• New landing page

• XP + Leaderboard

• DIY cards improved (with a very helpful and unique feature for Language learners)

• Import Anki

• Import Duolingo

• Import Source Material

• Card Database

• Deck Statistics

• Cram Mode (useful for exam prep and/or comprehensible input etc)

• Diagnostic

• Shaeda vs Others

• Pricing

• For early-access to Shaeda, please email me (shaedacode@gmail.com) or message below.

XP & Leaderboards

Whilst I’m generally not a fan of XP for most EdTech apps, after a few requests, I’ve conceded that for flashcards (which can often be perceived as ‘difficult’ or ‘boring’ [which I disagree with]) they are probably a smart addition. One of the (many) issues with, say, Anki for example, is that it’s quite a dull and uninspired experience overall. Having a per-card XP feature and a per-language / per-topic Leaderboard alongside should improve the experience without being overly gamified or dopamine-heavy. Generally, so long as gamification doesn’t disrupt the actual learning process, it is likely fine - the issue is that with most apps, it commonly does.

DIY Cards

For Academic learners: simply type the front and back.

For Language learners: simply type exactly the words or phrases you wish to practice and have the translation and audio generated at the same time. The reason this is helpful is that the language you’re learning is probably not the language of your keyboard. But now you do not need to go to other sites or apps to generate the word/sentence, you simply type it in your native (keyboard’s) language and the app will convert it into your target language.

I don’t know of any other flashcard app that allows for this feature. Even if possible (such as on Anki) there are many more steps and it can be quite technical involving third-party tools etc. On Shaeda you just type, click Generate and save.

Import Anki (+ generate related)

You can now import your Anki deck (with audio, for Language learners) and have the option to ‘Generate Similar’.

For Academic cards imported, this will either generated related questions, pre-requisite questions, or the same questions but just reworded. This last point is notable as failing to correctly answer the same card that’s differently worded can indicate that the user simply recognised the card rather than knew the content.

For Language cards imported, you have the option to generate similar words or similar sentences - again, all with Chirp3HD audio.

Import Duolingo (+ generated related)

Import Source Material

Note: this is intentionally text-only. LLMs have a much lower accuracy when reading other formats, such as PDFs or screenshots. This is another problem with other flashcard apps in that this is never mentioned to the user. The result is that students are pasting tonnes of complex documents, PDFs and screenshots etc but the AI the app uses has almost certainly misread parts or even missed sections entirely (See here for a non-technical article)

Another issue with other apps is that they always use old and very cheap models, which is why they advertise “unlimited monthly flashcard generations” as though it’s a good thing for effective studying - it is not. Put simply, Anki aside, flashcard apps online are not good. I will show an example at the end of this post.

If you are wanting flashcards generated from a textbook, simply type the text in. Not only is this going to be substantially more accurate, it is also going to actually offer a small benefit anyway with reading and typing the material prior to having the cards generated.

Card Database

Whilst this isn’t quite on Anki’s level for incredibly detailed statistical analyses, it’s substantially better than other flashcard apps and far more intuitive. Many don’t provide such statistics at all yet somehow charge $20p/m.

Deck Statistics

Cram Mode

Cramming can be incredibly beneficial when done correctly. For those studying Academic, this typically means daily cramming a week or so before the exam. This is the one time where it absolutely makes sense to disregard the importance of the “spacing factor” (waiting as long as possible until reviewing) and get in as many reps as possible. Outside of this specific area, the efficiency of cramming is going to be substantially reduced.

Cramming for Language learning can be done, and is essentially just a way to get level-appropriate and relevant-to-you Comprehensible Input (CI). There are pro’s and con’s to using Shaeda for CI versus consuming native films or videos.

Diagnostic (90% finished)

For both Language and/or Academic, you have the option to take a ~20 minute ‘test’ to assess your initial level and repeat every ~2 months to check and assess how your knowledge level is growing. Here is mine for Russian, and Nutrition Science.

Shaeda vs Others

Going back to what I mentioned earlier re other apps not disclosing which LLM they’re using (because they use very old and cheap) and the fact that they advertise “unlimited monthly flashcard generations” (because they use very old and cheap) as though it’s beneficial (it’s definitely not), let’s see below how Shaeda compares ($3p/m + credits) against the current most expensive app app ($20p/m) that’s been around for years and has 10s of millions of users. The screenshots below are the result from having cards generated using the exact same prompt: “Graduate-level atomic physics”

Pricing

Shaeda will be $3.50 p/m membership + credits on release. This is the most sensible pricing structure and also the fairest. The reason other apps utilise a much higher monthly price with “unlimited” cards is solely to maximise as much profit as possible with no regard for actual usage or value. Users have no choice because there are no alternatives (currently). With Shaeda, you get maximum value and transparency. Flashcard apps (Anki aside) have been borderline mocking users for a very long time. Hopefully Shaeda can put an incredibly small dent in that by providing a far more productive and effective experience for 2-5x less.

The reason it is the fairest is that, for some, $1 in credit costs could last an entire year or more. For these users, paying a high monthly fee (>$10p/m) is illogical. For others, perhaps studying MS-level Academic, it is of no use having the allure of “unlimited” flashcards when they actually need fewer but more detailed, technical and accurate cards. For these students, paying a higher monthly fee is illogical.

The fix is simple: a substantially lower membership fee plus (less-profitable) credits.

That’s everything for this update. If you know of anyone who may be interested in learning much more efficiently (and cheaper), feel free to consider sharing this update with them!

PS: Please consider leaving a like or sharing. 50% of users who have signed up for Shaeda have all posts go to their spam - this isn’t ideal. Sharing or liking, Substack say, will help emails be directed to inboxes instead. Thank you!

Share

Subscribe now

TLDR:

Shaeda is open again for Early Access. Please message below if interested.

• New design

• DIY self-made card feature added (ft. cat)

• Ask Gemini added

• Improved Flashcard Breakdown

• Improved Settings/Sidebar

Early Access

Shaeda is now ready for use and is good to go for daily use as a study tool (whether that’s for Language learning or Academic study)

If you would like early access to Shaeda, or perhaps to test the accuracy of one of the as-yet-unvalidated languages (must be native/native-level speaker), please get in touch by messaging me below - my only ask that you do provide feedback if possible.

New UI Design

Previously the Ask Gemini was somewhat sandwiched and cramped at the bottom. It did not make sense to have a chat component horizontal as there was minimal room to scroll which made reading the chat needlessly difficult. This, as seen below, has now been moved to the righthand side which makes for a much improved layout. To be honest, I’m not sure what I was thinking not having it here to begin with.

The Settings and Timer areas have been moved to a sidebar which will appear whenever the cursor hits the far righthand side of the page. There is no click required, so if you wish to change the voice, speed or background noise mid-review, it is just as simple as before.

DIY Self-Made Cards

I agree that for certain situations, DIY cards will undoubtedly make more sense than Gemini-generated ones. The issue, as always, is the trade-off. More formally this is known as Opportunity Cost. If self-made cards improve retention by, say, 20% per-card, but take 5x the effort to create (and thus the total volume of practice is substantially lower), it might not be suitable for some - especially perhaps adults who are short(er) on time.

