Read once.
Remember forever.
MemoryLynx is a study partner for the things you read, watch, and listen to. It distills each source down to a few ideas worth keeping — with the original passage cited beside each one — then quietly brings them back, just before you’d forget.
Three steps between
reading and remembering.
Most reading tools collect highlights. The highlights then sit in a folder, and the books quietly fade. MemoryLynx closes the loop — capture, distil, return — so the things you cared enough to read actually stay with you.
Capture it
Drop a link, a PDF, a podcast, a photo of your handwritten notes. Forward a newsletter. Eight ways in — whatever your reading life looks like.
Distil with citations
Claude pulls out the few ideas actually worth remembering — and pins each one to the passage that said it. Page numbers, timestamps, line offsets, all preserved.
Remember, gently
FSRS schedules each idea to come back just before you'd forget. Five formats — comprehension, cloze, multiple choice, Q&A, teach-it-back — chosen to fit each idea.
Bring in whatever you’re reading.
Capture friction is what kills most learning tools. So we made eight ways in — one for nearly every shape your reading life takes.
Web articles
Paste any link — The Atlantic, a Substack, a Wikipedia rabbit-hole. Article body extracted; the full text is kept so we can cite back to it.
YouTube lectures
Drop the URL. We pull the transcript with timestamps; the citation on each insight is a click that jumps you to that very second.
PDFs & ebooks
Research papers, book chapters, course readings, up to 50 MB. Page numbers preserved, so citations point at the page you can flip back to.
Podcasts & lectures
Drop in an MP3 or a recording from class. We transcribe it sentence-by-sentence; citations are clickable and play right back.
Photos & whiteboards
Snap a textbook page, a meeting whiteboard, a page of your own handwritten notes. Vision reads it like a careful student would.
Forwarded newsletters
You get a private inbox. Forward Stratechery, the Browser, your favourite Substack — it lands in your knowledge base by morning.
Notes & thoughts
Paste a meeting transcript, a long voice note, a journal entry, a conversation. The same gentle distillation runs over it.
Readwise & Kindle
Bring your highlight history with you in one import. Hit the ground running with a populated library on day one.
Browser extension
One-click capture from any page, with selection highlighting. Coming next.
Every idea, cited.
Every claim, traceable.
AI summary tools have one shared failure mode: they make things up. We built MemoryLynx so they can’t. Each insight remembers the exact passage that produced it — character offsets in text, timestamps in audio, page numbers in PDFs. If we can’t ground a claim, we mark it for your review rather than pass it off as fact.
The source · Cal Newport, Deep Work, ch. 2
The shift from open-plan office to social media to constant chat has, paradoxically, eroded our capacity to do the very thing knowledge work asks of us.
Attention residue is the cost paid for context-switching: when you turn from one task to another, a portion of your attention remains stuck on the prior task①. The cost compounds across the day.
Sophie Leroy’s 2009 study at Minnesota measured this directly. Subjects rotated between brand-evaluation tasks under two conditions — finished, or interrupted.
In the interrupted condition, performance on the second task dropped 17–24% even when participants reported feeling fully focused②. The brain doesn’t care what you think you’re paying attention to.
What MemoryLynx remembers · 2 of 4 grounded
Five ways to practise.
The right one, automatically.
A definition isn’t an argument. A formula isn’t a framework. Different ideas call for different kinds of practice — and your review session interleaves them, so you’re never just on autopilot.
The reading you forgot you did, made askable.
Three months from now, you’ll have eighty articles, a dozen papers, forty hours of podcasts in here. Asking your library is how you actually find any of it again.
Not flashcards.
Understanding that lasts.
The category looks busy — eighteen flashcard apps, all making the same flashcards. MemoryLynx is built for a different person, asking a different question.
Less than one new
paperback a month.
The free tier is generous enough to actually use. Pro removes the limits. No hidden seats, no enterprise calls.
For trying the loop on a few sources.
- ✓10 captures / month
- ✓5 AI extractions / month
- ✓5 review prompts / day
- ✓30 minutes audio / month
- ✓10 library queries / month
- ✓50 knowledge items total
- ✓One-time Readwise / Kindle import
Or USD $72 a year (≈ $6/mo). Cancel anytime.
- ✓Unlimitedcaptures, extractions & library
- ✓15 review prompts / day · longer sessions
- ✓20 hours audio / month · podcasts welcome
- ✓Unlimited library queries
- ✓Unlimited Teach-it-back evaluations
- ✓Anki, Notion, Obsidian & CSV export
- ✓Shareable insight cards & weekly digest
Questions, mostly answered.
How is this different from Anki, Readwise, or AnkiDecks?+
Anki is a flashcard runtime — you write the cards. Readwise stores your highlights and surfaces them. AnkiDecks generates flashcards from sources. MemoryLynx pulls a small number of high-level insightsper source, grounds each one to the exact passage that produced it, and reviews them with comprehension prompts — not recognition flashcards. Different unit, different practice, different outcome.
Won't the AI just hallucinate insights?+
It’s the failure mode we built against. Every extraction stores the offset of the source span that produced it — character offsets in text, timestamps in audio, page numbers in PDFs. If the system can’t ground a claim, it’s flagged as “synthesised — review carefully” rather than passed off as fact. We track the ungrounded rate as a quality metric and target under 10%.
What's the algorithm? SM-2? FSRS?+
FSRS-5. It’s the modern, ML-trained successor to SM-2 — better long-term interval estimates, more responsive to lapses. We use the same rating mapping (Again / Hard / Good / Easy) across all five review formats.
Can I get my data out?+
Yes. Pro includes export to Anki (.apkg), Notion, Obsidian and CSV. Free accounts can read everything via the API and self-export. We don’t lock anyone in — the goal is for the loop to be good enough that you don’t want to leave.
Is my reading data used to train models?+
No. We send your text to Anthropic’s Claude API for extraction and grading, with the no-training opt-out enabled. Your library stays your library — never shared, never used for fine-tuning.
Mobile app?+
The web app is a PWA — installable to your home screen, with offline review and notifications. Native iOS and Android are post-launch on the roadmap, prioritised by user feedback.
The reading you do this year
doesn’t have to vanish.
Capture three things you read this week. Look for them next month. See if they stuck. The loop only takes minutes a day.