Posted by trms 4 days ago
It’s been a surprisingly good middle ground between fully manual cards and fully LLM dumps. If anyone’s curious: https://wordwise.me
I've definitely hit walls with Anki over the years, and while the community decks help a lot, it's really nice to just tell Claude "can you take this assignment my tutor gave me, extract all the infinitive verbs, and then make cloze style cards for conjugations at an A1/A2 level?" and get it all done in a couple minutes.
In a way making the cards helps a ton to learn the content and decide what's really important to retain. On the other hand, it's such a slog that I usually end up relying on community cards, or skipping it altogether. The MCP idea may be a nice middle ground. Will give it a try for an upcoming exam.
> with provisions in place to ensure that Anki remains open source and true to the principles I’ve run it by all these years.
I really hope this holds.
And I recently wrote about making my own Anki Japanese cards in my blog[1]
[1] https://alt-romes.github.io/posts/2026-01-30-from-side-proje...
Ftr Danish is a category 1 language, while Japanese is category 4 ("https://2009-2017.state.gov/m/fsi/sls/orgoverview/languages")
You caught me. ;-)
(AnkiDroid has always been run independently, which is good, considering the state of the iOS client, which has always been neglected.)
> We’re currently talking to David Allison, a long-time core contributor to AnkiDroid, about working together on exactly these questions. His experience with AnkiDroid’s collaborative development is invaluable, and we’re grateful he’s willing to help us get this right. We’re incredibly excited to have him join us full-time to help propel Anki into the future.
Not sure whether you'd still be entitled to the source code under the previous license then.. can a copyright owner revoke a previously issued license to the code? Haven't heard of it, but wouldn't surprise me if it's legal.
You can't revoke previously issued licenses (unless the license allows it, which it doesn't in this case).
and then now why, of all times, when a solo developer is never more productive, would the lead maintainer cede ownership? the antidote for programming burnout has just been invented, just take it haha
Writing code is fun. Solving interesting problems is fun.
Debugging deep problems is fun.
Debugging slop code is a painful suffering experience, having to constantly double check that the AI agent didn't just change the unit tests to "return true" and lie to you is tiring, and the feeling that you can't significantly improve the tool burns me out hard.
That last one can't be overstated. When I find a weird behavior that looks like a bug in the linux kernel or rustc or such, I find it exhilarating to read code and understand what the bug is, how it got there, and to feel like I can fix it and never see it again.
When claude code gives me a "wrong" output for my prompt, I don't feel like there's any possible way I can go and find what part of the Opus 4.5 model resulted in it not being able to give better output.
I feel helpless to debug what went wrong when claude code spirals into the deep end.
I can add more initial context, add skills, but those are tiny heuristic tweaks around the giant mass of incomprehensible weights and biases that no human understands.
The antidote for programming burnout is not to replace all the fun parts of programming with painful probabilistic suffering.
To amend, I got way more than 30€ of value out of it. I learn a new language, and vocab training with Anki works better than anything else.
For example, copy and paste retains the text color (probably by design). So, sometimes I get black text on a black background, when the app is in dark mode.
The editing process to remove the formatting is pretty annoying.
It takes me time to find the edit button, which is buried in the menu but prominent in the desktop version. Then, I have to toggle the HTML mode and delete the retained tags, which on a phone takes time. The desktop version, instead, has a button to remove all formatting.
> I keep pressing the second button to OK a card, I rarely use the 3rd and 4th. But if I fail a card, that button becomes a NOK, and I keep pressing it out of reflex
> I can't help interpreting "card was a leech" notification other than "how dumb can you be". Fortunately there is no way to turn it off.
> It keeps phoning home for some reason, each time it gets into the foreground. It is really great when you are behind a proxy, and it keeps complaining that there is no network, every single time. Of course that call can't be turned off. Also, have no idea what it sends home. I try to trust that it's not some nefarious.
> Some years ago, for some reason Anki changed DB format, in a backwards incompatible way. There was a notification at the start of the app, that if I don't want it, I shouldn't update my app. I did turn off auto-update. A few weeks later it bricked my deck (my deck got updated to the new format, even though that old iOS app was the only way I accessed it), also trashing my 3 years long strike.I don't mind so much that it's paid, given how much use I get for the price, but it sucks knowing it sucks and not being able to help make it better.