For any language learners using Shaeda for targeted, appropriate and relevant listening practice (commonly referred to as Comprehensible Input [CI]), it is unclear to me how DIY cards are a net positive tradeoff on the whole long-term: Whether the user writes “El gato callejero es tan chiquitito.” (which means the stray cat is so tiny) or Gemini writes “El gato callejero es tan chiquitito.”, those words can and will only ever be those words. Whether the user writes them or Gemini generates them, the words are still just the words: gato will still mean ‘cat’, callejero will still mean ‘stray’, and chiquitito will still mean ‘very small/tiny’.

DIY cards, however, will come into play for much more nuanced Academic topics or perhaps slang variants of words/sentences that one might have heard from natives and/or native material.

The key to remember is they likely will come with a higher per-card retention, but whether this is a worthy trade-off is not necessarily a given for everyone - context is always important.

Ask Gemini Added

The Ask Gemini feature has been added, along with a choice of models depending on the situation.

As seen below, there will also be some potential questions that appear. Usually elsewhere outside of Shaeda these suggested questions I maybe use only 25% of the time: I had a very quick question and, once answered, I did not need to know more. With Shaeda this will be higher as the app is aimed at understanding, and not necessarily just pure memorisation (which is still, in my opinion, a hugely important aspect to learning - however controversial that may be to some)

Improved Breakdown

Previously, the breakdown feature only analysed the specific inflected form of the word used on the card. This could be confusing, or just unclear, if you weren't already familiar with the word’s dictionary form. Now, the app automatically identifies and displays the base/lemma form (e.g., the infinitive for verbs or nominative for nouns), hopefully providing a clearer foundation.

Further, it is now easier to save individual words from the Breakdown box with the Settings button to configure which deck you would like the word being saved to, or Quick Add if you are saving multiple all to the same deck.

Sidebar

The previous design was too overwhelming and cramped the key features (Breakdown and Ask). Keeping the Settings and Timer as very (very) easy access is still important though, so there are no extra clicks or navigation required to access them. Just move your cursor all the way to the righthand side and the side panel loads instantly. The actual layout and readability of the Settings section has also been improved with it having more (vertical) space and no longer cramped inside a smaller container.

Negative Feedback, Scaffolding and Gamification

So far the only negative feedback I have received is that the app has no scaffolding or gamification:

Re scaffolding, the app does provide scaffolding with Breakdown, Ask, Paperclip, DeepDive and Lookup features - it is just on the user to utilise these. If you are asked a question that you kind of know the answer to, but not as much as you would like, it is wise to use all features available to you. This is what separates Shaeda from other flashcard apps, in my opinion. Unfamiliar with a key term? Use Breakdown to isolate it and add it to your deck. Ask Gemini what it means if unsure. If it is a complex or nuanced area, use Flash or Pro models. If you are still unsure, use LookUp which will perform a Google search and allow you to read articles on the topic. Then use Paperclip (or the Search bar) to test your new understanding when you return to the app. Remember that flashcards are just questions, and answering questions is both a form of demonstrating knowledge and acquiring knowledge due to the Testing Effect (Wiki, Articles)

Re gamification (or lack thereof), indeed the app has zero gamification and will continue to have zero gamification. For anyone looking for gamification I recommend downloading Duolingo and paying 5x more than Shaeda (‘SuperDuolingo’ is $13p/m) for Super Gamification.

Ultimately there is no free lunch in learning (or anywhere, for that matter). Learning (and retaining said learning) is inherently difficult and slow and there has to be some level of discomfort in order to progress. In this regard it is no different to one starting the gym: it would be inefficient and ineffective to go and never break a sweat.

400

PS: Shaeda recently crossed 400 signups - thank you all very much for the interest! However, unfortunately only around 50% of signups actually receive these email updates. Likes and Shares, therefore, are hugely appreciated as it improves the ‘health’ of the Substack (so they say), meaning more people who have subscribed will actually receive the emails in their inbox as opposed to their spam.

Subscribe now

Share

If you are interested in testing - and providing feedback - please get in touch either via the Message button below or emailing me at shaedacode@gmail.com.

-- Again, note that the app is intended for desktop use and not mobile or iPad

I’ll post some new screenshots of the app below

PS: If you know anyone who may be interested in using (or testing) Shaeda, please consider sharing with them.

Thank you!

Share

Subscribe now

• As the release of Shaeda nears, I wanted to write a quick, laid-back (opinion-based) post addressing some arguments against flashcards that I’ve seen over the years.

• I think arguments against flashcards are either not well thought out, or simply misguided due to a potential lack of awareness

• Shaeda update coming this week. It is 90% ready for pre-beta. If you’d like to be a part of stage 2 testing, please get in touch below:

“They promote memorisation, not true learning”

I am starting with this one. Not only because it is the most common but also due to the fact that there multiple problems with the statement - as I see it.

I’ll try to bullet-point some of my thoughts below:

1a: How exactly can new learning take place if prior learning has been forgotten? This risks a House of Cards-type scenario with one’s knowledge. It would be illogical to teach a child about multiplication and division prior to them having learnt addition and subtraction. The same applies just the same as one progresses.

1b: If a student demonstrating that they can recall xyz is not valid learning, then what exactly is and, importantly, how could they display it?

• “I know the answer. Trust me.”

1c: What exactly is ‘true’ learning?

• Typically this ‘just’ means “a deep understanding”.

• That is, proponents of this belief seem to suggest that learning x via flashcards is largely futile if you do not learn everything about x.

• 1. Why exactly flashcards cannot be used for deep understanding is not clear to me. If anything, the ability to repeatedly answer questions (many of which new) on a subject is the very definition of possessing a deep understanding. Simply rereading the textbook and saying “Yeah, looks about right” is not the same for reasons covered in previous posts (passive =/= active). Perhaps being able to implement something in the real world? Maybe, it depends on the subject. But even then, how will any given thing be implemented if the prerequisites have been forgotten? Regardless, the act of trying to implement something is still an indirect form of recall.

• But let’s just go with the statement as is purely for arguments sake:

• Deep understanding, as suggested by the name itself, is inherently incredibly difficult: it takes tens of thousands of hours of active work and practice.

• 2. To suggest that flashcards will not help with studying simply because someone is of the belief that they might not directly lead immediately to a deep understanding seems akin to suggesting that a young child learning their 7 times tables should stop as this will not make them the next Von Neumann or Ramanujan.

• What exactly does lead to a deep understanding that does not involve thousands of hours of further active study and practice? Nothing, of course.

• Therefore this is not a valid argument against flashcard study, it is a flawed argument against all forms of non-PhD-level study.

1d: What if the student does not require or even seek a deep understanding? (Aimed mostly at casual language learners and/or for students studying ‘side’ subjects)

• The belief that all learning requires extensive depth is misguided.

• Would it be nice to have deep expertise in everything one begins to learn? That’s debatable but sure, let’s go with it. (Un)fortunately our brains did not evolve for this - and luckily they didn’t have to.

• Take for example our language: English. We all speak English. We all read English. We all use English. But let’s say a (very bored, likely retired) linguistics professor started quizzing us on it one night. We would probably struggle to explain English1. Does this mean we don’t know the language? Such a suggestion of course would never be made. For some students, they are actually simply aiming for a certain grade in a particular subject or class in a years time to then be done with it to never touch again (probably Cellular Biology). Ironically, it is actually these very scenarios where flashcards possess even more utility: a decrease in interest and allocated study time necessitates an increase in the efficiency and effectiveness of the study.

