ChatGPT French travel poster

A Second Shot at Perfect Posters – ChatGPT’s Image Tweaker

The big ChatGPT news in recent weeks is about images, rather than words. The AI frontrunner has added a facility to selectively re-prompt for parts of an image, allowing us to tweak sections that don’t live up to prompt expectations.

In essence, this new facility gives us a second shot at saving otherwise perfect output from minor issues. And for language learning content, like posters and flashcards, the biggest ‘minor’ issue – the poor spellings that crop up in AI image generation – makes the difference between useful and useless material.

Rescuing ChatGPT Posters

Take this example. It’s a simple brief – a stylish, 1950s style travel poster for France. Here’s the prompt I used to generate it:

Create a vibrant, stylish 1950s style travel poster featuring Paris and the slogan “La France”.

I wanted the text “La France” at the top, but, as you can see, we’ve got a rogue M in there instead of an N.

ChatGPT generated image of a French travel poster

To target that, I tap the image in the ChatGPT app. It calls up the image in edit mode, where I can highlight the areas that need attention:

ChatGPT image editing window

Then, I press Next, and can re-prompt for that part of the image. I simply restate the slogan instructions:

The slogan should read “La France”.

The result – a correct spelling, this time!

ChatGPT French travel poster

It can take a few goes. Dodgy spelling hasn’t been fixed; we’ve just been given a way to try again without scrapping the entire image. Certain details also won’t be retained between versions, such as the font, in this example. Others may be added, like the highly stylised merging of the L and F in the slogan (a feature, rather than a bug, I think!).

But the overall result is good enough that our lovely 1950s style poster wasn’t a total write-off.

Another case of AI being highly imperfect on its own, but a great tool when enhanced by us human users. It still won’t replace us – just yet!

Image tweaking is currently only available in the ChatGPT app (iOS / Android).

Neon robots racing. Can Claude 3 win the AI race with its brand new set of models?

Claude 3 – the New AI Models Putting Anthropic Back in the Game

You’d be forgiven for not knowing Claude. This chirpily-named AI assistant from Anthropic has been around for a while, like its celebrity cousin ChatGPT. But while ChatGPT hit the big time, Claude hasn’t quite progressed beyond the Other Platforms heading in most AI presentations – until now.

What changed everything this month was Anthropic’s release of all-new Claude 3 models – models that not only caught up with ChatGPT-4 benchmarks, but surpassed them. It’s wise to take benchmarks with a pinch of salt, not least because they’re often internal, proprietary measures. But the buzz around this latest release echoed through the newsletters, podcasts and socials, suggesting that this really was big news.

Tiers of a Claude

Claude 3 comes in three flavours. The most powerful, Opus, is the feistiest ChatGPT-beater by far. It’s also, understandably, the most processor-intensive, so available only as a premium option. That cost is on a level with competitors’ premium offerings, at just under £20 a month.

But just a notch beneath Opus, we have Sonnet. That’s Claude 3’s mid-range model, and the one you’ll chat with for free at Anthropic reports that Sonnet still pips ChatGPT-4 on several reasoning benchmarks, with users praising how naturally conversational it seems.

Finally, we have a third tier, Haiku. This is the most streamlined of the three in terms of computing power. But it still manages to trounce ChatGPT-3.5 while coming impressively close to most of those ChatGPT-4 benchmarks. And the real clincher?

It’s cheap.

Haiku costs a fraction of the price per token of competing models to developers. That means it’s a lot cheaper to build it into language learning apps, opening up a route for many to incorporate AI into their software. That lower power usage too is a huge win against a backdrop of serious concerns around AI energy demands.

Claude and Content Creation

So how does it measure up in terms of language learning content? I set Claude’s Sonnet model loose on the sample prompt from my recent Gemini Advanced vs. ChatGPT-4 battle. And the verdict?

It more than holds its own.

Here’s the prompt (feel free to adapt and use this for your own worksheets – it creates some lovely materials!):

Create an original, self-contained French worksheet for students of the language who are around level A2 on the CEFR scale. The topic of the worksheet is “Reality TV in France“.

