to my personal Library is the one titled The Savage Girl. It has a bright blue checkmark and appears underneath the INSIDE YOUR LIBRARY heading.

There are lots of other forward links mentioned in The Umami Theory of Value though. Those that another at least one other Reader user has previously saved are placed in the Suggested links section such as How William Gibson Keeps His Science Fiction Real and The Premium Mediocre Life of Maya Millennial. We assume that this collaborative filtering is sufficient signal to conclude that the forward link is worthy as opposed to trash. For each of these forward links, you can hop to the mention inside the current document or click the + button to quickly save the document to your Library. Finally, there's a section called More links (not shown in the screenshot) where all other forward links mentioned go.

As anyone who's been nerdsniped by a networked tool like Roam Research knows, this v1 of Links only scratches the surface of what Reader could look if conceptualized as a knowledge graph. But we're glad to have taken the first step!

PDF Upgrades 📄💅

Mati has been quietly shipping a series of PDF enhancements to make the PDF reading experience more delightful. This includes subtle features like persisting zoom and rotation settings on a per-document basis across sessions. It also includes more visible features like thumbnail-based navigation by default:

This is only on web right now, but Mati is building the same for mobile/tablet.

In order to make way for the thumbnails, we refactored the top bar of the web app to show page numbers and enable jumping to a specific page. We adapted this top bar to all documents in the web app (not just PDFs) which resulted in a much desired UI-free reading experience.

No user interface is visible once the top bar hides!

Mati is now making a round of updates to mobile after which we'll conclude PDFs v2 with a snapshotting feature (ie a way to "highlight" images and tables in PDFs), better cover images, and maybe – just maybe, if the technology exists – a way to deal with cross-column highlight ordering and janky PDF highlight line-breaks.

Note on inverted PDFs in dark mode: We've seen a handful of inquiries come in about PDFs being inverted in dark mode. Our assumption is that if you want the user interface to be dark, then opening a bright white PDF will be harsh on the eyes. Accordingly, PDFs are automatically inverted when you have your appearance set to dark mode. If you want to un-invert, you can switch to light mode (Cmd/Ctrl + Alt + T). We're considering if we should add an extra setting for PDFs or just a notification to reduce confusion going forward.

Parsing & RSS Upgrades 🧼📃

Last but not least, we went through a few month period there without anyone manning our parsing fix workflow, but under Tadek's tutelage, Krzys is now knocking down parsing glitches left and right.

Here's a non-exhaustive list of fixed domains I was able to pull from our ticket management software: theverge.com, ben-evans.com, smashingmagazine.com, densediscovery.com, mrmoneymustache.com, hyperboleandahalf.blogspot.com, solidot.org, gov.uk, ignant.com, cs.stanford.edu, discord.com, noizz.pl, punctuationmatters.com, themuse.com, eatplant-based.com, brief.pl, thurrott.com, alexmurrell.co.uk, yle.fi, shorthand.com, mironov.com, nysfocus.com, opensearch.org, gerador.eu, developers.redhat.com, soundonsound.com, photopills.com, cc.cz, johnsalvatier.org, blog.readwise.io (😬), logseqtimes.com, book.stevejobsarchive.com, questionablecontent.net, lynalden.com, scottaaronson.blog, blog.alexewerlof.com, powazek.com, patrickcollison.com, css-tricks.com, fujilove.com, xrds.acm.org, bloomberg.com, reflets.info, austinkleon.com.

As part of this, Tadek and Krzys have also set up new testing infrastructure for parsing fixes which detects regressions in previous parsing fixes.

Finally, Tadek has been refactoring our server-side RSS pipeline so that RSS documents parse and show up in your Feed much faster. He's already fixed a major issue that was causing many RSS documents to show "loading" forever.

Coming Up 🔜

• Performance – I've said it before and I'll say it again: app performance (i.e., speed of the user interface) remains our #1 priority. Adam is well underway on a massive refactor that we hope is finally coming to a conclusion. It's just a beast of a project.
• More Ghostreader v2 – As mentioned above, Hannes is doing more with GPT summaries on mobile, upgrading Ghostreader to new GPT models, and fixing the behavior of highlight-level responses.
• More PDFs v2 – Also as mentioned above, Mati has some mobile/tablet PDF enhancements already underway. After that, he plans to add snapshotting and take a stab at highlighting issues idiosyncratic to PDFs (multicolumn highlight ordering and line-breaking).
• Quoteshots – Artem is well underway on building quoteshotting into Reader which will transform a highlight into a sexy image for sharing elsewhere.
• Pagination – We're finally in position to begin building one of our most desired outstanding features: pagination. We can't wait to get this out there for long-form reading.

The items above are major features already underway. After these ship, we'll turn our focus to deeper unification of Readwise 1.0 into Reader and Search v2 among other things.

Minor Improvements 🦐

• Hannes significantly upgraded our in-app feedback feature with more options, user categorization of feature suggestions and bug reports, and GPT-4 labeling to help our team respond more effectively.
• Tristan, Hannes, and Tadek tinkered on the database after an incident on Thursday making it faster than ever.
• Mitch and Artem (with a contribution from Adam) enabled a v0.1 TTS using non-English languages.
• Artem pushed a performance refactor on TTS which significantly reduces battery usage and makes the bottom menu more responsive when hiding it via scrolling.
• Bruno and Tristan added a List API endpoint to fetch your documents from Reader.
• Artem fixed a painful issue with long notes on Android where they were hard to read or scroll.
• Adam added language metadata to documents enabling more accurate word counts and reading time estimates in the future.
• Our first summer intern, Rena, removed empty Views from displaying their headings on Home on web to match mobile UX.