• If someone is already naturally engaged with a subject 24/7, the utility of flashcards actually does decrease as the student is constantly cycling through the material in their brain performing continuous mini-recalls. This is why, for example, a native speaker would get virtually zero value from flashcards to learn their own language, but immense value if they were to study a new language or subject - and especially if early on in their journey. As the learner reaches higher levels (say, B1 and above), they’ll probably start to feel more comfortable engaging with both natives and native material more.

(To reiterate, it is not clear to me why exactly flashcards cannot be used for ‘deep understanding’ - I strongly believe that not only can they be, they’re arguably one of the most efficient means of doing so, if not the most efficient)

“They’re boring”

• 1. This is likely just a function the subject being studied. That’s fine, but it’s important to understand the difference.

  • If it is the subject, then as previously outlined, the student should still be using flashcards for that particular subject even more. We can simplify their choices down into the following 3 options:

    • Study effectively and efficiently, but for less time (flashcards)

    • Study ineffectively and inefficiently, but for more time to make up (reread, highlight notes)

    • Refuse to study (scroll TikTok)

• 2. If it isn’t the subject, and the student really does feel it is just the flashcard that are boring them, it’s likely that they’ve been drawn to certain (unnamed) heavily-gamified apps that have lulled them into a false sense of learning via constant easy dopamine hits.

• Indeed, such apps are incredibly entertaining and addictive. Of course we’ve touched on the risks here in previous posts about how this is misleading as there is minimal learning actually occurring: the student is just 99% gaming.

• Further, once this 99%-gaming ‘study’ session is over, they’ll typically then go and engage in 100% gaming (or TikTok’ing). Thus the total time spent learning or studying is actually substantially less despite spending considerably longer on the gamified apps.

• Given children are the main audience of these apps, they will simply not yet have developed the meta-cognitive awareness to recognise that they aren’t actually learning anything.

• Worse still, the parents see that their child is ‘studying’ on said apps, likely collecting substantial points. They’re therefore happy to continue paying for the membership for their child’s ‘education’. And they’re then happy for their child to go and relax/game once their ‘study session’ is over. The problem becomes apparent.

“They’re too difficult”

Well, being asked questions repeatedly certainly is difficult. But:

• This is a feature, not a bug. The fatigue experienced during flashcards is a useful proxy to indicate that you’re brain is beginning the slow process of learning. To avoid this cognitive challenge is no different than going to the gym and avoiding the physical challenge. The fatigue is a large part of what drives the adaptation.

• Too difficult is a function of the specific card itself, and not of flashcards as a whole. Flashcards are very commonly formatted poorly (too long, not atomic, irrelevant, too advanced etc) which will needlessly lead to poor(er) retention.

• If you’re brand new to learning a new language, and, say, the 4th card of the deck you’ve downloaded online prompts you to translate “He walked to the shop yesterday”, of course this will be too difficult at this early stage. First you need to learn and remember how to say all of “he”, then “walked”, so on and so forth. With Shaeda this won’t happen. You customise exactly what you want to learn, and if a card appears that you don’t want (for whatever reason), you just skip.

Learning something new without undergoing discomfort is just not possible. Are some people naturally good at certain things? Sure, perhaps2. But that does not mean all things come naturally to them - even within their domain.

A young math student spends hours on long division

A senior student spends hours factoring polynomials.

A PhD student spends hours solving equations that are as long as their arm.

But the degree of effort and difficulty involved is relative and consistent throughout.

“Physical cards are better as you need to write them”

• Outside of very specific cases, it is illogical to opt for handwritten cards in this day and age.

• Advances in algorithmic scheduling far (far) outweigh the potentially slightly increased retention of card-specific hand-written cards.

• Another thing to bear in mind with “handwritten cards leads to better retention” is that this comes with much, much slower creation. That is, I don’t know if the slightly increased retention is a function of the act of hand-writing, or just a function of taking longer to write it.

• Further, if a handwritten physical deck comes with, say, 10% better retention but with only 10% of the total cards you’d like to learn, the tradeoff is clearly not worth it: the student has missed the forest for the trees. Remember that New Cards should be gradually added over time whilst maintaining Review of Due Cards. Trying to balance this, keep track of each day’s Reviews and make time for New Cards becomes unfeasible with even the most basic of scheduling intervals. This means that even if the student does make time (and keeps topping up their paper and pen reserves), they’ll quickly actually lose that time and more with the inefficient intervals.

• With Shaeda, you have FSRS built-in to determine the best time to review the card, as well as additional unique, one-click features such as Transliterate/Breakdown/Look Up/Deep Dive/Ask/Chirp3 TTS/Audio speed/Toggle Text visibility/Autoloop etc. None of these currently exist within flashcard apps that I’m aware of. They do loosely exist as individual add-ons, but this is again even more friction.

“They don’t work for certain subjects”

• Every time I’ve encountered this, the actual subjects differ. This to me suggests that the issue isn’t the tool (flashcards), but rather the implementation (how the cards are made and used) and/or a misunderstanding of the subject’s learning requirements.

• However, subjects that do sometimes pop up are Maths and CS.

• But I think this is just a misunderstanding of what flashcards are: questions.

• To suggest, then, that Maths and CS are not subjects that can be studied by answering questions would of course be quite strange.

• On the Maths side, you just choose an appropriate level and prompt it accordingly. Kyle Kabasares actually has some interesting videos on this where he tests the ability of new models’ to answer incredibly complex (probably bleeding-edge) PhD-level Physics questions. Would I recommend LLMs for this level of learning? Probably not. But most resources online are probably not going to provide sufficiently accurate material at this level.

• And on the CS side, for some examples of what it can look like when CS students either do not think foundational knowledge is important, or simply no longer possess the knowledge due to forgetting, consider watching this video (which is just one of hundreds on this channel - here’s another). Importantly, these students are not outliers. Does this mean that students who do use flashcards know everything? No, of course not3. In fact, quite the opposite: they’ll actually be hyper-aware that they most definitely do not know everything and will then be able to take the necessary steps to correct that. Students who study only via passive means will only make this realisation when it’s slightly too late. Again, flashcards are just questions.

• To any 3rd/4th year college/university Math/CS students here, consider if you can answer the following:

“Reviews pile up too fast, causing stress when thinking about each day’s reviews”

• This is a function of how, not what.

• If the flashcards are poorly formatted, retention will be low, frustration will be high and Daily Due Cards will climb excessively high.

• If the flashcards are formatted correctly, but the level is too high or they’re not relevant, retention will be low and Daily Due will climb.

• If the flashcards are formatted correctly, relevant, and the level is correct, but the amount of New Cards added each day is too high, the Daily Due will climb.

Remember that the (recommended) daily nature of flashcards is merely a reflection of our brains - both on the habit-forming side but also on the minimise forgetting side4 . Cards appear daily because we forget things daily.

Flashcards are daily, highly-relevant, well-formatted questions that leverage the proven two most effective learning techniques with modern algorithms determining when the optimal time for review is to extract maximum retention.

It’s (cognitive) science.

That’s everything - thank you for reading.