The worksheet format is as follows:

– An engaging introductory text (400 words) using clear and idiomatic language
– Glossary of 10 key words / phrases from the text (ignore obvious cognates with English) in table format
– Reading comprehension quiz on the text (5 questions)
– Gap-fill exercise recycling the same vocabulary and phrases in a different order (10 questions)
– ‘Talking about it’ section with useful phrases for expressing opinions on the topic
– A model dialogue (10-12 lines) between two people discussing the topic
– A set of thoughtful questions to spark further dialogue on the topic
– An answer key covering all the questions

Ensure the language is native-speaker quality and error-free.

Sonnet does an admirable job. If I’m nitpicking, the text is perhaps slightly less fun and engaging than Gemini Advanced. But then, that’s the sort of thing you could sort out by tweaking the prompt.

Otherwise, it’s factual and relevant, with some nice authentic cultural links. The questions make sense and the activities are useful. Claude also followed instructions closely, particularly with the inclusion of an answer key (so often missing in lesser models).

There’s little to quibble over here.

A language learning worksheet created with Claude 3 Sonnet.

A Claude 3 French worksheet. Click here to download the PDF!

Another Tool For the Toolbox

The claims around Claude 3 are certainly exciting. And they have substance – even the free Sonnet model available at produces content on a par with the big hitters. Although our focus here is worksheet creation, its conversational slant makes it a great option for experimenting with live AI language games, too.

So if you haven’t had a chance yet, go and get acquainted with Claude. Its all-new model set, including a fabulous free option, makes it one more essential tool in the teacher’s AI toolbox.

Masses of digital text. AIs with a large context window can process much more of it!

Gemini’s Long Context Window – a True Spec Cruncher

Maybe you’ve noticed that Google’s Gemini has been making gains on ChatGPT lately. Of all its recent impressive improvements, one of the lesser-sung features – at least in AI for Ed circles – is its much enhanced context window.

The context window is essentially how much text the AI can ‘remember’, and work with.  Google’s next model boasts one million characters of this memory, leaving other models – which count their own in the hundreds of thousands – in the dust. It blows open the possibilities for a particular kind of AI task: working with long texts.

Language learners make use of all kinds of texts, of course. But one particularly unwieldy (although hugely useful) type where this new feature could help is the exam spec.

Exam Spec Crunching with AI

Language exam specs are roadmaps to qualifications, listing the knowledge and skills students need to demonstrate linguistic competency. But they have a lot of fine detail that can bog us down.

As a content creator, one thing that challenges me is teasing out this detail into some kind of meaningful arrangement for student activities. There is a mass of vocab data in there. And as systematic as it is, abstract lists of connectives, temporal adverbs and helper verbs don’t make for very student-friendly lesson material.

With a massive text cruncher like Gemini, they are a lot easier to process. Just drag in your spec PDF (I’ve been playing around with the new AQA GCSE German doc), and tease out the material in a more useful format for planning:

Take this German exam spec, and create an outline plan of three terms of twelve lessons that will cover all of the thematic material.

Additionally, it can help in creating resources that cover all bases:

Create a short reading text to introduce students to the exam topic “Celebrity Culture”. It should be appropriate for students aiming for the top tier mark in the spec. In the text, make sure to include all of the prepositions from the prescribed word list.

With a long textual memory, it’s even possible to interrogate the spec after you’ve uploaded it. That’s literally just asking questions of the document itself – and, with that bigger window, getting answers that don’t overlook half the content:

If students have one year to learn ALL of the prescribed vocabulary in the spec, how many words should they be learning a week? Organise them into weekly lists that follow a broadly thematic pattern.

Supersized Context Window – Playing Soon at an AI Near You!

For sure, you can use these techniques on existing platforms straight away. However, due to the smaller context window, results might not always be 100% reliable (although it’s always fun trying!). For the new Google magic, we’ll have to wait just a little longer. 