Bug Fixes 🐛

• Fixed issue where the mobile apps were unusually slow, battery-intensive, and crash-prone. (If you haven't already, strongly recommend upgrading to 3.20.)
• Fixed bug where tapping a TTS button on Android would click through to the underlying document sometimes causing unwanted behaviors.
• Fixed bug where Chinese characters were erroneously being parsed into the metadata of English PDFs.
• Fixed bug where using Enter to input kanji characters caused the annotation process to prematurely stop.
• Fixed bug where saved YouTube videos were sometimes grabbing the wrong transcript.
• Fixed bug where using keyboard shortcuts to zoom PDFs would cycle back to 10% after reaching 110% instead of continuing to increment.
• Fixed bug where Reader’s in-app browser would continue to play audio even after the browser was exited.
• Fixed bug where notes would appear in front of the adaptive top bar.
• Fixed bug where the right sidebar would cover the sort context menu.
• Fixed bug where pressing the ... button on a list item would open the document instead of the context menu.
• Fixed bug where tapping the < Feed back button returned to Views instead of Manage feeds.
• Fixed bug where the keyboard shortcut M wasn't working in list view.
• Fixed bug where bold default fonts caused some button text to get truncated.
• Fixed bug where highlights in the Notebook on mobile did not correctly display spaces.
• Fixed bug where note input on mobile was disregarded if you added a tag before "saving" the note.
• Fixed bug where the search query input would sometimes switch focus and incorrectly trigger keyboard shortcuts.
• Fixed bug where custom swipe actions on mobile stopped working properly.
• Fixed bug where tags with capitalized characters inputted via the share sheet weren't properly saving.
• Fixed bug where View headers with long titles were breaking formatting.
• Fixed bug where PDF page numbers weren't displaying correctly.
• Fixed a series of bugs where the Android app was crashing for some users.
• Fixed bug where the TTS audio bar would show incorrectly after navigating into a PDF.
• Fixed a series of visual issues on web including the sidebar overlaying incorrectly, misaligned buttons, and capitalization inconsistencies.
• Fixed big where code pre blocks would format incorrectly, break highlighting, and interfere with scrolling.
• Fixed bug where connecting your Twitter account on mobile would fail.
• Fixed bug where Roam syncing would duplicate some highlights.
• Fixed series of bugs with logging into the Android app.
• Fixed series of bugs with the share sheet crashing on Android.
• Fixed bug with Twitter bookmarks failing to sync with Readwise 1.0.
• Fixed some bugs with syncing highlights from Kobo to Readwise 1.0 (but there are still some ongoing issues on Kobo's end which we are continuing to work through with them).
• Fixed bug where deeplinking into Reader from Readwise 1.0 on Android was not opening the right document.
• Fixed a series of bugs where TTS audio would stop randomly.
• Fixed bug where the loading screen of a newly saved document formatted poorly.
• Fixed bug where highlights in the sidebar were not appearing in the same order as they do in the document.
• Fixed a series of bugs with custom shortcuts including an issue affecting the up and down arrows.
• Fixed bug where sometimes selecting text on Android would highlight extra words incorrectly.
• Fixed bug where clicking the x on a tag didn’t properly delete the tag.
• Fixed bug where the sidebars on mobile would too easily slide open by accident.
• Fixed a series of syncing bugs that were causing apps or browser extensions to be stuck without saving changes.
• Fixed bug where bulk actions could be applied to the wrong documents on mobile if activated from search.

Creator Content 📼

Metamuse

Tristan and I just gave our most in-depth interview to date, chatting with Adam Wiggins of Muse and the Metamuse podcast about read-it-later apps, the history of reading tech, and our philosophy behind Readwise. If you're into productivity software, you might really enjoy this podcast!

Linking Your Thinking

We're always grateful for the opportunity to collaborate with community leaders such as Nick Milo. We joined Nick's conference a few weeks ago to give a live walkthrough of how Reader can be used in a Linking Your Thinking workflow.

Matt D'Avella

The prolific YouTuber Matt D’Avella just made a video about the productivity system that organizes his life. Spoiler alert: Tiago Forte, Readwise, and Reader all make an appearance 🤩

Tiago Forte

Speaking of Tiago, he shares his experiments with using ChatGPT to augment his post-reading summarization workflow. He concludes that highlights from Readwise are a key ingredient to producing a valuable, personalized summary.

Maneetpaul Singh

We still have a long ways to go to get the long-form reading experience where we want it to, but when the king of Kindle devices suggests stop using Kindle, and start using Reader… maybe we're on the right track!

Justin DiRose

Filtered Views are one of the most powerful primitives in Reader, but they're admittedly not intuitive as they need to be.  Justin DiRose is here to show you their power.

Nicole van der Hoeven + Napkin

Nicole shares her love of Napkin's AI-based auto-tagging which automatically imports Readwise and Reader highlights through the integration they've built.

NotionDiva

Speaking of integrations, NotionDiva gives our Libby integration some love in her latest video :)

Farewell 👋

Whew, that was a long one. I'll be sure to not let so much time go between updates next time. In the meantime, if you or someone you know might be a good fit for the Customer Success & Technical Writer or Senior Growth Engineer roles, hit us up! We have a lot of fun 🙂

Blarney Castle in Ireland

Thank you again for your continued support and please reach out any time 🙏
– Dan, Tristan, & the Readwise team