Please consider leaving a like or sharing. 50% of users who have signed up for Shaeda updates have all posts go to their spam - this isn’t ideal. Sharing or liking, Substack say, will help emails be directed to inboxes instead.

Share

Subscribe now

PS: Shaeda update is coming in the next few days. I have added some more screenshots to the landing page under ‘Latest’ for a look at these. As always, if you would like to be a pre-beta tester, simply get in touch.

1

A quiz on the English language. Do we need to know these in order to speak English?

Answers

• First group of testers are getting access to pre-beta very soon, likely this week

• If you’d like to be part of 2.0 pre-beta testing, please let me know.

• Speed improvements added

• Chirp3HD TTS added

• Model Selection added

• Voting Section added

• Basic Heatmap added

• Academic Subjects have undergone some more validation (and will continue to do so)

• Save To Database feature is very close

• I’m in need of some more language teachers. See bottom for more.

v1 Pre-Beta Testing

I’ve been in touch with 10 of you who will be first testers of pre-beta. Exciting times. Over a year of work.

Now, due to the amount of signups and emails I get enquiring about release, I am going to stagger the releases slightly. I don’t like putting timeframes on these as a rushed product is generally a bad one, but it’s looking like early December time for when full beta will be out for everyone. Of course, if you cannot wait this long and are desperate to try, do get in touch (if this is you, sharing this page or the app is very much appreciated). Of course, expect bugs if you are trying during pre-beta.

Yes, pre-beta will be free for full access.

Full beta will have a free tier*, but the initial pricing of $3p/m will likely be in place by then. This will increase over time, with early members maintain their pricing.

*Free tier = full access to A0 (see below for the new design) and 5 free flashcard generations per day using Flash-Lite only.

Speed Improvements

Certain actions on the site used to take a second or two, but we’ve managed to get those cut down quite a bit without increasing how ‘heavy’ the app is to run. As each version of pre-beta is released and feedback comes in, these actions should become even faster as we work out where bottlenecks are and optimise further etc. You might even notice the landing page loading much faster now. Again though, it’s just me and a developer, not Microsoft.

Chrip3HD TTS

A very popular use-case of Shaeda will be for language learning. More specifically, for improving listening comprehension. A lot of work has been going into this. Put simply, Google’s Chirp3HD is the best TTS out there. You can have a listen to various languages here (note: this link loads incredibly slowly and depending on your device and internet, may not load at all). Some languages are not supported, and for these a fallback to Microsoft’s Azure TTS will be added.

An important note here is that you may experience a delay of 1-2 seconds the first time you see a brand new card after generating. If it is a card you find valuable and you want to save (which should be almost all of your cards), once you save it to your database and login to review each day etc, it will be near-instant on playback. The majority of user’s time on Shaeda will likely be spent reviewing already-seen cards. I’ve touched on this before already but this is by design and is another intentional feature not a bug: A common complaint against non-self-made flashcards online is that they have a lower retention (users forget them easier). However, this is a function of 2 things:

• 1. How those cards were added to the users database (all at once without ever seeing), and

• 2. Having no control over what is on the card (it could be a single word, it could be a complex question, it could by a short sentence, it could be completely irrelevant to your interests etc).

  
  In Shaeda, it is of course different:

• 1. You will see each card at least once if you want to add it to your database. This serves as a ‘primer’ for when you then revisit it in Review, meaning it has not spawned out of nowhere, and

• 2. You have complete control over what you see. If you’re learning, say, Thai, and want to focus on learning Muay Thai related words, with Shaeda you can. If you then feel confident on your progress of single words and want to try on current-tense Muay Thai-related short sentences, you can. You have control. You can listen to whichever voice you want. You can turn the speed right down. You can even add 'Background Gym sounds’ to improve real-world listening ability. Once you’ve became comfortable with listening, you can click ‘Flip’ and now you can practice speaking those very same cards. With other options online, you click ~‘Download’ and you’re left studying whatever is on the deck. It’s very hard to modify, add new, add related, flip etc.

  If any of this isn’t clear, feel free to get in touch. Essentially, a lot of thinking and work has gone into common areas of ‘flashcard friction’. The result is that Shaeda will make for far more productive and enjoyable studying. No more “Oh no, this annoying card again…”.

Model Selection + Further Testing

The above (Chirp3HD) will not be relevant if you are not planning on using Shaeda for language learning (but you should, it’s great), but the ability to select which model you want to use most certainly is. I’ve provided some guidance on the app for which model you might want to use (of course you can switch back and forth).

Very generally speaking, if you’re planning on using Shaeda for very difficult content (I.e., areas such as: MS level content in Quantum Computing; C2+ long-sentence “Political Negotiations” in Mandarin/Japanese; Subjects that are not officially listed as validated etc), then you may want to consider using a ‘Smarter-But-Slower’ model.

Again, for the vast majority this is likely not necessary. A lot of testing has gone into Shaeda. Google’s Gemini models along with the behind-the-scenes work has made Shaeda great - testers’ words not mine.

I’m sometimes asked what kind of testing the languages/subjects go through. As an example, Physics has already been validated by a professor and a current PhD student from Japan approaching the end of his degree. Once Google release Gemini 3, which is anytime in November or December, this will be yet another upgrade on the model’s ability and again Shaeda will undergo even more validation for even higher levels of study. If you have concerns over LLM-generated content, you can relax when using Shaeda.

In cases where the topic or language has not been validated as much, a message will be shown to advise you on how you should use.

For all languages validated, the model used was Flash-Lite and all native-testers and teachers have responded that it is incredibly accurate at ~99-100% even for advanced content (I’ve paid for human-made content that wasn’t even that accurate at 10x the cost). I cannot speak for other apps as I’m not sure what validation they have undergone nor what models they are using.

Voting Feature

I’ve added an option for users to vote for which features you would like to see added first. It’s not an exhaustive list, so the option is there to email in if you would prefer something that is currently not added.

However, given that it is still another 1-2 months until full beta, I’ll provide a poll below if anyone would like to start voting now.

Basic Heatmap

It works. It’s not the most glamorous currently, and I had to simplify it a lot to focus on higher priority work, but it’s there.

I Need Teachers

If anyone knows of teachers who are familiar with flashcards and are able to use Google Docs for the following languages or subjects, please email me at shaedacode at gmail dot com

That’s everything, I think.

I’m sorry that the last update was so long ago. Hopefully you haven’t forgotten about the app. As you can see, work has very much not stopped.

Thank you all very much.

If you’re reading this and are excited for Shaeda to help you with your studying, please do consider sharing with others you know who may be interested in the app. Only around 60% of people who have signed up are actually receiving these updates in their email. We’re nearly at 300 signups. I was expecting maybe 20 when I first created the Waiting List, of which I would be have been one. I don’t plan on spending money on advertising, I’d rather invest into building and improving the app further, so any sharing with others that may find value as well is hugely valuable and goes directly to making Shaeda even better. I see a few have already done this, so thank you very much.

Subscribe now

Share shaeda.io

Thank you all so much for your interest and emails.

I really do hope you all love Shaeda when it’s ready, and I’m very confident the app will change how studying is done.