But from the initial signs, it definitely looks worth the wait!

Gemini’s new supersized context window is available only in a limited released currently, and only via its AI Studio playground. Expect to see it coming to Gemini Advanced very soon!

Lots of owls - not quite the Duolingo one, but they're looking strict nonetheless!

Going Cold Owl (Once More, With Feeling)

Well, friends – I finally did it. I beat my Duolingo Leagues addiction.

Yes, it’s been months in the making. It was back in May last year that I wrote about that tipping point where Duolingo’s gamification switches from incentivising learning, to subverting it. After more than two years in Diamond, my main Duo activity had become all about chasing points, rather than finishing courses.

I’d even sussed out a ‘good’ trick – an easy lesson on Hindi alphabet characters – that I could mindlessly run through to amass points in my idle moments. Then I’d justify it with the two or three daily lessons I’d do that were relevant to my language goals.

Only it wasn’t good at all, in the grand scheme of things. What is was, was a great way to de-direct my learning time on what is otherwise a nice little supplementary app for my language plan.

Surely my idle moments were worthy of something more.

Letting Yourself Fall (and Trusting in a Soft Landing)

Gamified apps like Duolingo are adept at creating a sense of worth out of otherwise valueless tokens. A league title here, an achievement there – it feels good to notch these accolades up. But the real value of a language learning app is never the dressing, it’s the language-y filling in the centre. When it stops being about that, it’s time to stop.

So, in a final moment of take-the-plunge bravery, I stopped chasing the demotion cut-off last week. I let myself fall. And I won’t pretend I didn’t have several fleeting moments of panic when I thought – it’s not too late, I can still catch up!

But nope – I was resolute. And when, on Monday morning, I woke up to – gasp – Obsidian and not Diamond, do you know what? It felt liberating.

I’d broken the cycle.

didn’t feel like I’d left an exclusive club. I didn’t feel like I’d lost anything valuable. I just felt relieved.

Duolingo – Done Right (aka all things in moderation)

Since then, I’ve kept up my lengthy Duolingo streak. But not through mindless tap-tap-tap lessons – instead, through a couple of Swedish and Gaelic lessons to progress through the courses. Just as it should be.

If you’re in this autopilot pattern yourself, ask yourself: what do I gain? Reframing any kind of daily screen-addictive behaviour in this way is the first step in changing it.

Keen on more tips for breaking digital additions? Check out:

Foreign alphabet soup (image generated by AI)

AI Chat Support for Foreign Language Alphabets

I turn to AI first and foremost for content creation, as it’s so good at creating model foreign language texts. But it’s also a pretty good conversational tool for language learners.

That said, one of the biggest obstacles to using LLMs like ChatGPT for conversational practice can be an unfamiliar script. Ask it to speak Arabic, and you’ll get lots of Arabic script. It’s usually smart enough to work out if you’re typing back using Latin characters, but it’ll likely continue to speak in script.

Now, it’s easy enough to ask your AI platform of choice to transliterate everything into Latin characters, and expect the same from you – simply instruct it to do so in your prompts. But blanket transliteration won’t help your development of native reading and writing skills. There’s a much better best of both worlds way that does.

Best of Both Worlds AI Chat Prompt

This prompt sets up a basic conversation environment. The clincher is that is give you the option to write in script  or not. And if not, you’ll get what script should look like modelled right back at you. It’s a great way to jump into conversation practice even before you’re comfortable switching keyboard layouts.

You are a Modern Greek language teacher, and you are helping me to develop my conversational skills in the language at level A2 (CEFR). Always keep the language short and simple at the given level, and always keep the conversation going with follow-up questions.

I will often type in transliterated Latin script, as I am still learning the target language alphabet. Rewrite all of my responses correctly in the target language script with any necessary grammatical corrections.

Similarly, write all of your own responses both in the target language script and also a transliteration in Latin characters. For instance,

Καλημέρα σου!
Kaliméra sou!

Do NOT give any English translations – the only support for me will be transliterations of the target language.