(I think the data model underneath Anki is...showing its age (and lack of explicit design) and building that on top of it would not be too easy. I've thought about it a few times.)
It improved my grades so much in college that I spent the 25 bucks as a broke student so I could have it on my second hand iPad. This was before AnkiDroid even existed so it's amazing the price is still the same.
Anki app has an interface for adding/editing cards, and can absolutely be used without AnkiWeb or syncing. In fact, this is how I used it myself for years. I would argue that using AnkiWeb and syncing is an advanced feature for people who got the taste of having own decks and don't want to loose it.
I think this "problem" is somewhat made up
This fictional person you describe also has access to the web and Anki on there.
Anki is not a very useful tool if you are not making or editing your own decks or a teacher is for you. This is an incredibly painful experience on mobile.
I would agree with you, but he doesn't seem to mind
Very appealing, makes people try app immediately. :)
It seems a lot like saying nobody should use GMail unless they agree to pay for premium Google Services.
We've got a setting [developer options only] which completely reimplements them, it won't be live for 2.24, but hopefully 2.25
https://github.com/ericli3690/gsoc-ankidroid-report/blob/mai...
Later used Repetitions (iOS / Mac / web) for the steps, EM boards, and informatics boards most recently.
Only within the last year finally tried Anki -- and this time for language.
and the classic method was the inspiration for Anki to begin with: making your own flashcards on index cards! You could do a version of spaced repetition by shuffling the deck.
Not sure the digital version is actually easier or more effective
> Governance and decision-making: How decisions are made, who has final say, and how the community is heard
> Roadmap and priorities: What gets built when and how to balance competing needs
> The transition itself: How to bring in more support without disrupting what already works
In other words: they have no clue what to do next (https://forums.ankiweb.net/t/ankis-growing-up/68610/2#p-1905...)
Honestly, for a program like Anki, starting out by saying "we need to figure out what good governance looks like, as well as what might be agreeable and possible for everyone involved" is a much stronger positioning than coming up with something that may or may not fly to try make a strong first impression. Communities do not follow the conventional rules of American business.
I have no deep insight here, but given the times I have seen a negative reaction from the community with ownership changes or similar, giving yourself time to figure out how to do things may actually be a good thing.
Done right, commercial interests can often have a very positive symbiotic relationship with open source. Almost all the largest and best open source projects out there have substantial involvement from commercial interests.
I do think though that from a structural/governance perspective it's not a good idea for Anki to be owned by AnkiHub. Anki is a community project, not a corporate product, and while it sounds like the license will continue to reflect that, I personally think it would be best if the corporate structure did too. If Anki were spun out as an independent foundation (like Blender, Linux, etc) receiving most of its support and development work from AnkiHub, rather than owned outright, I think that would allow a much more robust governance structure than just having everything be under AnkiHub's direct control with some pinky promises about listening to the community.
It seems like the core things that Anki needs are new user experience improvements, and algorithm updates. SM2 really shows its age as compared to other algorithms.
If changing their SPS or the promises about an SPS motivate the learner, then great, they're putting in the work and time to learn, but I doubt that the effect of changing the SPS is as large as some people claim.
For example I used a tool that supposedly uses FSPS, but it did have a low maximum for the duration you don't have to practice a word, and no way for me to "ban" a word, so that it asks me in 6 months or something, and simple words kept coming back, especially after not learning for a few days. I didn't make much progress using the tool, even though it had FSPS.
That said, even SM-2 is probably vastly superior to just not doing SRS at all.
They state in contrast:
We do not dismiss the work behind FSRS. It is a commendable open-source effort and a marked improvement over ancient algorithms like SM-2.
For context, Anki uses SM-2's algorithm (albeit apparently heavily modified for various special cases) if FSRS is not enabled.
For me really learning a word means:
(1) Knowing how to say it.
(2) Knowing how to write it, meaning the Chinese characters, of course.
(3) Still remembering (1) and (2) after at least a month.
(4) Being able to actually use the word correctly.
You learn the words for a day (you're able to match the sounds and meaning to it). You will forget a lot of them tomorrow, so now you have to re-learn them. This is just how Anki works. You keep learning and re-learning until they stick for a prolonged period of time. It's common for Japanese learners to add between 20 or 30 words to their learning queue.