TLDR Study 1:

• 90 final-year undergraduate medical students were randomly split into two groups over 6 months: one group will use Flashcards, and the other will carry on with ‘Traditional Study’

• Both groups began the study with the same knowledge levels (potentially slightly higher in the Traditional Study group)

• After 6 months, the Flashcard students test performance improved significantly, while the traditional group’s did not improve virtually at all.

• Over 90% of students in the Flashcard group reported better retention, engagement, and confidence, with ~65% saying that they felt so “very strongly”.

TLDR Study 2:

• This was an observational study (researchers watched from afar) looking at 36 medical students who were already using flashcards

• Mature cards (cards that have a long gap [interval] due to repeated success in consistently recalling) were the strongest predictor of exam performance (knowledge)

• High flashcard users (users that had amassed ≥9,390 mature cards) scored 71.5%

• Low users (<9,390 mature cards) scored 60.0%

• ~10 percentage point difference between high and low users

• Overall, students averaged ~179 hours on flashcards over the study period (6 weeks)

Shaeda Update:

• Work is still ongoing to make it as good as possible (Again, within reason. There is no large full-time PhD team working on it, just me and a Web Developer)

• The app is not being vibe-coded with AI as I plan on maintaining and improving it for a long time. This means that things need to be set up properly in a way that will allow for smooth updates in the future. This means that it takes longer. And that means that more care will have gone into the app to make it as good as possible.

• I’m looking for around ~5-10 people to test an early early version (language learning, academic study, or both). If you are interested in this, please reply to this email and I will be back in touch very soon. All I ask is that you do give feedback, good or bad.

Substack Stuff:

• Sorry to be that guy, but Likes and Shares help immensely with getting these emails to all subscribers. Still only around ~55% of subscribers get these emails into their inbox. Substack tell me that the way to increase this is to get more interaction on posts.

• If you find this an interesting and valuable read, please consider liking and sharing with someone who you feel may also find value in it - and also in Shaeda.

Share

Subscribe now

• In the post below, I have highlighted key sections in different colours and occasionally provided some commentary below certain parts using the same colours. I believe this has improved readability.

• I’ve a slight background in Statistics, but I am not a professional. If you spot an error in any of my interpretations, please feel free to get in touch and I will happily update and credit.

A couple of months ago, we covered some research on what effective studying looks like (according to a lot of research) and whether or not flashcards are effective. You can take a look here.

Since then, more research has came out that has tried to explore this further.

This will be a quick blog post that goes into these new studies. It will follow a similar structure to the previous blog post in that I will take screenshots, highlight relevant sections, and commenting where I feel potentially relevant.

Both papers are open-access, and are linked and cited below.

Study 1:

Notes on Method

Notes on Results

Now let’s get into the bulk of the paper. Feel free to scroll past all this if you’ve seen enough and would like to see about study two.

Some Quick Thoughts on “challenges such as time constraints and inconsistent adherence”

Some Notes:

• The flashcards were custom made by senior staff, professors, or postgraduate students.

• This is why all subjects and languages on Shaeda have undergone validation and is part of what, I believe, will separate Shaeda (as an ‘AI-Flashcard’ application) from the rest in this competitive space.

  • (Although per the previous blog, a custom/DIY flashcard feature will also be added into Shaeda as well.

  • For info on what ‘validation’ looks like (or rather, what it results in), see below:

    

• As I’ve already touched on before, though, there is always a trade-off with everything in life. That is, yes, custom-made flashcards from 1:1 tutoring with a tutor will result in pitch-perfect cards, with perfect audio (for language learning), or perfect examples/explanations (for academic learning). Whether this translates, however, into better knowledge and retention is another question.

  • It will also, naturally, cost significantly more - along with being a much slower process of obtaining new cards. This is the trade-off. Unfortunately, there’s no such thing as a free lunch.

Study 2:

Quick Thoughts on Table 5

Other Potential Limitations

• No baseline knowledge assessment

  • Students may have entered with vastly different foundational knowledge, making it impossible to isolate Anki's effect from prior learning

• No control for baseline study skills:

  • Students with better existing study habits might naturally gravitate toward systematic tools like Anki

This blog ended up being longer than I planned - apologies!

Reminder of the Shaeda update: work is still ongoing, and I am looking for ~5-10 more people to test an early early version. If you are interested, please get in touch by replying to this email.

Another small reminder (sorry) that Likes and Shares are hugely appreciated as it improves the ‘health’ of the Substack, meaning more people who subscribed will actually receive the emails in their inbox - as opposed to their spam.

Thanks for reading.

Share

Subscribe now

Citations:

Vagha, K., Choudhari, S., Taksande, A., Tembhurne, J., Vagha, J., & Vagha, S. (2025). Implementation of a spaced-repetition approach to enhance undergraduate learning and engagement in paediatrics. Frontiers in medicine, 12, 1601614. https://doi.org/10.3389/fmed.2025.1601614

Winter, V., Ames, E., Jacobs, M., Georges, C., & Simanton, E. (2025). Exploring the Impact of Spaced Repetition Through Anki Usage on Preclinical Exam Performance. Journal of Medical Education and Curricular Development, 12. https://doi.org/10.1177/23821205251369705 (Original work published 2025)

User Login

Account Creation has now been setup and is ready to go. This took longer than planned. I apologise for the delay on this.

Save Card Feature

Now that the User Login/Account Creation has been set up, next we’ll be setting up the Save Card feature. This is quite self-explanatory: it’s the ability to Save Cards to your decks to then review over time. The whole point of flashcards is to enable Spaced Practice (and Retrieval Practice, simultaneously). Spaced Practice is only possible if you ‘Save Cards’ to a deck of yours, and then review said deck daily.

Optional Subtopics for Academic Study

Streak Counter

Upload Own Sources (Demo Only)

(With the caveat it may not necessarily be better [to use one’s own sources] or even required. See previous Update for more on this)

Deliberate Errors Research

A newly published study by Qiang, Ma, and Li (2025) indicates that correcting deliberate errors leads to better long-term learning than traditional restudy. This is likely due to the fact that this strategy:

• creates a ‘cognitive conflict’ that must be resolved (strengthening memory encoding)

• forces learners to explain why something is wrong and what makes the correct answer right (deepening understanding)

• and introduces desirable difficulty that makes learning feel harder in the moment but produces stronger retention over time

  • This is essentially leveraging our brain's error-detection and correction mechanisms to differentiate between similar concepts and solidify knowledge

  • That last part is very important: common methods of study (rereading, highlighting etc) feel great. Students will feel like they are learning. And to be clear, there likely will be some very small degree of learning occurring. It is just not nearly as effective as other methods, such as Retrieval Practice etc. Again, a really useful way to think about this may be if one has ever attempted a diet. It will feel difficult. That is part of it. If learning things (as in, really learning) was easy there’d be no need for School or University.

For more information on this study, consider reading this blog link here

Overall this paper was really cool for me to see. When I first had the idea of Check It there was nothing like it currently anywhere in any educational apps - but why? It makes complete sense to have something like this.

For any parents especially it may be particularly helpful in helping your child study (perhaps with homework etc) when trying to decipher whether your child knows the topic or just knows the card, CheckIt Mode will enable this. If they have been answering their flashcards correctly on the topic of, say, ‘Gravity’, but they failed to spot the error in a card which states “Gravity is the force that pushes objects up to the sky”, this is valuable information that can then be rectified.

