What's new

• Adding an item as “Title by Creator” now finds it. The creator is separated from the title before research starts, so a shop page that does not repeat the author’s name in its own heading is no longer rejected — one item went from a bare file list to a full description, a cover, a source link and the seller’s own words.
• A shop that never filled in its brand no longer becomes an item’s publisher.
• Shop navigation, cart menus and footers no longer end up in an item’s description. Where a page turns out to carry no real description at all, the section is simply left out and the item keeps its contents list — measured across the library, twenty-three items were publishing a shopfront.
• Copyright notices are stripped from descriptions, and unauthorised copy sites are no longer named inside one.
• Descriptions are now read from shops that build their pages in the browser. One large distributor was returning nothing but its own menus and signup box, which is what a few items showed instead of a description — they get the real text now, and several other stores came back fuller too.
• The description block is called “About this item” rather than “What the seller says”. For an out-of-print item, the copy that survives is often not on a seller’s page at all.
• Unauthorised copy sites are no longer credited as an item’s publisher, or named as the source of a description. Their text is still used where it is the only record of an item that exists.
• Cataloguing now draws only on sources that permit it. Two reference sites an early version of Maven consulted have been switched off while we ask their owners for permission properly. Nothing already in your library is altered or deleted: when a re-catalogue would replace a fuller record with a thinner one, Maven keeps what you have and says so. One honest consequence: those were the sites that published page numbers, so newly catalogued items name the routine and its creator but generally will not tell you the page it starts on.
• Books and magazine runs that publish their contents as a list are now read in full, which recovers a great deal. A complete magazine run that had been cataloguing to a single entry now comes back with more than 1,600, each credited to whoever devised it. This affects older and hand-built shop pages most, which tend to be exactly the ones holding an index nothing else has.
• Sharing a link to Maven now shows a picture of the library instead of a bare address. Paste it into Facebook, X, WhatsApp, iMessage, Slack or a forum and you get the app itself, with a line saying what it is.
• Fixed a fault that threw away video links found while researching, so a creator’s own trailer or lecture upload can be read again.
• A rebuild can now clear a seller quote that turned out to be the shop’s own listings rather than a description of your item. Thirteen records were carrying a rail of other people’s products, with prices, under “What the seller says”, and nothing could remove it. Real copy is kept, even when it is only a sentence long.
• Research no longer borrows another product’s contents. If the deep search cannot point at a page that is genuinely this item, what it says is discarded rather than merged — one single trick had picked up fifty-one entries belonging to a different work of the same name.
• Re-cataloguing a lot of items at once can no longer hollow out a good record. When a source refuses us mid-run — which is exactly what a big batch provokes — an item whose contents come back much shorter now keeps what it had, and tells you why.

• Your phone or tablet can now hold a library of its own, rather than only reading one from a computer. Setting up a device now asks where your library lives and you choose: on your computer, which is still the right answer for most people and lets you play your videos over Wi-Fi, or on the device in your hand. Devices you have already set up carry on exactly as they were.
• Saving a backup is a proper feature now instead of a plain download. On a phone it opens the share sheet, so the file goes into Files, iCloud Drive, or wherever else you keep things, and every backup is named with its date so you can tell them apart.
• Wherever Maven says something needs Chrome, Edge or Brave, there is now a “Which devices?” link to a page explaining what works where. The computer holding your video and PDF files needs Chrome, Edge or Brave, and it streams them to everything else you own. Everything else about Maven works on every device.
• When your library is kept inside the browser rather than in a folder, Settings now tells you whether that browser has actually promised to keep it, and the one thing that would fix it if not. It also says when you last saved a backup.
• A backup no longer tries to include your videos and PDFs. It saves your catalogue: every item, its contents, and your own notes, which is the part that exists nowhere but Maven. Your media files are already sitting on your own disk, and on a large library trying to bundle them in could leave the app hanging for a very long time.
• You can now import your library straight from a Vanishing Inc. MLI file. MLI is a new open format for moving a magic library between services, and Maven is the first app to support it — choose “From an MLI file” when adding, then either pick the file or paste it in. It reads on your device, shows you exactly what it found, and adds nothing until you say so.
• Items imported this way arrive with the shop’s own product page attached, so the research pass starts from the page the seller actually published rather than trying to find it. That link is kept, and a later re-build will not wander off it.
• Importing the same MLI file twice updates what you already have instead of creating duplicates, even if you have since renamed an item — it matches on the shop’s own record, not on the title.
• Items marked as borrowed are left out of an MLI import by default, since an export file is only a snapshot and not proof that something is yours. There is a tick-box if you do want them.
• MLI files carry titles, creators and store links but no chapter lists, so everything in one needs researching. Maven now says how many items that will use before you add anything.
• You can also export your whole library as an MLI file, from Settings → Library. It is free, uses none of your items, and it is what makes your catalogue portable: any other service that reads the format can take it. Items that came in from a shop go back out exactly as they arrived.
• The MLI export is deliberately a record of what you own, not a backup of what Maven built for you — it carries titles, creators, types and store links, and not your contents lists or your own notes. The Back up button next to it is still the one that keeps all of that.
• Fixed Ask answering “Nothing matching on your shelf” for every question when the AI was actually unavailable. It now tells you what went wrong instead of reporting an empty shelf, which was worse than an error: it looked like a confident answer about your library.
• Big batches no longer lose a chunk of their items. Adding several hundred at once could push the research service past what it will serve at once, and every item in flight when that happened failed — a third of a large batch, in the worst case. Maven now notices, drops to one item at a time for a breather, and works back up as things recover, so the batch finishes instead of leaving you a pile to retry. An item that trips over this is still never charged for.
• An item that fails on a wobble now gets several attempts, spaced out, rather than one straight away. The first retry used to land two seconds later, in the middle of the same problem, and fail the same way.
• When your included cataloguing runs out part-way through a batch, the item it stopped on is now listed as not yet catalogued rather than as failed. Nothing was spent on it and nothing was lost — top up and hit Retry.
• When cataloguing fails on an item, the automatic error report now names that item, so the reason can be found and fixed instead of showing up as an anonymous count. It is only ever an item that failed, never the rest of your library, and /privacy says so.
• Better wording when a device cannot find your library. It now says the two things that are actually wrong most often: the two devices are not on the same Wi-Fi, or the device you are holding has no connection at all. It used to show a raw browser message.
• Ask Maven is now simply Maven: The Modern Grimoire. Nothing about the app has changed and your licence, your library and your devices carry on exactly as they were. If you have it installed, the name in your dock or Start menu updates the next time you reinstall it.
• Search results now show every relevant routine inside a book or video set, not a handful of them. If a search matches twenty-four things inside one work, the strong ones are all listed and the rest are behind a “Show all” link on the same card, so a good match can no longer be missed for sitting late in the index.
• How many are listed adapts to the search. Maven compares each match against the best match in that same work and lists everything close to it, so a work with three real matches shows three or four rather than padding the list out with things that only share a word.
• “Showing a sample, more inside” has stopped appearing on searches that were not sampling anything. It was based on how long a book’s index is rather than on the search, so a long book carried the label whether or not anything had been left out. It now appears only when there really were more matches across your library than one search could rank.
• Every listed routine is clickable, not just the ones Maven wrote a line about.
• Re-cataloguing an item now shows what it is doing. The step it is on appears in the item window as it works — looking up the contents, reading a page, writing it up — instead of a spinner that looked the same whether it was busy or stuck.
• An item that cannot be found online no longer gets invented contents. If all Maven has to go on is the files on your disk, it now lists those files and nothing else — where before it could turn a filename into a few plausible-sounding chapters that no source ever mentioned.
• Items by creators who publish under initials are catalogued properly. A two-letter name could never be matched against a page, so those items were left thin even when their page had been found and read.
• A performer’s Penguin LIVE lecture and their LIVE ACT are no longer mistaken for each other. They are different products, and either one could previously take the other’s cover, description and contents.
• Some covers that showed as missing were there all along. They had been saved as an archive link, which the archive rate-limits, so the picture never loaded; re-cataloguing such an item now points it at the original image.
• A wrong page can no longer be cited as a source. Two ways it could happen were fixed: a shop we trust answering with a different product, and a page that failed to load being read as though its address were its title.
• A record no longer contradicts itself. An item whose research came back empty says so — but if it already had the seller’s own description on file, that line is simply dropped, rather than claiming nothing could be found directly above four paragraphs of it.
• And an older record cleans itself up the next time you re-catalogue it, even when the shop page it came from has since disappeared.
• Shop clutter no longer leaks into your records. A quoted product description used to be able to start with the shop’s own headline, its price, its category and tag lines, and to run on into a rail of other people’s products, which could even be listed as the item’s contents. The quote now begins at the seller’s first real sentence and stops where they stop talking about the item.
• Self-published and small-dealer items are found far more reliably. A creator selling from their own storefront, or a dealer whose shop runs on a blog, was often invisible to research even when the page was easy to find by hand; those pages are now looked for directly, and where a dealer’s listing is only a teaser, the link to the product’s own page is followed for the full description.
• Items distributed by Murphy’s Magic can be catalogued. Their own site has no public catalogue, so nothing could read it; research now finds those titles through a shop that stocks them.
• Cataloguing costs about half what it did, with slightly better results. The expensive web search is no longer spent on items where a free search has already looked, which on a nineteen-item test cut the spend from 44 to 21 cents while finding MORE contents, not less.
• Review copies are simpler. Everyone given a free copy now gets the same short setup page: claim it, set it up, done. An older sign-up flow was still asking some people to create a Gumroad account for an affiliate programme that no longer exists.
• Settings now has an Items panel where the AI panel used to be. Same place, clearer name: it is where you see how many cataloguing items you have left, top up, and check what they were spent on.
• Two new buttons in the header, so the things people were hunting for in Settings are one click away: one for sharing your library with your other devices, and one for sharing Maven itself. Each opens the panel scrolled to the right card.
• Share Maven with another magician and you both get 100 free cataloguing items. Your link never expires and you can share it as often as you like.
• The version number moved under the Maven name, and your email moved to the foot of the page, which leaves the top corner for buttons you actually press.
• Clicking a routine inside a book or video set now opens that item scrolled to the routine, with it briefly highlighted, instead of dropping you at the top of a three-hundred-entry contents list. Clicking the card itself still opens at the top.
• Weak matches are now dropped rather than hidden. A search for two common words could match four hundred rows inside one book, and “Show all” opened every one of them; anything scoring under 15% of that book’s best match is no longer offered at all, so the rows behind the link are ones worth reading. Use “Search within” for an exhaustive pass.
• A library search that finds nothing now offers an Ask AI button. The library box matches titles, authors and tags, so a typed question finds nothing even when the answer is in three books on your shelf; the button hands the same words to Ask, which searches inside your items.
• A large add loses fewer items, and one that slips through no longer costs you anything. Where the research had actually finished but the answer could not be handed back, it now gets through instead of being thrown away, so you are not charged for work you never received.
• A big batch no longer grinds to a halt around a handful of stubborn items. An item waiting to try again was holding up everything behind it, so a few slow rows could park a run of several hundred for minutes at a time.
• Video from another device stops giving up on a slow connection. It allowed twenty seconds to start and then switched to the slower way of playing, which needs far more of the same connection that was already struggling. It now waits for as long as video is still arriving, and only counts a start as failed when nothing is coming at all.
• And a video that started slowly once is no longer stuck on the slow path for the rest of the session. One bad minute of Wi-Fi was enough to send every later attempt at that file the long way round.