Let’s start off the conversation by talking about the weather.

This prompt worked pretty reliably in ChatGPT-4, Claude, Copilot, and Gemini. The first two were very strong; the latter two occasionally forget the don’t translate! instruction, but otherwise, script support – the name of the game here – was good throughout.

Try changing the language (top) and topic (bottom) to see what it comes up with!


An illustration of a robot scribe writing AI prompts for ChatGPT or Gemini

Power to the Prompts : Fave Tweaks for AI Worksheets

Content creation is what AI excels at. That’s a gift to language learners and teachers, as it’s the easiest thing in the world to create a set of prompts to run off original, immersive worksheets.

AI isn’t great at everything, though, which is why our prompts need to tweak for its weaknesses – the things that are obvious gotchas to language folk, but need explaining to a general assistant like AI. Fortunately, in most cases, all you need is an extra line or two to put it right.

Here are five of my favourite prompt-enhancers for worksheets, covering everything from vocab to copyright!

Five Tweaks for Perfect Prompts

Cut Out the Cognates

Cognates are generally great for language learners. As words that are instantly recognisable, they’re extra vocab for free. But for that very reason, they’re not the ones that worksheets should be making a song and dance about. It’s not particularly useful, for example, to have “der Manager” picked out in your German glossary as a key word. Yes, I worked that one out!

Try this in your vocab list prompts, bearing in mind that not all platforms will work equally well with it (Gemini aced it – ChatGPT sometimes gets it):

Don’t list items that are obvious English loanwords or cognates with English.

Highlight the Good Stuff

You know what I mean by the good stuff – those structures and snippets of languages that are really frequent, and really reusable. I like to call them sentence frames – you can learn them, and switch in other words to add to your own linguistic repertoire.

You can ask AI to draw attention to any really pertinent ones in your target language texts:

Highlight (in bold and italics) the most pertinent key words and phrases for the topic, and provide a brief glossary of them at the end.

Make It Colloquial

Vanilla AI can sound bookish and formal. That’s no good as a model for everyday speech, so polish your prompts with a wee push to the colloquial:

Make the language colloquial and idiomatic, in the style of a native speaker.

Include an Answer Key

It might seem obvious, but if you’re making materials for self-study, then you will find an answer sheet indispensable. It’s an often overlooked finishing touch that makes a worksheet truly self-contained:

Include an answer key for all questions at the end.

Covering Your Back With Copyright

Copyright issues have been bubbling in the background for LLMs for some time now. They produce texts based on vast banks on training data, which isn’t original material, of course. But in theory, the texts that pop out of it should be completely original.

It can’t hurt to make that explicit, though. I like to add the following line to prompts, especially if I’m intending to share the material beyond personal use:

Ensure that the text is completely original and not lifted directly from any other source.


What are your favourite tweaks to make perfect prompts? Let us know in the comments!

A collage of lots of word and picture cards.

Treating Leeches – Strategies for Suspended Anki Cards

How do you deal with leeches?

I’m not talking about traditional medicine here (not to downplay the modern application of the age-old treatment at all!). The leeches I’m more concerned with on the day-to-day are those Anki cards you forget so persistently that the app takes charge, suspending them from your deck.

It’s an apt description for an item that sucks away your time and motivation. I don’t know about you, but I also get that sinking feeling of failure when “card was a leech” pops up baldly.

Catching Leeches

First of all, fight that feeling. Leeches can creep up for a number of reasons, and your memory lapse is the least of them. Despite the cold rebuke, Anki means well. It suspends the cards to save you wasting any more time on part of your learning strategy that isn’t quite working. So, for now, let them go.

Instead, schedule a review of leeches regularly. Once a month or so seems about right if you’re a prolific language learning user – I always have a couple to deal with in that time span. In the Anki desktop app, head to Browse. Then, there are two ways to list leeches. You can simply highlight Suspended under Card State in the left-hand menu.