I spoke on this in more detail in the previous Update re a potential risk of ‘Overfitted’ knowledge when using flashcards.

That’s everything - this was just a short one.

Thank you for reading!

Share

Subscribe now

• A new paper just published may give more evidence that flashcards are both (very) effective and under-utilised

• A 2021 meta-analysis appears to find that only 2 methods of studying are highly effective (all others are of low or moderate effectiveness)

• Majority of students seem to consistently use the lowest-effectiveness methods (which will likely only get worse with the rise of AI)

• Some new shaeda updates

• A very small request at the end

• Research seems to indicate that a large percentage of teachers during their training were not made aware of the research on effective studying and learning

Are Flashcards Effective?

A new study by Ingebrigtsen and colleagues testing the effectiveness of flashcards on student learning was just published the other day (Ingebrigtsen et al., 2025). The discussion segment, which begins at the bottom of page 10, is particularly insightful without needing to read the entire paper necessarily. I will paste the Abstract below.

The below is a very quick summary:

• Students were randomised into whether they were placed into the flashcard group or not

• The exam was a first-year university exam covering Anatomy, Physiology and Biochemistry with a historically high failure rate of ~20-25%.

• Flashcard users scored on average +7.6 points greater on the final exam (out of 100)

• Flashcard users were nearly 3× more likely to pass

• Flashcard users were over 2× more likely to get an A

• The relationship between flashcard use and greater exam performance was maintained after adjusting for potential confounders (age, effort, study duration etc)

• Potential limitations did exist, such as low adoption rate of flashcards.

I will paste some screenshots of the paper that are of most relevance below for those interested in further reading without wanting to be overwhelmed by the full paper.

Do Related Studies Find Similar Results?

The study above builds on previous papers which also found positive effects of spaced practice/flashcards/testing:

This paper concluded “A total of 87% of flashcard users found the flashcards to be helpful, and 83% of flashcard users would recommend the flashcards to someone else. Flashcard usage was not associated with final exam scores.” (Sun et al., 2021)

Some notes here:

• They were restricted from actually publishing the exam results

• Flashcards were only used for a short period of time, thus the full effects were not likely to have fully materialised yet

The authors of this paper write the following in an ‘Essentials’ segment (Deng, Gluckstein & Larsen., 2015):

This paper concluded with “Repeated testing with feedback appears to result in significantly greater long-term retention of information taught in a didactic conference than repeated, spaced study. Testing should be considered for its potential impact on learning and not only as an assessment device.” (Larsen, Butler & Roediger 3rd., 2009)

This paper concluded “Spaced education significantly improved composite end-of-year test scores…Spaced education consisting of clinical scenarios and questions distributed weekly via e-mail can significantly improve students' retention of medical knowledge.” (Kerfoot et al., 2007)

Remember, at a fundamental level all flashcards are doing is easily implementing Spaced Practice and Self-Recall (a.k.a. retrieving from memory with minimal (0) assistance)

For more on these 2 methods, I will paste some quotes from a meta-analysis performed in 2021 below.

2021 Meta-analysis on Effective Study Techniques

The below is directly quoted from this Meta-analysis where the authors are referring to a previous meta-analysis done in 2013.

This article outlines a meta-analysis of the 10 learning techniques identified in Dunlosky et al. (2013a), and is based on 242 studies, 1,619 effects, 169,179 unique participants, with an overall mean of 0.56.

The most effective techniques are Distributed Practice and Practice Testing and the least effective…are Underlining and Summarization

The authors of a previous meta-analysis categorised 10 techniques into three groups based on whether they considered them having high, medium or low support for their effectiveness in enhancing learning.

• High support ✅

  • Practice Testing

  • Distributed Practice

• Moderate support

  • Elaborative Interrogation (generating an explanation of a fact or concept)

  • Self-Explanation (where the student explains how new information is related to already-known information)

  • Interleaved Practice (implementing a schedule of practice mixing different kinds of problems within a single study session).

• Low support

  • Summarization (writing summaries of to-be-learned texts)

  • Highlighting/Underlining (marking potentially important portions of to-be-learned materials whilst reading)

  • Keyword Mnemonic (generating keywords and mental imagery to associate verbal materials)

  • Imagery use (attempting to form mental images of text materials while reading or listening)

  • Re-Reading (restudying text material again after an initial reading)

In an accompanying article, Dunlosky et al. (2013b) claimed that some of these low support techniques (that students use a lot) have “failed to help students of all sorts” (p. 20), the benefits can be short lived, they may not be widely applicable, the benefits are relatively limited, and they do not provide “bang for the buck” (p. 21).

Practice Testing is one of the two techniques with the highest utility. Practice Testing instead involves any activity where the student practices retrieval of to-be-learned information, reproduces that information in some form, and evaluates the correctness of that reproduction against an accepted ‘correct’ answer. Any discrepancy between the produced and “correct” information then forms a type of feedback that the learner uses to modify their understanding. Practice tests can include a range of activities that students can conduct on their own, such as completing questions from textbooks or previous exams, or even self-generated flashcards. According to Dunlosky et al. (2013a), such testing helps increase the likelihood that target information can be retrieved from long-term memory and it helps students mentally organize information that supports better retention and test performance. This effect is strong regardless of test form (multiple choice or essay), even when the format of the practice test does not match the format of the criterion test, and it is effective for all ages of student. Practice Testing works well even when it is massed, but is even more effective when it is spaced over time. It does not place high demand on time, is easy to learn to do (but some basic instruction on how to most effectively use practice tests helps), is so much better than unguided restudy, and so much more effective when there is feedback about the practice test outputs (which also enhances confidence in performance).

Many studies have shown that practice spread out over time (spaced) is much more effective than practice over a short time period (massed)–this is what is meant by Distributed Practice. Most students need three to four opportunities to learn something (Nuthall, 2007) but these learning opportunities are more effective if they are distributed over time, rather than delivered in one massed session: that is, spaced practice, not skill and drill, spread out not crammed, and longer inter-study intervals are more effective than shorter. There have been four meta-analyses of Spaced vs. Massed practices involving about 300 studies, with an average effect of 0.60 (Donovan and Radosevich, 1999; Cepeda et al., 2006; Janiszewski et al., 2003; Lee and Genovese 1988). Cepeda et al. (2008) showed that for almost all retention intervals, memory performance increases sharply with the length of the spacing interval.

Further, Spaced Practice is more effective for deeper than surface processing, and for all ages. Rowland (2014) completed a meta-analysis on 61 studies investigating the effect of testing vs. restudy on retention. He found a high effect size (d = 0.50) supporting the testing over restudy, and the effects were greater for recall than for recognition tasks. The educational message is to review previously covered material in subsequent units of work, time tests regularly and not all at the end (which encourages cramming and massed practice), and given that students tend to rate learning higher after massed, educate them as to the benefits of spaced practice and show them those benefits.

A follow-up and more teacher accessible article by Dunlosky et al. (2013b) asks why students do not learn about the best techniques for learning. Perhaps, the authors suggest, it is because curricula are developed to highlight content rather than how to effectively acquire it; and it may be because many recent textbooks used in teacher education courses fail to adequately cover the most effective techniques or how to teach students to use them. They noted that employing the best techniques will only be effective if students are motivated to use them correctly but teaching students to guide their learning of content using effective techniques will allow them to successfully learn throughout their lifetime.