• Adding an item by typing just its title works again. If you did not also fill in an author, cataloguing stopped straight away with a technical error message instead of researching anything. Pasted lists, spreadsheet imports, folder scans and re-catalogues were never affected, because each of those already knows who made the item.
• A creator who sells their own work from their own site now gets linked to it. A web address built out of the performer’s name, like joediamondlive.com or osterlindmysteries.com, is that performer’s own shop, and for a self-published lecture it is very often the only page that exists.

• Books that come with a published index keep all of their chapters. A chapter whose title began “Chapter One — …” was being read as a section heading rather than an entry, so one book came back with three of its fifteen chapters.
• An item is no longer marked “not yet researched” just because its contents came from a shop page rather than an archive. Over a full re-catalogue that mislabelled 44 finished records, including one with 48 entries.
• Items whose title is just a number — “22”, “3510” — can be catalogued again. They were being skipped as though the title were a stray page number, which is only true when nobody knows who made it.
• A batch re-catalogue now reports what it actually changed. It was comparing against the wrong thing and telling you almost every item had got worse when the records were correct.
• A review of a book is no longer mistaken for its table of contents. The sections a reviewer writes under are not the book’s chapters, and reading them as such was hiding the real routine list further down the same page.
• If you put the disc count in your own title — “13 Steps to Mentalism 6 dvds” — that no longer stops the shops whose listings do not mention it from being found.
• You can name a lecture the way you actually talk about it. “Mark Strivings Penguin Live” now finds its page on Penguin, whose own titles never mention Penguin — while still telling it apart from the same performer’s At The Table lecture somewhere else.
• Boxed sets are recognised however you have named them. “13 Steps to Mentalism DVD Set” now finds all three shops that sell it, instead of none: a shop writing “(6 DVDs)” was being read as a different volume, and the word “set” in your own title was being treated as part of the work’s name.
• A performer’s recorded LIVE ACT is no longer mistaken for their Penguin Live lecture. They are different products, and the two words are too common to tell them apart, so a page that calls it the other thing is now treated as a contradiction.
• Items whose title contains a number in words — “Six Impossible Things”, “Four Small Lies”, “Thirteen Steps to Mentalism” — now find their own product pages. We were looking for the digit and refusing the page that spelled it out, so 26 items in a library this size were coming back as though they did not exist online.
• Folding a scan into an item you already own now uses the same, much better cataloguing that adding a new item does. It was still on an older path that had quietly stopped being the best one.
• Covers now come through from Penguin and Vanishing Inc product pages. Both sites make us read them the slow way, and an internal limit was quietly refusing the last step — the one that fetches the picture.
• Cataloguing got faster, cheaper and better all at once. An item now takes about a quarter less time, costs well under half as much of your OpenAI credit, and comes back with MORE of its contents filled in.
• Self-published books now get their real table of contents. Where an author lists every piece on their own site with a line about each, all of it is taken exactly as they wrote it. One book went from 15 entries in our words to 72 in the author’s.
• A book whose contents sit behind a link on the author’s own site now has that link followed even when the site blocks us and has to be read through a proxy. Before, those books came back with a blurb and no routine list.
• Lectures keep their contents. When a store page’s customer reviews are the only place a lecture’s running order is written down, those reviews are now read in full rather than cut short by the length of the sales pitch above them.
• Cataloguing is quicker and cheaper where it can be: a DVD whose shop page already lists its chapters no longer goes looking for that list somewhere else, and the deeper search is skipped once the contents are found.
• Re-cataloguing now simply saves what it finds, like adding an item does. The preview and approve-each-change screens are gone; a rebuild that would replace a long contents list with a short one is still refused automatically, and each row tells you what changed.
• A book found in a published index now also gets a shop page looked up, so it has a real description and not just a contents list.
• Martin’s Magic Collection is back as a source, and it is the best one there is for what is on a DVD set. It had quietly stopped working: the site now blocks its own search, so nothing could be found there even though the pages themselves were still readable. A four-volume set’s page is now also cut down to the volume you actually own. It is consulted for books now as well, when the archives do not have one.
• The seller’s description no longer repeats the contents list back at you. Where a shop prints the routine names inside its blurb, that line is dropped, because those names are already listed below with their descriptions.
• Fixed a credit that was being attached to items it had nothing to do with. Six records said a routine used Marc Paul’s AAA Book Test; only one of them does.
• A contents list now reads in the order the source published it, with anything we could not place shown at the end rather than at the top.
• Two spellings of the same routine no longer appear as two entries when the only difference is a word like “in” or “to”, and the spelling already in your library is the one kept.
• “What the seller says” is left empty rather than filled with a shop’s shipping and payment terms. Some mirror pages carry nothing else, and a postage policy is not a description of a trick.
• When a shop lists the contents at the end of its blurb, that list is no longer quoted twice — once as the description and again as the contents.
• Seller descriptions no longer pick up advertising banners or “subscribe and save” lines from the shop’s own page.
• Contents lists no longer pick up stray lines from forum posts. A real contents row is a title; a lowercase phrase came from someone’s comment, and is now left out.
• Large collections read properly again. A page listing nine books and two hundred routines was coming back empty, and now it comes back with every routine filed under the book it belongs to.
• A collection that only lists the books inside it now gets each of those looked up, so you see the routines rather than a list of titles — wherever an archive has them.
• Books with a published index now keep their page numbers, and their front matter (prologues, forewords) is no longer dropped.
• One of our index sources was only being searched for authors A to I. Books by everyone from J to Z now find their contents there too.
• A re-catalogue can no longer replace a long contents list with a short one when neither version can point to a source. It keeps what you had and tells you.
• When you preview a re-catalogue of your library, anything that would lose something is shown first, so the changes worth reading are never buried behind hundreds of harmless ones.
• Re-cataloguing a lot of items is now several times faster. Items that need a reference archive are paced politely; everything else runs in parallel, so a whole library takes about an hour instead of most of a night.
• Batch runs now show a progress bar and an estimated finish time, can be stopped cleanly, and automatically try again at the end for anything a source refused.
• Fixed a fault that could empty an item’s contents during a re-catalogue. If the pages we managed to read cannot account for ANY of what is stored, that now means we read the wrong page, and your contents are left alone.
• Re-cataloguing several items from the same shop no longer overwhelms it, which was quietly making those items come back thin.
• The review screen now waits until the whole batch has finished instead of interrupting after the first item.
• Cover images now come through from Penguin Magic, Vanishing Inc and other shops that block automated readers. They were being dropped entirely.
• Adding an item now searches sixteen magic shops directly rather than relying on your existing library to know where to look, so your first additions find their product page as reliably as your hundredth.
• A title written with a number word now matches a shop that writes the digit, so “Four Small Lies” finds “4 Small Lies” and “Volume Two” finds “Vol 2”.
• Fixed a long-standing fault that cut short the contents we read from one of our reference archives. Books whose listings include a line about each routine were keeping only the first few. Pure Effect went from 4 entries to 17, and Corinda’s 13 Steps from 206 to 444.
• Effect descriptions in big books now come from the published index itself, written by someone who actually read the work, instead of being composed for you. That already covers 85% of The Jinx and 95% of Syzygy.
• Books with a subtitle now find their published index. A one-character bug meant “Prism: The Color Series of Mentalism” never looked up “Prism”, so it saved 15 lines instead of 63.
• Volume 1 of a series is matched to its index again — archives usually title a first volume without the “1” — while volume 2 is still kept off volume 1’s pages.
• Item types are decided by what a release TEACHES rather than the format it came in, so a single trick taught in an ebook is filed as a trick.
• Added a Lecture notes type, and a single magazine issue is now filed as a book while the whole run stays a collection.
• If you set an item’s type yourself, a rebuild will no longer change it back.
• An item mislabelled as a single trick is no longer stuck with a thin listing: when the shop states a routine count, or an archive has an index, Maven digs anyway.
• Video trailers are read in full now. YouTube only shows a short preview of a description to other sites, and the rest often lists exactly what a release teaches.
• Fixed a shop’s OTHER product leaking into an item. Two releases by the same performer with nearly the same name were being merged, so a lecture listed routines from a different DVD.
• The shop’s own description now sits above the contents and keeps its shape — its bullet lists and paragraphs render properly instead of running together as one block.
• Multi-volume books keep their volume headings, so you can see which volume a routine is in.
• An item sold on the creator’s own store is found even when a bigger shop sells something with the same name.
• Big magazine runs and multi-volume sets find their full published index again. An item saved as “The Jinx (issues 1-151, complete run)” was being refused its own 1,028-entry index because no archive describes a book that way.
• Mindhaus items now get their real cover art. We had been reading a homepage banner field that most of their products do not have, so most Mindhaus items had no cover at all.
• Product descriptions are now quoted from the shop that actually sells the item, and say where they came from.
• Cover art can now come from anywhere it survives, including sites Maven will never link you to — a real cover beats a missing one, and the link is still refused.
• A short title like “The Room” no longer picks up a stranger’s cover: the creator’s name has to appear too, or no cover is saved.
• Maven now opens the pages its research finds instead of taking the search’s word for them, so contents come from the shop or the reviewer rather than a summary of them.
• Descriptions that no page actually supports are removed rather than saved. An item may now show a routine by name alone, which means nobody could find out more — not that we made something up.
• Fixed a one-word title picking up an unrelated page: “Mirroring” had been catalogued with a list of Mac features.
• Multi-volume sets now list each disc separately with its own contents, instead of one merged list or the same disc twice.
• A lecture saved without a number now resolves to the first one rather than the sequel.
• Books and lectures now get their sources read right through rather than skimmed. A book whose listing named 11 of its 28 routines now catalogues all of them, under the book’s own chapter headings.
• Fixed a lecture being catalogued with another lecturer’s routines, which happened when a shop page listed both.
• An item saved under a shelf name like “Penguin Live Lecture” now resolves to the real title, so it stops matching every other lecture in the shop.
• Product listings are now quoted into your items word for word, under "What the seller says" — no paraphrasing, and every routine name the shop wrote is searchable.
• Books, lectures and collections always get a second research pass now, even when the first answer looked complete — a shop page that names ten of a thirty-chapter book used to be taken as the whole list.
• When nothing published says what a lecture teaches, Maven now reads the reviews, which is usually where somebody wrote the running order down.
• New: see exactly what a rebuild would change before it is saved. Rebuilding an item — one from its own card, or a whole selection at once — now shows you the difference first: which entries would be added, which would be lost, what the summary would become, and the old and new covers side by side. Approve it or keep what you have, one item at a time, with an approve-all when you have seen enough. Nothing is written to your library until you say so.
• New: Tidy up your library, in Settings under Library. It repairs things in records you already have that never needed research in the first place: contents lists that were saved as plain text (and were therefore invisible to search), stray reference markup left in a description, links that were not really links, and cover images from sites we will not show. It costs nothing and uses no AI items. You can see exactly what it would change before anything is written.
• Improved: cataloguing is markedly more accurate about WHICH item it found. It used to be possible for a product that merely shared a few words with yours — often something else by the same performer — to be taken as the item, which is how a listing could end up with the wrong cover art and a description of the wrong routine. Maven now judges a page by what the page itself says it is, pays attention to numbers so Volume 1 and Volume 2 are never mixed up, and knows the difference between a product page, a forum thread and a shop’s search results.
• Improved: a cover now only ever comes from a page that genuinely is your item. If nothing can be confirmed, you get no cover rather than somebody else’s artwork.
• Fixed: books with a published index keep every entry and every page number. Long indexes were quietly being cut short — a magazine run with over a thousand indexed effects was saving barely a third of them — and page numbers were being lost along the way. The contents list is now built directly from the published index, so nothing can go missing, and the AI’s job is limited to describing each entry.
• Improved: much more of what a shop actually says about an item now makes it into the listing. On big store pages the description was being crowded out by the shop’s own menus and hidden page code, so items came back with a vague summary or none at all. Maven now reads the part of the page that is actually about the product.
• New: for a multi-part course, each lesson’s own page is read too. Course pages often carry nothing but links to the lessons, with all the real detail one click away.
• New: customer reviews are read properly on Penguin Magic, including the ones past the first page. For a recorded lecture the shop’s own description is frequently just a biography of the lecturer, and the reviews are the only place the routines are ever named.
• Improved: Maven now asks a shop directly for its own catalogue when it knows which shop to ask, which finds items that no search engine has indexed at all — including courses and self-published releases on a creator’s own site.
• Fixed: an item can no longer be catalogued from a page that does not exist. A link that leads nowhere was being treated as a readable page, and whatever the shop’s "not found" screen happened to say could end up informing the listing.
• Fixed: re-cataloguing the same item twice now gives the same result. Several things were left to chance that should not have been, so running research again could return something different, and occasionally worse, for no reason.
• New: Preview before you re-catalogue. Rebuilding several items at once can now research everything and save nothing, showing you exactly what would change for each one — more entries, fewer, a cover gained or lost — so you can decide before anything is overwritten.
• Fixed: contents lists that arrived as a run-on paragraph are now proper lists, so the individual routines inside them can be found by search.
• Fixed: some videos played their sound over a black screen, and nothing anywhere said why. A handful of older releases are .mp4 files on the outside but were encoded years ago with something no web browser can show — the file is perfectly good, it opens in VLC, and your browser simply has no way to draw the picture. Maven now notices the moment it happens and tells you plainly, instead of leaving you staring at a black rectangle wondering whether the download was broken.
• New: Fix unplayable video now looks inside your files instead of just at their names, and lists exactly what it finds. It used to judge a video by its extension, which cannot reveal what is actually in it — so a shelf full of black-screen videos looked perfectly healthy and you were told there was nothing to convert. It now opens each one, reads a few bytes to see what is really inside, and groups the results by what is actually wrong: which ones play sound over a black screen, which ones cannot be opened at all, and which ones are not the format their name claims. It names the culprit too, so you can see at a glance that a file is, say, an old DivX encode. On a library of nearly two thousand videos the whole check takes about a second.
• New: it also catches files that are not what they say they are. A surprising number of magic downloads are named .mp4 but are really something else entirely — usually RealMedia — because somebody renamed them years ago. Those never play in a browser and there was previously no way to find out why. They are now found and repaired along with everything else.
• When a repair is needed, the file keeps its name and stays exactly where it is, so none of your items need re-linking afterwards. Anything already fine is left completely alone, and running the tool a second time costs nothing but a quick re-check.
• The repair script now starts almost immediately instead of spending several minutes re-examining your whole collection first. Maven already worked out which files need attention when it scanned, so it writes that list into the script and the script goes straight to them. It still double-checks each one the instant before it touches it, so nothing is taken on trust, and there is a setting at the top if you would rather it re-examined everything.
• You no longer have to install anything yourself. The script fetches the converter it needs the first time it runs, using the package manager already built into Windows, or Homebrew on a Mac. On Windows nothing asks for an administrator password, because it installs into your own user folder. So the whole job is now: save the script, double-click it.
• And if nothing is wrong, Maven simply tells you so. It used to walk you through saving and running a repair script even when every video in your library already played perfectly, and the script would then spend minutes re-checking thousands of files to reach the same conclusion.
• The repair is also much faster than it looks. Most of these videos have perfectly good sound already, so only the picture is re-made and the audio is copied across untouched. In the rarer case where it is the other way round — the picture is fine and the sound is in a format browsers cannot play — only the sound is redone, which takes seconds rather than hours.
• New: bring your library in from a spreadsheet. If you have ever catalogued your collection somewhere else — Zotero, Goodreads, LibraryThing, Calibre, or a spreadsheet you keep yourself — export it as a CSV and Add media will read it. It does not matter which program wrote the file or what the columns are called: Maven works out which column is the title, which is the author, and so on, then shows you what it found so you can correct anything before a single item is added.
• New: if your spreadsheet lists the chapters, tracks or routines inside each item, those items are added straight from your own file — instantly, and without using any of your AI items. Your list is almost always better than anything that could be found online, so there is no sense paying to look it up again. Anything in the file with no contents list is researched the normal way, and the Add button tells you how many of each before you commit.
• New: your own descriptions come with you. If your spreadsheet has a description, blurb or review for an item, Maven keeps it word for word under its own heading in the item, and research adds to that rather than replacing it. This matters most for obscure items: when nothing about something can be found online, Maven refuses to make anything up — so your description is often the only thing there is, and previously it was being thrown away. It is also searchable straight away, which is worth more than it sounds, because store blurbs frequently name the routines and the sleights inside.
• New: a "Imported, not yet researched" filter in the library, so you can come back to everything that arrived from a spreadsheet and fill it in later. Re-building one of those only ever ADDS to it — every chapter you brought with you is kept exactly as it was.
• Fixed: the demo video on the website would not play at all on an iPhone or iPad, and tapping it simply did nothing. Apple insists a video be fetched in pieces rather than all at once, and ours was only being offered whole, which every iPhone quietly refuses. It now plays everywhere, starts sooner, and is less than half the size it was, so it costs a lot less to watch on a phone.
• Fixed: a book could refuse to open at all when read from another computer, while opening perfectly on the computer that holds it. Fetching a book a few pages at a time asks more of the file than reading the whole thing, and some scans — usually older ones, or ones a converter has been unkind to — cannot be read that way. Maven now notices and quietly fetches the whole book instead, so it opens either way rather than showing you an error.
• Fixed: sound could drift out of step with the picture on a long video played from another computer, and scrubbing back and forth put it right. Each time it fetched the next stretch of video it was starting again from the nearest usable point slightly before where it had got to, so a second or two arrived twice; your browser patches over the repeat, and each patch nudged the sound by a hair until you could hear it. It now simply carries on from where it stopped, which sends less rather than more, so it is also a little quicker.
• Fixed: opening a video from another computer could sit on a spinner that never moved, with nothing to do but close it and open it again. Nothing was downloading at that point — the other computer was reading through the file looking for where to start, and on an unusually put-together file it could keep looking for minutes. It now gives up on that quickly and plays the file the slower, reliable way instead of leaving you waiting.