Exposing leeches via Suspended Cards in the Anki Browse window

Exposing leeches via Suspended Cards in the Anki Browse window

Otherwise, you can use the fact that Anki tags leech cards with the text leech to draw them out. Highlight one of your decks in the left-hand menu, then in the bar at the top of the Browse panel, add the text tag:leech to narrow the results to that set.

Exposing leeches via tags in the Anki Browse window

Exposing leeches via tags in the Anki Browse window

Now out in the open, we need to think of a rehabilitation strategy for our annoyingly helpful leeches.

Treating the Cause, Not the Problem

It’s tempting to just un-suspend by removing that leech tag, and pop the card right back in the deck. But there’s a reason Anki singled it out – something wasn’t working.

Often, it’s not simply failure to remember. Many of mine aren’t words I’ve forgotten, but words I get mixed up – either with other target language words, or with the wrong English translation. For example, in Greek, I leeched out παραδέχομαι (paradéchomai – admit) with αποδέχομαι (apodéchomai – accept), due to their similarity – same root verb, different prefix.

It’s not always just soundalikes, either, but happens with concepts. Left and right are a case in point in Swahili. I know both words very well – kushoto and kulia – but I’d always say one for the other, to the point that they were marked as leeches.

could recall them – it had just become 50/50 whether I’d say one or the other!

These cases of interference usually arises because there’s a lack of distinguishing information on the vocab card. The easiest way to fix that is to make your cards clearer and more precise. Any defining detail will do, and with language learning, context is key. Short sentences that embed the vocabulary items are perfect. To give the brain more to hang onto, you can expand them from basic X is Y types to X is Y, so/because…, and even make use of allegory and rhyme in your examples.

Taking the Swahili example, there’s a topical hook with those that adds layers of meaning: politics. There’s also a good rhyme for kulia in pia (also). So to my card, I add the sentence (and forgive the unpalatable mention of unpopular politicians here) Boris yuko kulia, na Rishi pia (Boris is on the right, and Rishi too).

And (of course) there’s a wee AI tip for that. If you struggle to fine rhymes – not unreasonable if you’re at an early stage in a language – then just ask your LLM of choice for rhyming pointers, or even entire couplets. It’s one of the things is does a pretty decent job of!

Asking AI for rhyming words in foreign languages.

Asking ChatGPT-4 for rhyming words in foreign languages.



Leeches are an initially frustrating but ultimately helping feature of the Anki lifestyle! Do you have alternative methods for bashing them? Let us know in the comments!

Lots of websites floating around a central AI sphere

Powered by AI : Some Favourite Tools

If you’ve been following my recent posts, you’ll know that I’ve been embedding AI deeply into my language learning routine.

There are some truly min-blowing ways to incorporate ‘raw AI’ – using direct prompts with LLMs like ChatGPT – into your learning, from live activities with personality, to custom content creation. But likewise, there are plenty of ready-to-run, AI-infused sites that you can use for language fun.

Here are a few of my favourites!

KOME.AI – YouTube Transcript Generator

I came across this when asked by a friend struggling to transcribe a long conference talk video for work. Surely there’s some way to automate that? And sure there was, and it’s 

It’s not the only transcription service out there – there are numerous ones, competing for supremacy – but it’s the most straightforward, it’s multilingual, and it’s free. It’s also fast, seemingly drawing on already-existing auto-captions where available, before kicking in with other tools where necessary. I pasted in a short news clip about the German teacher shortage – I had a transcript almost immediately., generating YouTube transcripts with AI

Delphi’s Digital Clones

Most of the best prompt strategies involve telling AI who you want it to be. has taken that a step further, by digitally cloning experts in their field – their words and big takes, at least – and making them available to the public. Think ‘coach in a box’.

While the site is set up for those wishing to clone knowledge-imparting versions of themselves (language coach, anyone?), you can browse and chat with many of their demo models. The philosopher collection is particularly enlightening.