The authors of this paper then essentially tested again this 2013 meta-analysis to calculate an individual effect size for each of the 10 learning techniques.

Method:

Research syntheses aim to summarise past research by estimating effect-sizes from multiple, separate studies that address, in this case, 10 major learning techniques. The data is based on the 399 studies referenced in Dunlosky et al. (2013a). We removed all non-empirical studies, and any studies that did not report sufficient data for the calculation of a Cohen’s d. This resulted in 242 studies being included in the meta-analysis, many of which contained data for multiple effect sizes, resulting in 1,620 cases for which a Cohen’s d was calculated (see Figure 1).

Results:

The below is a quick overview of this Table with slightly more detail:

In the first column we see the study technique.

In the second we see what the 2013 meta-analysis classified each as in terms of effectiveness.

The third, # Cases, represents the total number of effect sizes. Since a single study can report multiple results (e.g., for different groups or at different times), this number can be higher than the number of studies. We can see that there’s much more data for the top 2 techniques.

The fourth column, Unique N, is the total number of unique participants (subjects) across all studies for that technique. N is a standard abbreviation for number. Using a "Unique N" ensures individuals are not counted multiple times, giving us the true sample size. Generally, a larger N is better as it increases the statistical power and precision of the results.

The fifth, d (short for Cohen’s d), is arguably of most relevance and hence why I’ve highlighted it. This is the actual calculated effect size. It attempts to tell us how much of an effect did the study method have on knowledge retention. Below is loosely how to interpret:

• d of <0: Negative effect.

• d of 0: No effect.

• d of >0: Positive effect;

  • ~0-0.2 = Small positive effect

  • ~0.2-0.5 = Moderately positive effect

  • ~0.8 = ~High positive effect

The sixth column, SEM (Standard Error of the Mean), attempts to quantify the precision of the calculated mean effect size. It represents the estimated SD of the sample means if you were to re-run the study multiple times. It essentially serves as a measure of confidence in the result: the lower the SEM, the more precise the estimate and the more likely it is that our sample mean accurately reflects the true population mean.

The seventh column, Q, measures how much the results of the grouped studies varied from each other. I.e., how different their findings were. We can see that there was a lot of variation. For more on this would be beyond the scope of this email, but the authors address it in their paper with further analyses.

And lastly we have I² (I-squared) which is telling us what percentage of total variation in the results is due to the differences between studies rather than random chance. Again, more on this would be beyond the scope of this email but can be found by reading through the paper.

The below is now back to quoting from the paper:

The purpose of the current study was twofold: to provide empirical estimates of the effectiveness of the 10 learning techniques, and second, to empirically evaluate a range of their potential moderators.

The technique with the lowest overall effect was Summarization. Dunlosky et al. (2013a) note that it is difficult to draw general conclusions about its efficacy, it is likely a family of techniques, and should not be confused with mere copying. They noted that it is easy to learn and use, training typically helps improve the effect (but such training may need to be extensive), but suggest other techniques might better serve the students.

In their other article (Dunlosky et al., 2013b), the authors classified Summarization as among the “less useful techniques” that “have not fared so well when considered with an eye toward effectiveness” (p. 19).

What is now most startling is that it appears quite consistently that the least effective techniques are actually the most used by students when asked.

Urrizola, Santiago & Arbea in their 2023 paper titled “Learning techniques that medical students use for long-term retention: A cross-sectional analysis” write the following:

Remember that the authors of the first study on flashcard effectiveness also cited other studies and wrote similar:

Further Reading

For more reading on this for those interested, I can strongly recommend the following books which are all both recent and based on research.

Math Academy Way

• (note that Math Academy’s book is good, but for the majority of learners I do not recommend their platform currently.)

How Learning Happens

How We Learn

That is everything! Hopefully this may be helpful and reassuring to many who are confused about flashcards and/or wondering what the research finds regarding their effectiveness.

Next week I plan on writing about some common complaints and/or worries over using flashcards (friction in finding appropriate cards, confusion in setting up, concerns that flashcards are only usable for random fact memorisation etc).

If you know someone who may be interested in the research shared in this post (particularly the just-released study I mentioned at the beginning) or the future post on common concerns over flashcards, please do consider sharing this with them to help their education and learning too.

Share

Housekeeping

Due to this Substack being new, the first update I sent out last week for some reason only reached about 40% of recipients’ inbox, with it going into spam for 60% (including mine, amusingly). According to Substack, encouraging readers to reply to the email will fix it. So, I guess, if you have any questions at all (or if you’d just simply like to help), please do consider replying directly to this email - I read them all!

Thank you, & Happy Learning🧠

- shaeda

Subscribe now

Awareness in Schools

Bibliography

Ingebrigtsen, M., Odden Miland, Å., Basteen, J., & Grom Sæle, R. (2025). Effective, scalable, and low cost: The use of teacher-made digital flashcards improves student learning. Applied Cognitive Psychology, 39(1), e4254. https://doi.org/10.1002/acp.4254

Donoghue, G. M., & Hattie, J. A. C. (2021). A meta-analysis of ten learning techniques. Frontiers in Education, 6, Article 581216. https://doi.org/10.3389/feduc.2021.581216

Urrizola, A., Santiago, R., & Arbea, L. (2023). Learning techniques that medical students use for long-term retention: A cross-sectional analysis. Medical teacher, 45(4), 412–418. https://doi.org/10.1080/0142159X.2022.2137016

Larsen DP, Butler AC, Roediger HL 3rd. Repeated testing improves long-term retention relative to repeated study: a randomised controlled trial. Med Educ. 2009 Dec;43(12):1174-81. doi: 10.1111/j.1365-2923.2009.03518.x. PMID: 19930508.

Kerfoot BP, DeWolf WC, Masser BA, Church PA, Federman DD. Spaced education improves the retention of clinical knowledge by medical students: a randomised controlled trial. Med Educ. 2007 Jan;41(1):23-31. doi: 10.1111/j.1365-2929.2006.02644.x. PMID: 17209889.

Sun M, Tsai S, Engle DL, Holmer S. Spaced Repetition Flashcards for Teaching Medical Students Psychiatry. Med Sci Educ. 2021 Apr 6;31(3):1125-1131. doi: 10.1007/s40670-021-01286-y. PMID: 34457956; PMCID: PMC8368120.

Deng F, Gluckstein JA, Larsen DP. Student-directed retrieval practice is a predictor of medical licensing examination performance. Perspect Med Educ. 2015 Dec;4(6):308-313. doi: 10.1007/s40037-015-0220-x. Erratum in: Perspect Med Educ. 2016 Dec;5(6):364. doi: 10.1007/s40037-016-0312-2. PMID: 26498443; PMCID: PMC4673073.

Piza, F., Kesselheim, J. C., Perzhinsky, J., Drowos, J., Gillis, R., Moscovici, K., Danciu, T. E., Kosowska, A., & Gooding, H. (2019). Awareness and usage of evidence-based learning strategies among health professions students and faculty. Medical teacher, 41(12), 1411–1418. https://doi.org/10.1080/0142159X.2019.1645950

Demekash, A. M., Degefu, H. W., & Woldeab, T. A. (2024). Secondary school EFL teachers' awareness for formative assessment for effective learning. Heliyon, 10(19), e37793. https://doi.org/10.1016/j.heliyon.2024.e37793

To be completely honest the demand and the amount of signups and emails has been a little crazy - I definitely did not expect it to be this popular after just a month of adding the waiting list.