• A video playing from another computer no longer drops back to the slow way of fetching it because one request was slow. If your phone stepped out of range for a moment, or the other computer was busy, a film that was playing perfectly well with plenty buffered ahead was thrown away and started again the slow way — and stayed slow for the rest of that file. It now simply asks again, and only gives up if it genuinely runs out of video to play. It also stops waiting a full minute before deciding something has gone wrong, and the computer holding the file no longer spends minutes hunting through it for a place to start on the rare video that is put together unusually.

• Fixed: Maven could reload the page under you a moment after opening it, for no reason, when you were already on the newest version. Opening a link from Facebook was the reliable way to see it, because Facebook views pages through its own servers and was answering "which version is current?" with a copy of the answer from whenever the link was first posted, sometimes days out of date. Maven believed it and refreshed to catch up with a version it was already ahead of. It can now tell which of two versions is genuinely the newer one, and asks the question in a way that cannot be answered from somebody else’s saved copy.
• When a new version of Maven goes out, you now get a small "A new version is ready" card in the corner, and one tap loads it. Maven has always updated itself quietly in the background, but only while you had not touched anything yet, so that it never reloaded out from under you mid-sentence. Once you had, it waited for you to move to another screen, and on a phone where you open the app, tap once and put it down, that could be days. Dismissing the card gives nothing up: the new version still arrives the next time you move around the app.
• Updates now reach a device even if it was interrupted while saving its offline copy. Maven saves a couple of hundred files so it can open without a connection, and a single one failing, which a phone on a patchy signal manages easily, used to abandon the lot and quietly leave that device on the old version indefinitely. It now keeps what it managed to save and fills in the rest as you use the app.
• Fixed: Maven could open as a bare page with no styling and nothing working, on one device only, and reloading never fixed it. The cause was a file the browser had recorded as missing during an earlier update and then filed away for a year, refusing to ask for it again even after it came back. Maven now checks properly instead of taking the browser’s word for it, so a device in this state repairs itself the next time it opens. It also no longer shows you an older copy of a page when a device is short of storage.
• Where you got to now follows you between devices. Start a lecture at your computer, pick up your phone, and it is waiting in Continue learning at the point you stopped, with the same books ticked as read and the same things in Up Next. It travelled one way before: your phone told your computer, and your computer never told your phone. Your devices exchange this when they connect and whenever you come back to Maven, so it is up to date when you pick a device up rather than changing under you while you watch.
• Fixed: on a phone or tablet reading a library from your computer, Continue learning never appeared at all, and neither did the reading progress on each cover, the tick on books you have finished, Up Next, or sorting by what you opened last. Everything you did there was being recorded properly, it simply was never shown back to you.
• Opening a video or a book from another computer in the first moments after Maven starts is no longer far slower than opening the same thing a minute later. There is a short window while the app is getting itself ready during which it could not fetch a file a piece at a time, and rather than waiting the fraction of a second for that to finish, it fell back to fetching the entire file before showing you anything — which on a large scan or a long lecture is minutes instead of seconds. It now waits for the moment it needs. Most noticeable on a phone, where you tend to open the app and go straight to something.
• Reading a large PDF from another computer no longer quietly downloads the whole thing behind your back. It always opened on page one quickly, but it then carried on fetching the rest of the book whether or not you ever turned to it, which on a big scan is hundreds of megabytes. Since everything your devices say to each other goes down one line, that also slowed down whatever else you asked for while it was happening. It now fetches the pages you actually read. Books already saved to a device, and books in your own folder, are unchanged.
• A long video from your library now starts almost straight away on another device, instead of leaving you watching a black rectangle. The reason it was slow had nothing to do with your network: an MP4 keeps its contents page at the very end of the file, and nothing can show a single frame until the whole of it has arrived. On a three-hour lecture that is over eight megabytes, sent before anything happens, every time. The computer holding the file reads that page off its own disk in well under a second, so it now does the reading and sends your phone or tablet a few seconds of video at a time. Picking up where you left off is quick for the same reason: it starts at that point rather than at the beginning. Anything that cannot be handled this way, an unusual format or an older device, still plays exactly as it did before.
• If you keep a library on two computers, you can now read one from the other without giving anything up. Settings, Library has a new "Which library you’re reading" list: this computer, plus any other you have read before, and anything on your network that is sharing right now. Pick one and your shelf, your search and your notes come from that computer instead; press "Read this one" on This computer to come back. Nothing is disconnected while you look: your own library stays exactly where it is and your phone keeps reading it the whole time. Adding and cataloguing are off while you are reading somebody else’s, and notes and where you got to are saved over there, where that library lives. If that computer turns out to be asleep, Maven says so and shows your own library instead of leaving you with an empty shelf.
• While you are reading another computer’s library, Maven now behaves the way it does on a phone: Add media is gone, items are read-only, re-cataloguing is unavailable, and a video or PDF is fetched from the computer that holds it. Before, the Add button was still there and would have catalogued into this computer’s own library while somebody else’s shelf was on screen. Notices about new files in your own folder wait until you switch back, since they are about the library you are not looking at.
• On a phone or tablet, switching to a different library now works. Settings, Library lists the libraries that device knows and the ones sharing on your network, and picking one moves to it. The old button said it was connecting to a different device and then only measured how long a message took to get there and back.
• A device now connects to the library you actually chose. Before, every computer sharing a library heard every attempt and whichever answered first won, so with two libraries on one licence a phone could quietly reconnect to the wrong one. Attempts are now addressed to a particular computer, and refused if a different one answers.
• Fixed: a computer that shares its own library could end up trying to connect to itself. If you have a library on two computers and both are sharing, going looking for the other one on the network made the computer you were sitting at raise its own "wants to connect" box, showing the very same three symbols as the real one, so there was no way to tell. Approving it left that computer permanently listed among its own devices, after which every later attempt was a silent toss-up between reaching the other library and quietly reaching itself. Maven now ignores its own attempts, refuses to list itself as one of your devices, removes any such entry it already has, and stops a connection outright if it turns out to have come back to the same machine. If you hit this, the stray entry disappears on its own after updating.
• The three symbols you compare when pairing are now shown on both screens. Settings, Devices was showing a four-digit number instead, which the other computer never displays anywhere, so the check you were being asked to make was impossible.
• Allowing a device to connect no longer sits on "Connecting…" for the best part of a minute. The commonest case is that the other device stopped waiting before you got to the box, and the answer in that case is reassuring rather than a failure: it is approved, and it reconnects by itself. Maven now reaches that answer in a few seconds and says what it is doing while it waits.
• Watching a video from your library on a phone or tablet should now start much faster, keep playing, and let you skip around without waiting. Two things were wrong and both had the same shape: when your phone stopped wanting a video, nothing told your computer to stop sending it, and asking for “the rest of the file” was taken literally, so a whole film would be sent when only a few seconds were wanted. Your devices talk over a single line, so all that unwanted video sat in front of everything you asked for next, which is why pausing or skipping meant waiting all over again. Cancelling a download is immediate now for the same reason, and PDFs benefit a little too.
• Background mode now tells the difference between starting itself and being started by you. When your computer signs in, Maven opens out of sight, because you did not ask for a window. When you open it yourself from the Start menu or Launchpad, the window stays where you can see it. Close that window and it tucks itself away as usual, so your other devices keep reading your library either way.
• Background mode now looks like part of Maven rather than a stray script. Its shortcut carries the Maven icon on both Windows and Mac, and on a Mac the menu bar shows the app icon instead of two letters. Its menu also only offers the action that applies: you get Hide when the window is showing and Open when it is not, rather than one of each with one of them doing nothing.
• Fixed: scanned books opened in Maven could show a few pages and then nothing but white ones. The reader was missing the part that turns a scanned page into a picture, so every page that came from a scanner stayed blank while the occasional photo or cover page appeared normally. On one 272-page book that meant three readable pages out of 272. It now opens in full, and the same fix covers older scans and books scanned in colour. If you have set a book aside because it looked broken, it is worth another try.
• Fixed: on a phone or tablet reading a library from your computer, Settings, Library offered to give that device a library of its own. No phone or tablet can hold one, so taking up the offer disconnected you from your computer’s library, failed, and left you back at the setup screen having to connect again to get your shelf back. It is now only offered where it can actually work. On a computer, choosing a folder and then changing your mind at the folder dialog leaves everything as it was, rather than disconnecting you first and asking afterwards.
• Fixed: on a Mac, the background mode setup command said it could not find Maven on computers where it was plainly installed.
• Fixed: Settings no longer offers to install Maven on a computer where it is already installed. It was asking the wrong question, "am I running as an app right now" rather than "is the app on this computer", so opening Maven in an ordinary tab made it look uninstalled and sent you to the how-to-install steps.
• The command that makes Maven run in the background now appears as soon as Maven is installed on that computer, instead of waiting until you had turned on sharing. It could not be found when people went looking for it, which is the only reason it exists.
• Fixed: a phone showing your saved copy no longer also nags you to approve it. If the computer holding your library was switched off, your phone correctly fell back to the copy it had already saved and said so, and then a card sat on top of it asking you to go and approve this device on that computer, which was not even running. That card now only appears when a device really is waiting to be let in for the first time.
• ⚠ BETA: you can watch your library on the television, using your phone as the remote. It is marked beta in the app because it is new and still rough, so treat it as something to try rather than something to rely on. Turn the TV on, tap TV in the bottom bar of your phone, and pick your Chromecast. From then on anything you play goes there instead of to your phone, and your phone becomes the remote: play, pause, skip and a scrubber. Tap TV again to stop, and it plays on your phone again. It picks up where you left off in both directions, so you can start something in bed, put it on the TV, and later carry on from exactly where the TV got to. Only video goes to the television, so opening a book while casting simply opens it on your phone. ⚠ Two things worth knowing: the video comes from the computer holding your library, so that computer has to be awake with Maven open, exactly as it does for watching on a phone. And it needs a real Chromecast or Google TV — a Fire TV cannot do it even with a casting app installed, and neither can an Apple TV.
• The Maven icon now tells you when something needs you, without you having to look at the window. If you set Maven to open by itself when you sign in, the computer holding your library is a window nobody looks at, so the one thing that really matters there was invisible: if your browser wanted permission to open your folder again, your phone and tablet quietly saw an empty shelf and nothing on the computer said why. Now a small mark appears on the app icon in your taskbar or dock, and hovering over it says what it is. It also shows a count while items are downloading to a phone or tablet, and a mark if a download failed, if a device is waiting for you to approve it, or if there are new files in your folder to look at. It only appears when there is genuinely something to see, which is the point: a mark on the icon always means something. It needs Maven installed as an app.
• Maven now marks its own icon when a long job finishes while you are doing something else. Cataloguing a stack of books, or asking a real question of your library, can take a minute or more, and most people switch to something else while they wait. If you do, the icon lets you know it is done. If you sat and watched it finish, nothing appears, because you already know.
• If you like, the app icon can also show how many of your devices are reading your library right now. It is off by default and lives in Settings, Devices, under the sharing switch. Left off, the window title still tells you whenever you hover over Maven in your taskbar.
• The “reconnect your library” screen no longer looks like bad news. Every so often your browser asks Maven for permission to open your folder again, after you quit it, restart the computer, or leave it a while. The screen now says so plainly, names the folder you already chose, promises nothing has moved or been deleted, and offers “Open my library” instead of “Reconnect”, which sounded like you were about to set everything up from scratch. There is a matching answer in Help, too.
• When a device can’t reach your library, Maven now tells you why instead of just saying it isn’t connected. Your devices talk to each other directly over your own Wi-Fi and never through the internet, which is what keeps your library private, but it also means a phone on mobile data, or on a guest network, simply cannot see your computer. It now says so plainly, and says it outright if it can tell you’re on mobile data. The computer holding the library also speaks up now when a device tried to connect and couldn’t, with the reason, rather than leaving you to guess from the other end.
• A device that needs signing in again now says that, instead of looking like a connection problem. It also stops retrying, because retrying was never going to fix it.
• Maven can open by itself when you sign in to your computer, so your phone and tablet can always reach your library. Your other devices can only read it while Maven is open on the computer that holds it, which is easily the most common reason sharing appears not to work. In Settings, Devices there is now a small helper you save into your library folder and run once: after that Maven opens minimised every time you sign in, and running it again undoes it. It needs Maven installed as an app on that computer first.
• Maven now opens on a phone with no signal at all. It was meant to already, since your library is kept on the device, but opening it in airplane mode could stall on the loading screen instead, because it was keeping only ONE saved copy of the app and handing it back for whichever page you opened, which the app then couldn’t make sense of. It now saves each screen separately, and if anything does go wrong while starting up you get a real message instead of a spinner that never stops.
• On a phone, Maven is now an installed app. Once your licence is active you’ll be asked to add it to your Home Screen: one tap on Android, three short steps on iPhone and iPad. It is not a formality: iPhones only allow an app to notify you about a finished download once it’s on the Home Screen, downloads are far less likely to be cleared away to free up space, and it keeps running longer when you switch apps. You can still activate your licence and read anything on the website in an ordinary mobile browser.
• On a phone you can now see a download without hunting for it. The More tab shows a small ring while anything is still outstanding, and opening it lists what is downloading: the file, how far along, and how many are waiting behind it, with a tap straight through to Downloads. It also stays visible when a download is paused, which is what happens whenever you switch apps, so a download that quietly stopped no longer looks like one that vanished.
• Tapping a download notification now opens Maven on the Downloads page. Before, it did nothing at all.
• Fixed the download bar at the top of the screen being see-through, so the page scrolled visibly through it.
• Sharing your library with your other devices works again. Turning on “let my other devices read this library” looked like it worked, since the switch moved and the QR code appeared, but the choice was never actually saved, so the computer quietly never started listening, and a phone scanning the code sat waiting for a library that was never going to answer. ⚠ Two things to do once after updating: switch sharing back on, and pair your phone again. Anything you approved while this was broken was not recorded either, so it needs re-adding. After that it stays put.