AI’s whole bag is text generation. Now, the big tech talk a bit game about these platforms also being digital assistants, but they’re basically content whizzes, and can still be lacking in other task performance areas. Searching seems to be one of these blind spots, which you’ll have realised quickly when faced with Bing’s sometimes laughably off-topic search results. aims to change all that. The developers have taken an LLM, and purpose-designed it for finding sources and answering questions. Consequently, it’s much more useful for learners, educators, researchers, and anyone who doesn’t want their AI to completely miss the point. It’s the future of search.

Web search infused with AI from


AI-generated music has been sneaking up on us all very quietly. It was text generation that was making all the bolshy fuss, up to now. Music was still very much experimental, and out of the question unless you were running models on a powerful testing machine.

But suddenly, we have services that can create whole songs – including lyrics – from a simple prompt. not only gives you that for free – you have ten tracks a day for nothing – but it’s fast, and uncannily good for an early release. And, although they don’t shout from the rooftops about it, it’s also a polyglot!

These aren’t just great, handy, fun sites to use. They also show how broad the brush of AI is, and will be, in the future. They offer a taste of how embedded the tech will become in all sorts of areas of our lives in the coming years.

Are there any emerging AI services you’re a fan of? Let us know if the comments!

Two AI robots squaring up to each other

AI Worksheet Wars : Google Gemini Advanced vs. ChatGPT-4

With this week’s release of Gemini Advanced, Google’s latest, premium AI model, we have another platform for language learning content creation.

Google fanfares Gemini as the “most capable AI model” yet, releasing benchmark results that position it as a potential ChatGPT-4 beater. Significantly, Google claims that their new top model even outperforms humans at some language-based benchmarking.

So what do those improvements hold for language learners? I decided to put Gemini Advanced head-to-head with the leader to date, ChatGPT-4, to find out. I used the following prompt on both ChatGPT-4 and Gemini Advanced to create a topic prep style worksheet like those I use before lessons. A target language text, vocab support, and practice questions – perfect topic prep:

Create an original, self-contained French worksheet for students of the language who are around level A2 on the CEFR scale. The topic of the worksheet is “Reality TV in France“.

The worksheet format is as follows:

– An engaging introductory text (400 words) using clear and idiomatic language
– Glossary of 10 key words / phrases from the text (ignore obvious cognates with English) in table format
– Reading comprehension quiz on the text (5 questions)
– Gap-fill exercise recycling the same vocabulary and phrases in a different order (10 questions)
– ‘Talking about it’ section with useful phrases for expressing opinions on the topic
– A model dialogue (10-12 lines) between two people discussing the topic
– A set of thoughtful questions to spark further dialogue on the topic
– An answer key covering all the questions

Ensure the language is native-speaker quality and error-free.

I then laid out the results, with minimal extra formatting, in PDF files (much as I’d use them for my own learning).

Here are the results.


ChatGPT-4, gives solid results, much as expected. I’d been using that platform for my own custom learning content for a while, and it’s both accurate dependable.

The introductory text referenced the real-world topic links very well, albeit a little dry in tone. The glossary was reasonable, although ChatGPT-4 had, as usual, problems leaving out “obvious cognates” as per the prompt instructions. It’s a problem I’ve noticed often, with other LLMs too – workarounds are often necessary to fix these biases.

Likewise, the gap-fill was not “in a different order”, as prompted (and again, exposing a weakness of most LLMs). The questions are in the same order as the glossary entries they refer to!

Looking past those issues – which we could easily correct manually, in any case – the questions were engaging and sensible. Let’s give ChatGPT-4 a solid B!

A French worksheet on Reality TV, created by AI platform ChatGPT-4.

You can download the ChatGPT-4 version of the worksheet from this link.

Gemini Advanced

And onto the challenger! I must admit, I wasn’t expecting to see huge improvements here.

But instantly, I prefer the introductory text. It’s stylistically more interesting; it’s just got the fact that I wanted it to be “engaging”. It’s hard to judge reliably, but I also think it’s closer to a true CEFR A2 language level. Compare it with the encyclopaedia-style ChatGPT-4 version, and it’s more conversational, and certainly more idiomatic.