I appreciate some have been waiting for a while, so below is just a very quick update on where the app is in no particular order:

TLDR:

• Beta release was pushed back slightly to fix and improve some parts.

• After I add the main core feature (the ability to save cards to one’s own personal database for future review) and sufficiently test it, the app will be released.

• Some new features, languages and ~STEM topics have been validated (and improved).

• I am debating adding a “My Sources” feature.

• The plan was for beta to be completely free but due to the (very) large amount of sign ups it would just be financially impossible. As such, there will likely be a very (very) small monthly fee that will be far cheaper than other AI-related study apps that lack the features shaeda does and have not been tested and validated extensively for accuracy.

Beta Release

• The launch has been delayed slightly while the ‘save-to-database’ functionality has been fully implemented and is working as it should.

• This ‘save’ feature is essential for the spaced repetition system that makes flashcard learning effective (and the bugs annoy me)

  • I could release beta now without this, but I’d prefer to wait and have the main feature in place to allow effective studying right away. Saving of cards that one deems beneficial and important is key to the effectiveness (and time-efficiency) of flashcards - and thus the app

  • This also gives more time for testing and further validating.

Development

• Testing a “Choose your AI/Model” feature

  • For most users the default model will be more than enough, however the ability to have a smarter model may be desirable for those studying at higher proficiency levels for the most complex languages and/or topics at ~MSc level in very niche areas.

    

  • This comes with trade-offs of course (much slower, unknown if it will generate meaningfully more accurate content, much more expensive etc)

  • For context here, one of the most difficult languages for AI models to accurately translate is Arabic. If you invest a lot of time into the prompting and model settings though, you can get good results. I’ll provide the accuracy report for Arabic below:

    

    

• ‘Paperclip’ feature has been added

  • Generates related words or sentences based on the flashcard currently viewing

  • This will be really helpful for situations where there’s very similar words or ways to phrase something that you may prefer to save instead

  • For example, Paperclipping ‘words’ on a flashcard that says “Today I felt tired” would generate new cards such as: asleep; to sleep; bed; morning; night; to rest, etc.

    

• Testing a “Check It” feature for both Language study or Academic study

  • Some cards will be generated intentionally containing 1 or more errors. Some cards will not contain any errors. You would then Check It (the card) and try to identify any potential error/s.

  • The ‘back’ will either reveal the error/s or state that there is none.

  • It’s designed to test actual understanding and comprehension rather than just rote memorisation of the card, an occasional worry if considering using flashcards.

  • Example for Spanish:

    • Front - El perro es muy bonitas.

    • Back - 2 errors: gender and number agreement. El perro es muy bonito.

  • Example for Physics:

    • Front - In a DC circuit, if you keep the resistance constant, the voltage is inversely proportional to the current.”

    • Back - Misstates Ohm's Law (V=I×R). In a DC circuit, if you keep the resistance constant, the voltage is directly proportional to the current

      

• Debating adding a “My Sources” feature

  • Here users simply upload either their own PDFs/notes and the app will generate flashcards based off of this content rather than it’s own.

  • Hundreds of apps of this kind already exist online and there is seemingly little to differentiate them (or justification for the price)

  • However, I see a few potential issues with this approach, some of which I’ll summarise below:

    • Hidden gaps due to technical details:

      • LLMs suffer from a little-known problem known as “Lost In The Middle”. As the name suggests, the model effectively remembers the ‘middle’ parts of a large input worse than it remembers the beginning and end. This affects the length of content one should send in. If the user is not aware of this and inputs, say, 20 pages of a textbook (or worse, 20 pages of hand-written notes) the app will generate a collection of cards, but it would likely have missed (or misremembered) the middle parts. The result of this is a collection of, say, 60 cards, of which 90% were from, say, pages 1-7 and 14-20 and the user would likely not have noticed. Or in the unlikely event they had, they now need to go back into the “submit” section, go back to pages 8-13 and now resubmit these middle pages and go through it again. This is all needless friction and wasted time.

    • Narrow Knowledge:

      • The user would likely assume that any cards generated reflect the entirety of what they need to (or should) know for that particular segment. As such, their knowledge essentially becomes overfitted on this very particular set of cards, or ‘spikey’. They learn this fixed set of cards incredibly well, but outside of this isolated knowledge they have (hidden) gaps. This ‘gap’ risk can also be seen in research.

      • This is part of the reason behind “Deep Dive”. Does the user know the topic or just know the card?

    • Opportunity Cost:

      • It’s not clear to me whether this way (creating cards from one’s own notes or sources etc) is even necessary or beneficial for many languages/subjects

        • As said, there’s a hidden opportunity cost built-in. That is, time spent rewriting/rereading (to produce the notes) is time that could have been better spent on more effective methods. If this time investment did not produce a meaningful improvement in quality (which is what I’m positing for the majority) then it’s actually had a less beneficial effect. Note that this does not mean the user would have not learned anything. Just not as much as they could have if they had instead re-read/wrote slightly less and ‘flashcarded’ more.

        • However a definite use-case for a “My Sources” could be for languages and/or very complex niche subjects that have not (yet) been validated.

          • Although even on this, below I will provide some feedback on niche subjects at MSc level and in other languages (i.e., essentially a 3-fold level of complexity: the subject, the level and the language all simultaneously)

            

            

            

(More on) Accuracy Validation

• Many more languages and ~STEM fields have been validated

  • You can view the list on the updated landing page here (newer signups to the waiting list may already have seen these)

  • I still require the following languages: Czech; Finnish; Georgian; Greek; Hindi; Hungarian; Indonesian; Korean; Swedish; Dutch

  • If anyone is either a native speaker and/or teacher of any of the above and would be interested in testing please email me.

  • Some languages actually failed. For example, one teacher deemed the Mandarin to be ‘correct’ but only up to around B1/B2. After this it would sometimes word too literally rather than how natives would say. Fixing or improving this shouldn’t be too difficult though, so I’ll have the 2 teachers retest after making some updates.

    

    

Pricing

• Initially I was planning on making beta release free for all but there is just far (far) too many signups now that it would be financially impossible, so I will clarify the plan for beta access in the coming weeks.

  • Some important notes here:

    • Very early members who do sign up to any paid tier will be grandfathered in, meaning as the app becomes more developed with better features and AI models gradually become smarter and faster etc the price you pay remains the same for as long as the subscription is maintained (along with of course having a greater say on future updates and features etc)

    • 10% of proceeds will be donated to charities supporting, but not limited to, animal welfare, those with Down syndrome and those with Alzheimer’s or dementia

    • All remaining revenue will be going straight back in to development

    • You can view some additional pricing info on the landing site here

That is all, thank you for reading and again thank you for the (slightly crazy) level of signups.

If you know of anyone who may also be interested in trying beta, feel free to share this with them as well.

And if anyone has any questions at all please feel free to reply to this email.

Thank you!

Subscribe now

PS: If anyone is interested in joining and supporting the development of shaeda, please do reach out.

Subscribe now