• Maven no longer describes the wrong item when two things share a name. Add a release with a short title — Mirroring, Anchors, Silence, Fate — and the shops are full of unrelated products called the same thing, usually ranking far above the one you own. Maven would read whichever it found first and write it up with complete confidence: a description of gaffed cards on a mentalism PDF, and the wrong cover to go with it. It now insists that a page actually names the creator you gave it before it will believe a word of what that page says, and if the only pages it can find are a different product of the same name, it says so and searches again for the right creator instead — looking at their own store and their other releases the way you would. About one item in five has a title short enough for this to have been a real risk.
• Self-published items sold on Gumroad now work. Gumroad is where most independent magic and mentalism is sold, and Maven did not recognise it as a real shop — so it found the creator’s own page, discarded it, and then, having nothing legitimate to point at, credited whichever pirate mirror it had also read as the “publisher”. It now treats Gumroad, Payhip and Ko-fi as the creators’ storefronts they are, reads their pages properly (a Gumroad page hides its entire description in a place Maven was ignoring — 2,600 characters of it), and knows the right web address to look up a title on them. Payhip and Ko-fi shops are covered too.
• A shop is no longer mistaken for a publisher. If an item was self-published — as most modern mentalism is — the publisher is the creator, not whichever dealer happens to stock it. Maven had been filling that field with the name of the first shop it saw, which for independent releases meant a download site, and occasionally something with no connection to magic at all. It now names the creator, or leaves it blank when it genuinely cannot tell.
• A pirate site can no longer end up listed as an item’s publisher. Maven reads those pages for facts when nothing better exists — that has always been deliberate — but it was then repeating their shop name as if it were the work’s publisher. It now leaves the field blank instead, because no publisher is better than the wrong one.
• Credits are kept now. If a shop page says a routine builds on someone else’s work — “combines methods from Marc Paul’s AAA Book Test, used with full permission” — that stays in the record, naming the creator and the work. It is often the most useful line on the page: you may already own the underlying method, and it is how you would find it. The same goes for how something is done, as far as the page says: gimmick or sleight, what is supplied, whether it resets, angles, difficulty, and any named principle.
• Customer reviews are read properly now. For a recorded lecture the shop’s own page usually never says what is taught — but the reviews often do, routine by routine. Maven now reads reviews and buyer questions as part of the page rather than skipping past them, so lectures can finally get a real routine list.
• Maven stops pretending a lecture has a routine list when nobody published one. Penguin LIVE and At-the-Table pages almost never say what is actually taught — they carry a blurb, some quotes and a biography of the lecturer — but Maven had been told those pages always list the routines and never to come back empty-handed. So it filled the gap with plausible guesses. It now leaves the list empty when no real one exists, and looks for the routines where they genuinely turn up: the reviews on the page, Martin’s Magic, and discussion threads.
• Single tricks now get a proper write-up instead of two vague lines. For a book, the chapter list carries the detail — but a single effect has no chapter list, so its description IS the record, and Maven was still summarising it down to a sentence or two. It now keeps what the shop’s page actually says: what the audience sees phase by phase, what you have to supply, what can be swapped or customised, whether it resets, what’s in the download, and who endorsed it. It also uses the page’s own words for the plot — “book test”, “PK touches”, “ACAAN” — because that is what you would search for.
• Maven reads shop pages far more accurately. On a typical product page, only about a fifth of what a reader sees is the actual description — the rest is menus, “you might also like” rails and footers — and Maven was hunting the useful part inside all of that. It now strips the shop away first: on one real page that took 22,000 characters of clutter down to 4,300 of pure description, which is why the write-ups above got much better.
• Maven no longer marks a description as complete when it only came from the AI’s own memory. If the shop page couldn’t be read, anything written from general knowledge alone now counts as unverified rather than confirmed — and the fact-checking step that is supposed to strip unsupported claims was being handed that same guesswork as its “source”, so it approved things like “close-up, minimal setup, impromptu” on a routine that is actually a stage book test. It now checks only against pages it genuinely read.
• Fixed a formatting bug that could squash an entire contents list onto one line, separated by commas.
• Every button in an item panel is now properly greyed out while research is running, instead of looking clickable and doing nothing.
• Buttons in an item now show what they did right where you pressed them. Re-research, Cover image and Linked file all live at the bottom of the panel, but their status message and their controls were appearing at the very top of the item — so on anything with a long contents list you would press Re-research and see nothing happen, unless you thought to scroll back up. The “working…” message, the cover box and the file picker now open directly above the button you pressed.
• Fixed: items from Vanishing Inc, Tannen’s, Alakazam, Ellusionist, Theory11, Stevens and a dozen other shops were coming back with nothing at all. Those shops block automated readers, so Maven has always had a fallback that reads the page properly — and the fallback itself was being turned away, on every single attempt, because of one wrong line in how it introduced itself. The result looked exactly like “there is nothing online about this item”, when in truth the page was perfectly readable. Of 25 shops that were failing, 24 now work. If you have items that saved empty or thin from any of those shops, re-building them should now fill them in.
• Re-building an item can no longer make it worse. Shops go down, rate-limit, or just refuse to answer — and when that happened mid-rebuild, Maven would replace a perfectly good entry with a hollow one, which is easy to miss when you are re-building a whole shelf at once. It now compares what came back against what you already had: if the new research can’t point to a single page it read and the old entry could, your entry is left exactly as it was and it tells you why. If a contents list shrinks a lot, it says so on the row rather than quietly changing it, so you can decide.
• Maven uses your own library to find obscure releases. If a creator you own several things from publishes on their own small store, Maven now knows that — it looks there for their other titles instead of trusting a web search that buries a $9 PDF under a mass-market trick with the same name. It works out the store from the items you already have, including ones where the creator is only a co-author, and goes straight to the right page.
• It no longer guesses what an item is from its name. With nothing to go on, Maven used to turn the title itself into a description — “Mirroring” became a routine “about reflection and psychological techniques”, when it is actually a PK Touches piece. Magic titles are chosen to sound good and often have nothing to do with the method, so it now says plainly what it can see (who made it, what the files are) and admits when it couldn’t find out more, rather than writing a confident paragraph out of thin air. Same for the tags — a made-up tag makes an item turn up in searches it has nothing to do with.
• When Maven can’t be sure, it now says so rather than sounding certain. An item is only marked as fully catalogued if a source actually confirmed it was the right item, and the publisher is left blank rather than filled in from a page Maven wasn’t willing to link to. There is also a new entry under “Needs attention” in the library — “Description with no source” — listing items that read confidently but have no page behind them, which is where an older mix-up would be hiding. Re-building one is usually all it takes; most of them are fine.
• Contents entries have to actually say something now. If Maven found a chapter list but no description of what each chapter contains, it used to stop there and save the bare names — which tells you nothing when you search. It now goes looking for the real product description, on every add and every rebuild, and writes a short “what it is” against each entry while keeping the published list exactly as it was.
• Your file names help write the description, rather than just sitting at the bottom. Maven cross-references them against what it finds online, so an entry names the routine AND points at the file you would open to watch it. For something obscure that nothing online covers, the file names alone are now enough for a modest, honest summary — it will say what the files show it contains, and still won’t invent how a trick works.
• Scanning a folder now links the files as it goes. Cataloguing found your items and then left them unlinked, so you had to go to Settings and run “Find my media” afterwards before you could open anything. Maven knew exactly which files it had just read; it simply forgot before saving. Now the item and its files arrive together, in one step.
• New files are noticed without a reload. Drop something into your magic folder, click back to Maven, and it tells you what turned up — so you can sort your drive with the app open beside you and it keeps up. Hiding the notice with ✕ now sticks for the rest of the session, but only for what you hid: add something else and it says so.
• Maven notices when you add things to your folders. Drop a new release onto your magic drive, open Maven, and it tells you what turned up — by name — and offers to catalogue just those files. No hunting down the same folder in a file dialog again, and it only looks at what is new, so adding to a folder you have already catalogued costs almost nothing.
• Your own file names now appear in an item’s contents. If a set’s files are named after the routines on it — and they usually are — those names become part of the record, so searching for a routine finds the item and tells you which file it is in. Before, the contents came only from what was written about the item online, which for a video set is often much thinner than the names already sitting on your drive.
• Rebuilding an item keeps what is on your disk. Re-researching a linked set used to replace its contents with whatever a shop page said. Now your real file names go into the rebuild as well, so it can only add to what you already have — worth running on anything you have linked to a folder.
• Links repair themselves. Move your library folder, plug a drive into a different computer, or reorganise your folders, and your items no longer break: Maven looks for the same file in the other folders it knows about and quietly fixes the link. If a batch of files has moved, it offers to update them all at once.
• One list of folders, instead of three ways of saying the same thing. Settings now shows everywhere Maven looks for your files — your library folder first, then any drives — each with the same button: check for new files. A drive that is unplugged says so, instead of looking broken. And “Find my media” is now called “Match files to notes”, because that is what it is: a repair for the odd item, not a step you are meant to remember.
• Two fixes worth knowing about if you keep your media on a separate drive. Re-scanning that drive used to send every file back to the AI and charge you for it all over again, only to throw the results away as duplicates. And adding a second folder with the same name as one you already had (two folders called “Magic”, say) could point every link on the first one into the wrong drive.
• A big library is much faster to browse. Maven now keeps its own small copy of each cover the first time it sees one, instead of loading the full-size picture from the shop it came from every single time. On a 750-item library that took the amount of image the browser holds from about 1.6 GB down to 40 MB, and scrolling is now completely smooth. It happens quietly while you browse — nothing to switch on and nothing to wait for — and the covers actually look sharper, because they are now sized for the shelf instead of being squeezed down from a poster.
• Your library shelves itself. Two new sort options group everything into headed sections as you scroll: by author, or by subject — card magic, mentalism, coins, stage and so on, worked out from what each item is about. Authors with three or more items get their own heading, the biggest shelves first, and everyone else is gathered at the end so you don't get a page of headings with one book under each.
• A "Continue learning" row across the top of the library picks up whatever you were part-way through, with a bar showing how far in you are. Tap one and it opens straight back to where you stopped. It only appears when you have something on the go, and only for items with a file attached — a shelf of paper books never sees it.
• Your progress shows on the library itself now, not only once you open something. Anything you are part-way through has a bar across the bottom of its cover.
• The author list is a panel you can put away. It used to hold the left side of the window open permanently, even though most people find an author by typing rather than scrolling a list — and the button that opens it now says "Authors" instead of "Filters", because that is what it is. Hiding it fits another column of covers on screen rather than just making the existing ones bigger.
• One library view instead of two. The list view and the covers/list switcher are gone: the covers view now carries the progress bars, the shelves and the artwork, and the list was the same information with all of that taken away.
• The website now shows what reading your library on your other devices actually looks like. There is a section for it, "Your own personal magic Netflix", with your library on a computer and your phone playing off it alongside. It covers the three things people ask about: nothing gets copied anywhere, it remembers where you stopped on whichever device you pick up next, and you can pull items onto a device to watch with the Wi-Fi off. It also says separately that the computer holding your library needs Chrome, Edge or Brave, while the devices reading it can be anything with a browser.