That attention to idiom is apparent in the glossary, too. There’s far less of that cognate problem here, making for a much more practical vocab list. We have some satisfyingly colloquial phrasal verbs that make me feel that I’m learning something new.

And here’s the clincher: Gemini Advanced aced the randomness test. While the question quality matched ChatGPT-4, the random delivery means the output is usable off the bat. I’m truly impressed by that.

A French worksheet on Reality TV, created by Google's premium AI platform, Gemini Advanced.

You can download the Gemini Advanced version of the worksheet from this link.

Which AI?

After that storming performance by Gemini Advanced, you might expect my answer to be unqualified support for that platform. And, content-wise, I think it did win, hands down. The attention to the nuance of my prompt was something special, and the texts are just more interesting to work with. Big up for creativity.

That said, repeated testing of the prompt did throw up the occasional glitch. Sometimes, it would fail to output the answers, instead showing a cryptic “Answers will follow.” or similar, requiring further prompting. Once or twice, the service went down, too, perhaps a consequence of huge traffic during release week. They’re minor things for the most part, and I expect Google will be busy ironing them out over coming months.

Nonetheless, the signs are hugely promising, and it’s up to ChatGPT-4 now to come back with an even stronger next release. I’ll be playing around with Gemini Advanced a lot in the next few weeks – I really recommend that other language learners and teachers give it a look, too!

If you want to try Google’s Gemini Advanced, there’s a very welcome two-month free trial. Simply head to Gemini to find out more!

An illustration of a cute robot looking at a watch, surrounded by clocks, illustrating AI time-out

Avoiding Time-Out with Longer AI Content

If you’re using AI platforms to create longer language learning content, you’ll have hit the time-out problem at some point.

The issue is that large language models like ChatGPT and Bard use a lot of computing power at scale. To keep things to a sensible minimum, output limits are in place. And although they’re often generous, even on free platforms, they can fall short for many kinds of language learning content.

Multi-part worksheets and graded reader style stories are a case in point. They can stretch to several pages of print, far beyond most platform cut-offs. Some platforms (Microsoft Copilot, for instance) will just stop mid-sentence before a task is complete. Others may display a generation error. Very few will happily continue generating a lengthy text to the end.

You can get round it in many cases by simply stating “continue“. But that’s frustrating at best. And at worst, it doesn’t work at all; it may ignore the last cut-off sentence, or lose its thread entirely. I’ve had times when a quirky Bing insists it’s finished, and refuses, like a surly tot, to pick up where it left off.

Avoiding Time-Out with Sectioning

Fortunately, there’s a pretty easy fix. Simply specify in your prompt that the output should be section by section. For example, take this prompt, reproducing the popular graded reader style of language learning text but without the length limits:

You are a language tutor and content creator, who writes completely original and exciting graded reader stories for learners of all levels. Your stories are expertly crafted to include high-frequency vocabulary and structures that the learner can incorporate into their own repertoire.

As the stories can be quite long, you output them one chapter at a time, prompting me to continue with the next chapter each time. Each 500-word chapter is followed by a short glossary of key vocabulary, and a short comprehension quiz. Each story should have five or six chapters, and have a well-rounded conclusion. The stories should include plenty of dialogue as well as prose, to model spoken language.

With that in mind, write me a story for French beginner learners (A1 on the CEFR scale) set in a dystopian future.

By sectioning, you avoid time-out. Now, you can produce some really substantial learning texts without having to prod and poke your AI to distraction!

There may even be an added benefit. I’ve noticed that the quality of texts output by section may even be slightly higher than with all-at-once content. Perhaps this is connected to recent findings that instructing AI to thing step by step, and break things down, improves results.

If there is a downside, it’s simply that sectioned output with take up more conversational turns. Instead of one reply ‘turn’, you’re getting lots of them. This eats into your per-conversation or per-hour allocation on ChatGPT Plus and Bing, for example. But the quality boost is worth it, I think.

Has the section by section trick improved your language learning content? Let us know your experiences in the comments!