If you understand how Anki works, you will also understand how the word learning is used in relation to its flashcard mechanism.
And with all due respect, someone claiming they learn 20 words per day, in Mandarin, is an almost outrageous claim. If you think that "learning" is commonly agreed upon to mean "memorize for a couple of hours", then please show me the research into the meaning of the word, that proves your claim. While I have explicitly stated, what _my personal_ definition is, you are claiming to be knowledgeable about a "commonly agreed upon" definition. That is an impressive claim in itself. Let us all hear that definition, that is so commonly agreed upon, so that we can gain from that.
What you call learning, I call "training" or "practicing" or "revising". Now the onus is on you to prove to me, that indeed as you claim there is some commonly agreed upon definition, specifically in the area of learning Mandarin, that proves, that my definition is off.
And I will have you know, that I am learning Mandarin for some 10+ years, and have a lot of experience in that area. I know what counts and what is important.
Why do you keep harping on Mandarin in particular? Do you think it's harder than other languages to learn new words? It's not like you have to learn new hanzi for every word. Most are compounds. It's like being surprised someone easily learned how to spell "lighthouse" because it's got a silent "gh" and a silent "e" and the "ou" is not pronounced the way you'd expect, and the "th" in the middle isn't pronounced like "th" should be.
The learner already knew how to spell "light" and "house" so it was effortless to learn "lighthouse."
My experience with Japanese is that you hit around 800 or so kanji and new vocab comes very easily. Even new kanji come extremely easily because they're all made up of the same parts ("radicals").
EdIT: One hour a day devoted to language study will yield 20 new vocab words a day that, over time, you'll have around 85% recall, which translates to over 6,000 new words per year (over 7,000 but then you adjust downward because of the 85% factor).
The issue is that people want to learn a language in five minutes a day, but they don't bat an eye at playing a video game an hour a day to be able to beat some level. I remember playing for hours to be able to get good at 1942 on the NES back in the early 90s.
The original claim was about 20 Mandarin words.
> The learner already knew how to spell "light" and "house" so it was effortless to learn "lighthouse."
This kind of comparison doesn't work properly for learning Chinese characters. Simply combining characters like that only works ~half of the time or less.
> EdIT: One hour a day devoted to language study will yield 20 new vocab words a day that, over time, you'll have around 85% recall, which translates to over 6,000 new words per year (over 7,000 but then you adjust downward because of the 85% factor).
Delusional for Mandarin, unless you have some kind of special brain putting you in some 0.001% of the population. Not even natives learn that many words in a year. That many characters they might know when reaching university, and then later forget many again. Most native adults don't know that many.
> The issue is that people want to learn a language in five minutes a day, but they don't bat an eye at playing a video game an hour a day to be able to beat some level. I remember playing for hours to be able to get good at 1942 on the NES back in the early 90s.
Well, at least on that we agree. If one doesn't put in the time and effort, then the results will reflect that.
When people in the language learning community say they "learn 20 words a day", they are referring to New Cards Added. It is a metric of input and initial encoding, not guaranteed permanent storage.
In Anki, "Learning" is literally a specific phase (the red cards). You introduce the card, you pass the initial threshold, and then the algorithm handles the retention over the subsequent weeks. You are conflating the process of learning (adding new information) with the result of mastery (long-term active recall).
Would you say a native English speaker doesn't know the word "they're" if they keep spelling it "their" even if they use it correctly 100% of the time?
How does this opinion hold up if you consider that spelling wasn't standardized three centuries ago. Did no one know any English words until spelling got standardized in the 1800s? Do illiterate native speakers not truly know any words? Do children not know words?
How would your opinion change if you knew that plenty of native Japanese and Chinese speakers can't write the characters they can read anymore? If you don't have to physically write anymore, you lose the ability to write the characters. This is true of even educated adults in Japan and China. When I was a university student (I'm not Japanese), I could write kanji that my 30+yo Japanese friends had forgotten, but no one would say I knew how to use the words better than they could.
EDIT: And in any case, 10 new vocabulary words per day is extremely easy. In my experience having studied two foreign languages at the university level, that's pretty much the bare minimum expected to get an B in class.
Bad example, but to roll with it: In that case I would say they don't know it properly, since it is apparent, that in their mind there is no difference between "their" and "they're" or even "they are".