• Opening an item now looks like opening something on a streaming service. The cover is the hero — blown up and softly blurred behind a crisp copy of itself — with the title, the author and one clear button to read or watch it. Everything for tidying up the record (editing, re-researching, changing the cover, the linked file, deleting) has moved behind a single ⋯ button, so the thing you actually came to do is no longer one of six identical grey buttons in a row.
• Your place in a book or a video is now shown on the item itself. A bar across the cover and a line telling you how far in you are — "Page 64 of 205", "50 min left" — and the button says Resume rather than Read or Play. Maven has always remembered where you stopped; it just never told you outside the player.
• The item window no longer changes size while you use it. The cover and buttons stay pinned at the top, the buttons at the bottom stay at the bottom, and only the description in the middle scrolls — so a book with a long contents list no longer scrolls its own title off the screen, and opening the ⋯ menu doesn't make the window grow.
• You can line up several downloads at once. Press "Keep offline" on as many items as you like and they queue up and come down one after another, in order — before, a second one fought the first for the connection and the progress bar could only show you one of them. Each item shows its own progress in the library, and Settings → Downloads lists the lot.
• Downloads pick themselves up again. Switching to another app on a phone cuts the connection — nothing any website can prevent — and that used to leave you with a failed download and a button to press. Now it simply pauses and carries on the moment you come back or your other computer is reachable again, starting from the next file rather than the beginning. Maven also keeps the screen awake while a download is actually running, and if you allow notifications it tells you when one starts and buzzes once when it is ready — updating the same notification rather than stacking up a new one for every item.
• You can stop a download. There was no way to before: a four-gigabyte set started by mistake ran to the end or you closed the tab. Stop is now on the item, on the bar at the top, and in Settings. Anything that had already finished stays on the device, so pressing Keep offline again picks up where it left off rather than starting over.
• Download a whole shelf at once. Filter your library however you like — one author, videos only, whatever you just searched for — and a bar above the results offers to download everything there that isn't on the device yet. It tells you how much space it needs and how much you have before it starts, and warns you if it would nearly fill the device. Handy the night before a flight.
• Settings has been rebuilt so you can find things. Each section now opens with a small illustration and one line telling you what it is for, and the settings under it are grouped into cards with real headings instead of a stack of identical boxes labelled in tiny grey capitals. Devices was the worst of it — seven equal panels covering three unrelated jobs — and is now sorted into this device, sharing your library, and reading someone else's.
• Your AI and your usage are one place now. "How many items have I got left" and "use my own OpenAI key" were on separate tabs, which is odd, because they are two halves of the same decision. The AI section now opens with what is included and how much of it you have used, then what it went on, then the optional key.
• Downloads has its own section, on the devices that actually have downloads. It shows what you have saved for offline, how much room is left, and — this is the new part — whether your browser has promised to keep them. That last one only ever appeared as a warning, so if you were fine, you had no way to check before getting on a plane. Now it tells you either way.
• Your licence page says what you actually bought: AI for cataloguing, and your devices, each with a link to where you manage it. It used to advertise "coming soon: Maven AI included" — which had, in fact, already arrived and was included in what you paid.
• Small grey text throughout the app is now readable. Explanations under settings, hints beside buttons, the notes on what each option does — a lot of it was set in a grey so dark it disappeared against the background, which is fine on a bright monitor in a dark room and not much use anywhere else. Every one of them has been brought up to a proper reading contrast.
• Videos turned sideways now show their play and pause controls. They were 84 pixels below the bottom of the screen — the player was sizing itself to a 16:9 shape instead of to the screen it was on.
• PDFs have a dark mode. One tap in the floating bar drains the colour and inverts the page, so a black-on-white scan becomes white-on-black for reading in bed. It sticks until you turn it off. How well it works depends on the PDF, but for the ordinary text-on-paper kind it is a big improvement.
• PDFs open showing a whole page, on whatever you are reading on. A tall phone fills the width as before, but a squarer screen — a folding phone opened out, a tablet, a computer — now fits the page top to bottom with margins either side, and that is also as far out as you can zoom. Rotate, fold or unfold, or hide the contents list and it re-fits itself; once you pinch to zoom it leaves your zoom alone.
• You can close the contents sidebar when reading. On a folding phone, or any wide screen, a PDF used to open with the contents list taking a chunk of the page and no way to put it away. The button in the top-left hides it and remembers that you did — and it slides out of the way rather than snapping.
• Adding a second device is smoother. Following one of your device links now drops you straight at your library and asks one question — does this device hold your library, or read one over Wi-Fi — instead of occasionally flashing a “choose a folder” screen at you after you had already answered. That flicker was a timing bug, and it is gone.
• While a device is looking for your library it shows a little animation of two devices talking over Wi-Fi, so you can see it is still trying rather than wondering whether it gave up. It also says “connecting” while it is connecting, instead of telling you to go and switch sharing on when it is already working — and it will never flash an empty library at you before your items arrive.
• Adding a device by scanning its QR code or opening its link no longer needs approving on the other computer. The link is the approval — you made it yourself, on the machine holding your library, and it only works once and only for a few minutes. Your phone now goes from scan to reading your library without you walking back to your desk. The computer that holds the library tells you it happened, and Settings → Devices can remove it.
• If you ever do add a device by typing your licence key instead, the two screens now show three matching emoji to compare rather than a four-digit number — quicker to check at a glance, and it is worth checking, because allowing a device is permanent.
• Setting up is shorter and clearer. There is no longer a step asking you to connect an AI account — your purchase includes it, and if you ever do run out, the app takes you to the right place at the moment you need it. So it is two steps on a computer, one on a phone.
• The setup screens have had a proper going-over. Each one now leads with a small illustration and a single clear line instead of a paragraph, the progress marker is out of the way rather than competing with the question, and on a phone the whole thing fills the screen instead of squeezing into a card.
• If you open Maven in Safari or Firefox, the explanation is shorter and straighter: it is a limitation on Apple and Mozilla’s side rather than something to wait for. Carrying on there is now marked "not recommended" so the trade-off is visible before you pick it.
• Setting up a phone or tablet is one visual screen: what it does, and one button to find your library. The instructions for the other computer are tucked behind a "nothing found?" link, where they belong.
• Installing Maven is now one tap. On Android and on Chrome or Edge, tapping the banner opens the browser’s own install dialog — it had been quietly handing out written instructions instead, because the browser offers that dialog earlier than the app was listening. On an iPhone, where no such dialog exists, tapping opens the two steps you need. Settings has a working Install button too, on every platform. Maven also has a proper app icon now, so installing it puts the Maven mark on your home screen rather than a picture of the page.
• Maven now asks you to install it. It always could be installed — its own icon, its own window, no browser bars in the way — but nothing ever offered, so unless you went looking in Settings you would never have known. On an iPhone or iPad it now offers BEFORE you pair the device, because a home-screen app has its own private storage: pair first and install second and you would have had to pair all over again. You can always say "set this up later".
• The notes list works on a phone. Every saved note was dimmed and refused to be tapped, and scrolling a note dragged the list about behind it. Both are fixed.
• Modals stay on the screen. A long item — one with a big table of contents — used to grow taller than the phone, so parts of it simply could not be reached, and the bug-report box opened jammed against the top edge. Everything now sits centred, at a size that fits, and scrolls inside itself.
• The page no longer scrolls about behind whatever you have open. Opening an item, a video, the filters or the contents list pins the page underneath, which also stops the address bar sliding in and out and making the panel look like it is coming apart.
• Turned sideways, the play and pause controls are actually on the screen. They were sitting just below the bottom edge, hidden behind the browser toolbar.
• Pinch to zoom a PDF. Two fingers zoom the document itself and it re-draws sharp when you stop, so the small print in a scan is readable. The +/- buttons are gone from phones — pinching is what you would reach for — and the space went to a Contents button, now part of the same floating bar as search and the page number instead of a second thing floating nearby.
• The documentation has a contents bar that follows you down the page on a phone, telling you which section you are in and opening the full list at a tap. It also stopped telling you to go and use a computer, which it was doing on the page that explains your phone is supported.
• On the notes page, Delete has moved away from Re-search, and confirming a delete no longer puts "yes" exactly where your thumb just was.
• Maven is now built for a phone, not merely squeezed onto one. Library, Ask and Notes sit in a bar along the bottom where your thumb already is, the top of the screen went from four rows of controls down to two, and the sort, view and filter controls now slide up from the bottom as a panel with every option shown at once instead of hiding behind small menus. Your items start near the top of the screen rather than a third of the way down it. Nothing changed on a computer.
• Watching and reading on a phone now works the way you would expect it to. Turn the phone upright for a video and you get the picture across the full width with the contents, files and bookmarks scrolling underneath it — tap any chapter to jump. Turn it sideways and the video fills the screen edge to edge, with the contents list a tap away over the top of it. PDFs get the whole height of the screen either way, since that is how a book wants to be read. It changes the moment you rotate the phone; there is nothing to set.
• Books with their own website now catalogue properly. A lot of self-published books and lecture notes have a dedicated page for the title, with the full list of routines a click away behind something like "for the complete contents, click here" — Maven now finds that page and reads the whole list, instead of stopping at the blurb on the front page. Titles that used to come back with nothing but a name should now fill in properly when you rebuild them.
• Adding a batch no longer quietly loses tables of contents. When you catalogue a lot of items at once, the sources Maven reads page numbers from would sometimes ask us to slow down — and the item was saved as though the book simply was not listed there. Every source now waits and asks again, and if one still cannot be reached, the item is marked as unfinished so you know to rebuild it rather than being told there is nothing to find.
• Covers now come from the book's own page. Maven looks at the item's official page first and takes the picture from there — including on self-published sites that just have a photo of the book on the page and nothing a computer would normally recognise as a cover. That is a much better result than the generic image search it used to fall back to, which sometimes attached the wrong book entirely.
• Research digs further into each item. Maven now measures how much looking each source actually takes instead of over-reserving for the worst case, so the leftover effort goes into reading more of the pages it found — in practice two or three more sources per item, which mostly shows up as fuller contents lists.
• The AI is now included — there is no OpenAI account to set up. Your purchase comes with 500 items of cataloguing and unlimited searching, so you can buy Maven, point it at a folder and start adding your library within a minute. No developer platform to sign up for, no billing to configure, no key to paste. If you would rather use your own OpenAI key you still can, and when you do, none of your included items are touched.
• Buying it takes one step now. You are brought straight back into a working app with nothing to copy, paste or find in an email — the licence sorts itself out. If you ever do need your key, or a receipt, Settings → License has a link to your account page.
• You can see exactly how much cataloguing you have left. Settings → Usage shows the number of items remaining out of what came with your purchase, and you can add more there whenever you want. Searching never uses any of it, however much you search — only adding something new to your library does.
• Adding a big folder no longer stops dead. If you point Maven at 750 items and have room for 500, it tells you before it starts, catalogues everything it can, and lists the rest as waiting rather than failing them. Add more items (or connect your own key) and one press of Retry finishes the job — nothing that already worked is researched again.
• Maven now updates itself. When a new version goes out you simply get it the next time you open the app — no clearing your browser cache, no hard refresh, no wondering whether the thing you just read about is actually there yet. If you are in the middle of something it waits until you move to another page rather than interrupting you, and your library, notes, saved progress and anything you have kept offline are untouched by an update.
• Your library now works on your other devices. Sign in on a laptop, tablet or phone with the same licence key and it finds the computer that holds your library over your own Wi-Fi — you approve it once (checking a four-digit number matches, the way you would pair headphones) and after that it just connects. Everything you own, every note, every table of contents, searchable from the sofa. Nothing is uploaded anywhere: the two devices talk directly, and it only works when both are on the same network.
• Watch and read straight from the other computer. Press play on a video and it starts almost immediately rather than downloading first, and you can scrub through it freely — the same for a big PDF, which opens on page one instead of making you wait for a 200MB scan. Files are not copied onto the second device unless you ask; watching something leaves nothing behind.
• Your place follows you between devices. Stop a video halfway through on the sofa and it picks up there on the computer, and the same for the page you reached in a PDF. Bookmarks, notes, and your watched/read ticks all travel both ways — and if the computer holding your library is asleep, your progress is remembered and sent across the moment it wakes.
• Ask works on the second device without setting anything up. No API key to enter, no settings to copy — searches run on the computer that holds your library, using the key already there, so it is all one bill and one usage log. Saving a search as a note works from either device.
• Take things with you for a flight. “Keep offline” on any item downloads it to that device, and from then on it plays with no connection at all. Settings → Devices lists everything you have kept with its size so you can clear space, and tells you honestly if your browser might remove downloads — worded for the device you are actually on, rather than warning everyone.
• Maven can be installed like a real app — its own window, no browser bars, and your library a click away. On an iPhone or iPad this matters beyond convenience: adding it to your home screen is what stops downloads being cleared after a while. Once installed it also opens with no connection at all, so a library you have synced and files you have downloaded are there on a plane.
• The player no longer offers to cast to your TV, because it could never actually do it. Casting works by handing your TV a web address to fetch for itself, and your files do not have one — they are on your own computer, which is the whole point. The button opened a device picker and would only have failed after you had chosen a TV, so it is gone. Putting a lesson on the big screen still works the ordinary way, by casting or mirroring your whole screen from your phone or computer.
• Settings on a phone or tablet only shows what that device can actually do. Reading your library from another computer means there is no folder here, no files of your own to link and no AI key to set up — so those panels are gone rather than sitting there doing nothing. The AI one was the worst of them: it asked for a key that would never have been used, because searches run on the computer holding your library. What is left is your devices, your licence, and a line telling you which computer your library is on — plus a way to give this device a library of its own if you change your mind.
• Setting up a second device takes one scan. Settings → Devices has an "Add another device" button that shows a code — point your phone or tablet at it and that device signs itself in and opens your library, with nothing to type and no key to find. There is a copyable link too, for the laptop in the other room. Both stop working shortly after you create them, and your licence key is not readable in either, so a link that goes astray is not your key going astray.
• Moving to a new computer is now a supported thing rather than a puzzle. Settings → License has "Moving to another computer?", which gives you a link that signs you in there — get it before you sign out here, because signing out first used to leave you stuck.
• Signing out of a device no longer affects your other ones. If your library folder syncs through Dropbox or iCloud, signing out on the laptop used to reach into that folder and remove your licence from it, so your desktop quietly lost the ability to sign itself back in. And in the other case — no OpenAI key of your own, which is now most people — signing out did not actually work: the folder put the licence straight back on the next visit. Signing out is now about the one device you are sitting at, every time.
• Setting up now starts by asking where your library lives, instead of assuming. “On this computer” is the same folder step as before; “On another computer” sets this device up to read a library you have already built elsewhere, over your Wi-Fi. That second path existed but was a single line of text at the bottom of another screen, which is no use if you are setting up a laptop. Either answer can be changed later in Settings, and it says so.
• And if you open Maven in Safari or Firefox, the first thing it does is tell you what that costs — before you have put any work in. You get one screen with three honest options: open it in Chrome, Edge or Brave (with a link to copy across), read a library that lives on another computer, or carry on where you are. Only the last one has a catch, and it spells the catch out before you choose it rather than after: your library would live inside the browser, clearing your browsing data would erase it, and your own PDFs and videos would not open. You see it once, not every time you launch.
• If you use Safari or Firefox, Maven no longer overstates how safe your licence key is. It keeps your key and licence in a hidden file in your library folder precisely so that clearing your browser does not cost you them — but in Safari and Firefox that “folder” is inside the browser, so clearing your data takes the lot. The app used to promise otherwise. It now tells you plainly that your key lives in this browser only, and Settings → Library says that an export is your only backup rather than merely suggesting you take one. Nothing changed about where anything is stored; what changed is that it no longer reassures you at the moment you are deciding whether to write your key down.
• Safari and Firefox no longer hand you buttons that cannot work. Maven keeps your library as real files in a folder you choose, and only Chrome, Edge and Brave let a web page do that — so in other browsers your library lives inside the browser instead. What changed is honesty: “Scan a folder” used to sit there and fail when you clicked it, and an item with a video or PDF attached claimed the file “may have moved or been renamed” when in truth that browser simply cannot open files on your computer. Both now say what is actually going on, and where something needs Chrome, Edge or Brave it says so on the spot rather than leaving you to work it out.
• The library page is far calmer to look at. The kind of item — books, videos, tricks, lectures, collections — is now a row of coloured buttons across the top with a count on each, and each kind keeps its own colour on every cover, so you can pick out your videos at a glance. The author list has the left-hand column to itself.
• You can finally see what you are filtering by. Everything you have narrowed to shows up as a small tag above your items — “Type: Book”, “Author: Osterlind”, whatever you typed — and clicking a tag removes just that one. There is a “Clear all” next to them, and the count tells you how many items you are looking at.
• Typed something into the library search and want a real answer instead of a title match? There is now a ✦ button in the search box that hands the same words to Ask, which looks inside your items rather than just at their names. Press “/” anywhere to jump into the search box, and Escape to clear it.
• Sorting says what it does. Instead of one button you clicked repeatedly to cycle through hidden options, there is a short menu listing “Title A–Z”, “Recently added” and “Last opened”. The two overlapping list layouts are now one.
• The tidy-up filters are out of the way until you want them. They sit in one “Needs attention” drawer at the bottom of the left column with a count of how many items need looking at, and each one is named in plain English now. “Needs cataloguing” gathers just the items whose file is actually thin — nothing captured, or a summary with no chapter list — so it points at the ones worth re-building rather than half your shelf. When you have filtered to the items that need fixing, one button ticks them all for a re-build. Closing the drawer clears its filters, so you are never left looking at a slice of your library wondering where everything went.
• Browsing by effect no longer misleads you. It was quietly showing only the routines taught in two or more of your items while the heading counted every entry, so the number never matched the list — and applying any filter silently switched it to showing everything. You now choose: “Taught in 2+ items” to find where your library overlaps, or “Every effect”, with an honest count on each and your filters left alone.
• The menu across the top is simpler. Ask, Library and Notes each have an icon, Settings and Help sit off to one side behind a divider, and the bug-report button has moved out of the way to the far right. Help now opens the documentation right inside the app instead of throwing you into a new browser tab.
• “Find my media” no longer links the wrong file. It used to treat a shared “live”, “lecture”, “volume” or even “by” as a real match, which is how one performer’s Penguin Live lecture ended up attached to another performer’s — often ticked and ready to apply. Lecture-style titles (“Bill Malone LIVE”, “… At The Table”) now only link to a file that actually names that performer, and a numbered volume will no longer attach itself to a different volume of the same set.
• Research no longer stops halfway without telling you. There is a hard limit on how much digging Maven can do for one item in a single pass, and busy titles were quietly hitting it — the entry got saved looking like nothing could be found, when really it had run out of room. It now spends that effort deliberately, on the best sources first, and if it does run short it says so and marks the entry as incomplete so you know to rebuild it rather than assuming the item is a dead end.
• Fewer items fall over in the first place. A research request that stalled had nothing to stop it, so it would hang until the connection was cut from underneath it — which is what produced that run of “AI request ended without a result” messages. Requests now give up cleanly if they take too long and simply try again, the connection is kept alive during the long stretches of research, and when something genuinely does drop you get a message that says so in plain English instead of a technical one.
• Adding a big batch survives a shaky connection. Items that fell over mid-research used to be lost for good — on a 400-item folder that could be sixty of them, and the only way back was re-scanning the whole folder and paying to re-research everything that had already worked. Each item now quietly takes a second run at it when the connection drops, and anything still missing at the end gets a “Retry failed” button that re-runs only those. Genuine dead ends (an item nothing online knows about) are not retried, so nothing is charged twice for the same answer.
• Maven now reads the author straight off the title, so far fewer items get held back for “missing info”. If a scan, photo or store import gives you “CBS by Paul Vigil”, “Bill Malone LIVE”, “Chad Long Penguin Live Online Lecture” or “Docc Hilford Lecture Notes”, the creator is filled in for you — you just glance down the list instead of typing a hundred names. It stays cautious on purpose: “Step by Step” and “The Lost Lecture” are left for you, because a confidently wrong author sends the research off after the wrong item.
• Filling in a missing author no longer fights you. The moment you typed a single letter, the row leapt out of “Missing info” into the list above and took your cursor with it, so you had to find it again for every item. Rows now stay exactly where they are while you type; a ✓ (or Enter) tucks a finished one up into the review list and drops you straight into the next one that needs an author. If you never press it, nothing is lost — anything with an author typed in still gets added.
• “Find my media” also stops offering a file that another note is already using, so you can no longer end up with two items pointing at the same video.
• “Find my media” finds more of your library in one go. When two notes wanted the same file, only one of them got linked and the other was quietly dropped — even when its own copy was sitting there unclaimed. Every note now gets a second and third choice, so a near-miss no longer costs you the match. Nothing that matched before matches differently; there is just more of it.
• “Find my media” also stops guessing when it only recognises the author. Matching the creator’s name used to be enough to tip a half-recognised title over the line, so the right author’s wrong book could get linked — “Impossible Three” landing on “Six Impossible Things”. The title itself now has to match; the author only decides between files that already fit. Spelled-out numbers count properly too, so “Volume Two” finds “Vol 2”, and “Six Impossible Things” will not settle for “Three Impossible Things”.
• Covers are far stricter about what they will attach. Maven now only uses an image if it can tell the image actually belongs to that title — and anything adult, pirated, or that is just an animation or icon is refused outright, at every step. If nothing suitable turns up you get a clean placeholder instead of a wrong picture, and you can always paste your own. Any unsuitable cover already sitting in your library disappears the moment you open it, no re-cataloguing needed.
• Settings is now a proper page with its own sidebar instead of a cramped pop-up box — easier to read, and there is room for the usage history and device list to breathe. Each section has its own address too, so you can bookmark or link straight to it.
• Settings is less cluttered: your library folder and your media folders now sit together in one Library section, and the old "Maintenance" buttons are gone. To re-research items, right-click any item in your library to start picking, then hit Re-build media — same job, fewer buttons.
• The app opens noticeably faster. Maven used to re-check your licence with the store every time you opened a page; now it confirms it once and remembers, quietly re-checking in the background about once a day.
• Working offline is more dependable — your licence keeps working without an internet connection, so a flaky café connection or a flight no longer gets between you and your library.
• There is now a full Privacy page spelling out exactly what leaves your device and what never does — including how store imports work without ever asking for your Penguin or Vanishing Inc login, and where your OpenAI key is kept. Linked from the site header and footer.
• The website is much clearer that Maven never touches your actual files. You tell it what is on your shelf; your PDFs, videos and downloads stay in your own folders, and reading, watching and casting all run straight from your own machine.