> How does this opinion hold up if you consider that spelling wasn't standardized three centuries ago. Did no one know any English words until spelling got standardized in the 1800s? Do illiterate native speakers not truly know any words? Do children not know words?
I am basing my personal definition of when I consider a word "learned" on reality, not on some "what if". If I had to map that idea of no spelling standardization to Chinese characters, then it would mean, that characters don't have standardized lines/components/parts. If there was no standard, then I guess I would consider this kind of making up how to write it on the fly sufficient for having learned a word. Thankfully there is standardization, so that is not a reality we live in.
Since I strive to not be an illiterate, I do not count being illiterate as having learned a word.
> EDIT: And in any case, 10 new vocabulary words per day is extremely easy. In my experience having studied two foreign languages at the university level, that's pretty much the bare minimum expected to get an B in class.
That depends very much on the language and course, but if it is your major, then sure, such a time commitment seems reasonable, since it is the thing you are doing. If your major is anything else and you just take an additional language course, where I come from you have once or twice a week that course. Then maybe 1 or 2 weeks you finish one chapter of a course book, which might have 20 new words, so that makes 20 words in 1-2 weeks, not in 2 days.
Typically for Mandarin the speed will also be slower than other easier to learn languages. For example at school I almost never had to learn vocabulary in English or Spanish. I just saw the words and memorized them somehow. Usage in class and often their sound and structure was sufficient , and always had good grades, often very good grades in those languages and always had good grades, often very good grades in those languages.
It doesn't work like that with Chinese characters. You are not gonna learn them (including writing them) by just looking at them a few times, unless you got an extraordinary visual, almost photographic memory. I consider myself already to have a pretty good visual memory, but still I need to put in the time and effort, and 20 words a day is way out of my league. But then again it was already cleared up in another comment, that the OP cuts out writing entirely. That's definitely a choice one can make and explains how 20 newly added words (I would still debate that that's "learned") make any sense.
I don't bother with the Hanzi past being able to recognize them. I want to be able to talk to people and, if I have to, use a pinyin keyboard to write basic sentences.
So only 1 & 4 are really relevant, 2 is what Anki is designed to do.
Writing a language makes you more skilled at living in the modern world. It's not a threshold past which you must travel to count as a speaker of that language.
By cutting out the memorization of Hanzi, I am able to accelerate my actual goal of having conversations with people.
In Silicon Valley speak, I think the term would be "ruthless prioritization" .
Unless you're applying for something in China, you don't need to know how to write hanzi ever, except for very one-off instances like "I can write happy new year in Chinese"
You know how many times I've written "real" Japanese by hand since 2005? Zero. I've written my name and stuff, sometimes I'll write 愛 to show my daughters. Nothing else. Because it's a worthless skill unless you live in Japan. Not even visiting. You live there.
Of course I type all the time. But typing is speaking + reading. It's not writing. You type phonetically (i.e., you know how to say the word), and then you hit spacebar until the correct kanji comes up (i.e., you can read kanji).
- Anki, as set up by dae aka Damien, is like the brand name and desktop implementation with the spaced repetition algorithm
- AnkiWeb is what I thought this hub thing was. It's where you download decks
- AnkiHub is a third party (started by "AnKing", now 35 employees) who sells decks as a monthly subscription and has their content on the deep web (you need to create an account and agree to terms to even see a listing of what's there besides a few featured parts). This is who is getting ownership of the former two. Because they write that Anki will remain open source at its "core", I presume that means that things will, at best, stay stable rather than anything (like AnkiWeb the deck sharing platform) becoming open
- AnkiDroid is a separate open source project (an Android app). The corporation is hiring the main developer, but it's not yet clear to me whether they're just going to get paid to work more on AnkiDroid or if they're also getting other tasks
----
To copy from my message on Discord:
> I’m moving to a full-time position working on Anki [incl. AnkiWeb & AnkiMobile]. I’m really excited about this, but there’s a mountain of pending, somewhat undefined work which will need to be done, and it’ll need my full-time attention for a while.
> I’ll still be contributing to AnkiDroid, but I won’t be able to commit as much time as I am doing currently (at least for the first few months while things stabilize). I’ll be here on evenings/weekends, and will be contributing in other ways (hopefully: unified Note Editor, JS addons etc… ), but I expect to slow down with code contributions to ensure I’m staying on on top of PR reviews & general force multiplier work. I’m definitely Org Admin’ing for GSoC over the summer [assuming Google gives us the greenlight], it’s historically been a VERY light role.