• Ripped DVDs now convert into a single watchable video. If you have DVD folders (a “VIDEO_TS” folder full of VTS_… .VOB files), the conversion tool stitches each disc’s parts back together into one MP4 (skipping the menu files) instead of leaving you a pile of fragments. A disc becomes one video, named after its folder.
• Converting a big library is much faster: the tool now converts several files at once (tuned to your computer’s processor). If your media is on a slower external/USB drive, you can lower how many run at a time with one setting at the top of the script.
• Clearer conversion progress: instead of technical ffmpeg readouts, you now see a tidy live progress bar with a percentage and count, and the titles currently being converted listed beneath it (so you can watch them fly by). Small files convert first for quick early progress, big DVDs are saved for last, and even a long DVD shows a live percentage so it never looks stuck. Any failures are summarized at the end rather than cluttering the screen.
• The media viewer’s Files list now groups by folder: open a multi-part release and its subfolders show as collapsible folders with the files tucked underneath, so a set with “Vol 1,” “Vol 2,” “Vol 3” folders is easy to browse instead of one long flat list.
• Long file and folder names in that list no longer just get cut off; hover over one and the title gently scrolls so you can read the whole thing without leaving the viewer.

• Play old video formats in the app. Got AVI, WMV, RealMedia, or MKV files that only offered a download? Open one and the “can’t play this format” card now has a Convert Files button that takes you straight to the setup; or start from Settings → Media → “Scan for unplayable video.” It finds them, saves a ready-to-run conversion script into your library folder (Mac and Windows — it even tells you how to install the free ffmpeg tool), and after you run it, re-links your items to the converted files in one click. To avoid duplicating a large library, each old file is replaced only after its converted copy is checked and confirmed complete — anything that fails to convert keeps its original untouched (and is listed for you). Prefer to keep both copies? One checkbox.
• Re-scanning a media folder is cleaner: anything already in your library is now simply skipped, instead of being listed again as something to review.
• Folder scanning keeps multi-part works, volume sets, and seasons together: “Kayfabe – Part 1/2/3” becomes one “Kayfabe”; a numbered series like “VideoMind – Parlor Mentalism – Vol. 1,” “VideoMind – Stage Mentalism – Vol. 3” collapses into a single “VideoMind” with each volume as a chapter; and a show with seasons (including files named the “S2E1, S2E2 …” way) becomes one item per season with the episodes/sessions as chapters — instead of a separate item for every file.
• Scanned items with no author are held in a new “Missing info” list and left out of the add until you fill in a creator — so you never accidentally catalog a title with no one attached. Add an author and it moves up into the review list automatically.
• Scanning now links your files in the same step: after a folder scan finishes adding items, Maven immediately matches them to the actual videos/PDFs and shows you the proposed links to confirm — no separate trip to Settings → “Find my media.” If you scanned a folder outside your library, it’s added as a media folder automatically so the links work.

• See what the AI costs you: a new Usage tab in Settings tracks every add, re-build, and search with a running total (and a Clear-history button), and adding a batch now shows a rough cost estimate before you start — so a big import (or a new store) is no longer a leap of faith. Your usage is saved as a plain usage.csv in your library folder.
• Scanning a media folder is smarter about the folder your library lives in: it skips Maven’s own catalog files, and a re-scan skips media you’ve already linked — so re-scanning to pick up new files won’t re-process (or re-charge you for) what you already have.
• Folder scanning now recognizes a creator’s folder of separate effects: if you scan a folder named after a performer that holds individually-titled clips (like “Invisible Deck by Jane Doe,” “Card to Wallet by Jane Doe”), each is catalogued as its own item — instead of being lumped into one “chaptered” item.
• Penguin “Tarbell” video lessons (Tarbell 71, 72, …) are now correctly credited to Dan Harlan, who presents them — not Harlan Tarbell, who wrote the original books. Already have them credited wrong? Filter to “tarbell,” select all, and Re-build to fix the batch in one go.
• Searching for lectures works better: asking for a lecturer’s lecture now also finds items titled “… LIVE,” “… LIVE ACT,” or “… At the Table,” which is how most magic lectures are named.
• Re-catalog several items at once: right-click any item in your library to start selecting, click (left or right) to tick more, then “Re-build media” to re-research the whole batch. Selection follows whatever you’ve filtered or searched to, so you can narrow to an author or series, “Select all,” and refresh them in one go.
• A refreshed look for Maven — “The Modern Grimoire.” A calmer, deep-slate dark theme with a single understated accent, matte surfaces, and less background motion, so the focus stays on your library.
• A more powerful library search box. “|” means OR (“cassidy | jinx | osterlind”); mix in a kind (“osterlind | type:book” = anything matching Osterlind plus every book, “type:book triumph” = just books about triumph); and exclude with a leading “-” (“osterlind -absolute -creasey”). A ✕ clears it, and your search now sticks in the address bar, so it survives a reload and can be shared as a link.
• A new documentation page at /docs walks you through the whole app, with a table of contents that stays beside you as you scroll.
• Signed-in customers get a Contact support button in the docs, so a question reaches us directly and we can reply to your account email.
• Fewer “the file kept changing on disk” errors when saving to a cloud-synced library folder (OneDrive, iCloud, Dropbox, Google Drive) — Maven now waits longer for the sync to settle before giving up.

Best on a desktop browser

• On phones and tablets, a banner now points you to a desktop Chromium browser (Chrome, Edge, Brave, or Arc), where connecting your library folder and the rest of Maven work best.

Bug reporting & cleaner imports

• New: a bug-report button (the 🐛 icon at the top-left) opens a quick form to send a report straight to the developer. It automatically attaches what’s needed to reproduce the problem — the page, app version, your browser, what you were just doing, and any recent errors — and the developer can reply to you directly. Your library contents are never sent.
• Store-account import now reads the page’s structure, so it can reliably tell your own purchases apart from the store’s sidebar ads and “bestsellers” — importing your library from Penguin and similar stores captures far more of what you actually own.
• Obvious non-items — page numbers and buttons like “Watch”, “Learn”, “Discuss”, “write a review”, and “archive” — are now skipped quietly during add, instead of showing a distracting “not an item” line for each one.
• The “Import from a store account” instructions are laid out more clearly.

Accuracy & safety

• Maven no longer invents details for a title it can’t actually find anywhere online — instead of a plausible-but-made-up entry, it now saves a lean, honest record (searchable by title) that you can fill in later with Rebuild. Items with real info online — including details found only in reviews or forum threads — keep the same full, deep catalog as before.
• Stray page numbers and site buttons are filtered out before cataloging, so they can never turn into bogus “products”.
• Cover images are more relevant: covers come from the product page or trusted magic sources first, and Maven now leaves a cover blank rather than attaching an unrelated image.
• Tidier, easier-to-follow “Import from a store account” instructions.

More reliable importing, clearer messages

• Importing a large purchase list is far more reliable — long, paginated store pages are read in full instead of quietly dropping items near the end, and anything that couldn’t be read is now reported to you rather than silently skipped.
• Stray page numbers and site buttons (like “Watch”, “Learn”, and “Discuss”) are no longer mistaken for products when importing from a store account.
• Recognizes Penguin “LIVE” / “LIVE ACT” lectures and Murphy’s “At the Table” lectures, and pulls their taught-contents list when cataloging.
• Clearer, plain-language messages when your OpenAI account needs billing or credits set up — with a link to fix it — instead of a raw error code.
• Adding a big batch now stops early with one clear message if your OpenAI key runs out of credit, instead of showing the same error on every remaining item.
• Folder scanning is more reliable on very large libraries and tells you if any files couldn’t be catalogued.
• Handles cloud-synced library folders (OneDrive, Dropbox, iCloud, Google Drive) more gracefully when a sync is mid-update.
• Easier to install the “Import from store” bookmarklet in Microsoft Edge, with a copy-and-add option if dragging it doesn’t work.