> In all honesty: I’m expecting things to be business as usual, I have more than enough capacity to keep up with the notification queue. Even if I completely dropped off the planet, we’re a great team and the improvements would keep on flowing. AnkiDroid’s bus factor has been >>> 1 for a LONG time now.
https://discord.gg/qjzcRTx => https://discord.com/channels/368267295601983490/701922522836...
Worth noting you don't need to use it. Anki comes with a syncserver implementation for a while now, and there are docker images too. It's worth it for the transfer speeds alone IMO.
Anki is under AGPL too, which has an anti-DRM clause, so many type of enshittification of anki or their addons (e.g. to prevent sharing of their decks) would be unenforceable too.
As such I see no obvious things that would be susceptible to enshittification here.
It's mostly due to time/resource/technical constraints [some of our strings come from a shared backend], but we can do better here, especially if there's now a lot more community interest in the feature.
Pull requests welcome! Do feel free to get in touch on the issue/Discord.
Overall this doesn't inspire much confidence in how solid and tested the procedure is.
https://github.com/ankitects/anki-manual/blob/main/src/sync-...
Full disclaimer - it's a feature which AnkiDroid supports, but isn't one which I use.
On that page though, the same issues are present. The pip install does not make use of any lock file.
pip install anki
SYNC_USER1=user:pass ~/syncserver/bin/python -m anki.syncserver
The repo already has a lock file for uv. It would be better to make use of that lock file, when using Python to install. And in fact, when one downloads a release of Anki for desktop and runs it the first time, it does make use of uv, creating a venv, and (unconfirmed) hopefully makes use of the uv lock file.
I see these kinds of issues very frequently in Python projects. As someone, who has previously worked on providing docker images for data science workflows, enabling reproducible research, I am quite sensitive to this. But also I hear from friends, that they are traumatized by Python projects installing things in system python and other shenanigans. In general there seem to be tons of people doing Python projects, who don't have a clear idea of how to make things safe and reproducible, which is giving Python projects in general a bad reputation. All while good solutions to these problems exist and existed for years.
It's 2026. Even JavaScript can do this.
pip is the de facto manager for the entire language. It should be better. With Node Package Manager for JS, the installation default is at the project level. You have to do a command line override to install globally.
PIP is the opposite. In fact, the only way to install at the project level is to create a virtual environment and trick PIP into thinking it's installing at the global level!
What language operates like this in 2026? Maven installs at the project level. Unison at the project level. Haskell at the project level. JS/TS at the project level.
The ecosystem is currently such that it seems hard to enshittify it. They say they have no intention of doing that and I believe it, but their vision of a healthy and good product might involve a fair price (for rich countries at least) whereas it was always free so far
Time will tell; it sounds like there's currently no plans either way, but it's also simply open enough that users can always just install the open source software and share decks with each other by whatever file transfer/sharing means. Everything that's already there won't simply go away. I'm going to keep using AnkiDroid and building the language deck I am working on
I also can’t imagine making cards on a phone, given how much switching between apps/windows is involved and how poor mobile platforms are at multitasking. It’s difficult to envision it being anything but maddening.
1. Anki isn't your everyday application with your everyday audience.
2. The number of people willing to splash $25 on an iOS flashcard app without first having tried it for free elsewhere, is incredibly small.
Where are you getting the stats that drive this claim? How are you measuring usage on platforms that don't necessarily collect usage metrics, e.g. desktop versions?
But upon reading this I think it's high time I exported all my notes in simple text format, just in case.
Maybe also try Fernando Borretti's flashcard app I saw (and dismissed) recently here
Already caveating with the "core" code. Even without PE and VC, it's clear that a company with 35 employees is bound to take this in a different direction than 1 guy, and not a good one. If there comes a day where those 35 employees can't be sustained anymore by revenue, and the choice is between enshittification and shutting down/firing everyone, we'll see what happens. That's the big difference - such a decision was never on the cards, or at least much less likely, when run by a single person. Now it will be.
Big conflict of interests too. AnkiHub makes money from selling paid addons. No chance any of those will ever end up in Anki now.
Also not a good look that they immediately locked the thread in their most popular community.