Would this kind of app be useful for T1Ds

[tgwynwa2024]

Hi all,


I’ve been thinking about a possible app idea that could support people with type 1 diabetes around mealtimes.


The general concept would be something that makes it easier to estimate insulin doses more accurately for meals, using a more personalised approach — especially for people who already know their insulin-to-carb ratio.


It could also learn over time based on previous meals, and offer gentle feedback or suggestions to improve long-term control. I’m trying to figure out if this would actually be useful or just another “nice idea” that wouldn’t get used.


So I’d love honest thoughts:

• Would something like this be helpful?
• Are there already tools out there that do this well enough?
• Anything that would make this kind of thing more (or less) trustworthy for you?

I’m not selling anything or collecting data — just at the “asking questions” stage. Appreciate any thoughts from people actually living with it day-to-day.


Thanks in advance

 

[Lucyr]

There’s many options that already exist to do this, and there are both free and paid for options already.

 

[tgwynwa2024]

There’s many options that already exist to do this, and there are both free and paid for options already.

Thanks for your feedback! You're right that there are existing apps, but my app is unique in its use of AI to suggest insulin dosages based on meal photos, which isn’t common in other tools. It also remembers previous meals, tracks blood sugar levels, and provides personalised tips for improvement.


These features—meal photo recognition and daily feedback—aim to make managing diabetes easier and more accurate, but I’m still working on refining everything. Would love to hear more thoughts on what could be most useful

 

[Bruce Stephens]

Would something like this be helpful?

Doubt it, to be honest.

I find the problem with many of these kinds of things is they're useful if I feed the information in to them, but the effort of doing that often feels much more than just doing whatever it is myself, so they don't actually provide anything useful.

Maybe someone would find it useful, but I'm guessing not me.

 

[tgwynwa2024]

Doubt it, to be honest.

I find the problem with many of these kinds of things is they're useful if I feed the information in to them, but the effort of doing that often feels much more than just doing whatever it is myself, so they don't actually provide anything useful.

Maybe someone would find it useful, but I'm guessing not me.

Totally understand that – a lot of apps do rely on users constantly inputting data, which can become a chore. That’s exactly the problem I’m trying to solve.


By using photo recognition and remembering common meals, the idea is to reduce that effort as much as possible – so it can offer useful recommendations with minimal input. The goal is to make the app feel helpful rather than like another task to do.


That said, I really appreciate your honesty – feedback like this is exactly what helps shape whether something like this could work or not!"

The whole reason for me looking into this idea, is that I feel the exact same way you do, it becomes tedious and I just do it myself.

 

[Standup]

I keep a food diary which basically achieves the same as you're proposing so I guess it could be useful. However, I'm struggling to see how a photo would workout the weights of food or know the difference between reduced sugar baked beans and normal beans (to pick a random example)?

And if I have to weigh food or enter the variant of food it is then there's lots of apps that do this. It also feels like I would need to enter CGM readings so the app can calculate the impact of different foods and their digestion time. Do CGMs have an API that could be used although its already starting to feel complicated.

 

[Inka]

It wouldn’t be useful to me at all. I know how many carbs are in my meals and the bolus I take not only takes into account those carbs, but what I’m about to do, my blood sugar, any corrections, etc etc. I don’t use any apps like that because they’re time-consuming and a faff, as well as doing a worse job than me in calculating my dose.

 

[Lauren]

This sounds interesting to me. I would certainly try it!

 

[Lucyr]

my app is unique in its use of AI to suggest insulin dosages based on meal photos

I’d absolutely not use it and not recommend it then. AI isn’t going to be accurate enough at carb counting. How would it know if I’m eating pasta from chickpeas or normal pasta? How about if my cake is made with sweetener or sugar? The weight of my rice or bread or whatever? If my porridge is made from whole oats or a sachet?

 

[Leadinglights]

I’d absolutely not use it and not recommend it then. AI isn’t going to be accurate enough at carb counting. How would it know if I’m eating pasta from chickpeas or normal pasta? How about if my cake is made with sweetener or sugar? The weight of my rice or bread or whatever? If my porridge is made from whole oats or a sachet?

You have pretty well said exactly what I was thinking, my example being a sausage where the carbs can be way different in what would appear to be an identical one in a photograph. Certainly I would think not reliable enough to be dictating insulin dose.

 

[tgwynwa2024]

I keep a food diary which basically achieves the same as you're proposing so I guess it could be useful. However, I'm struggling to see how a photo would workout the weights of food or know the difference between reduced sugar baked beans and normal beans (to pick a random example)?

And if I have to weigh food or enter the variant of food it is then there's lots of apps that do this. It also feels like I would need to enter CGM readings so the app can calculate the impact of different foods and their digestion time. Do CGMs have an API that could be used although its already starting to feel complicated.

Great points – and exactly the kind of feedback I was hoping for. You're right that the photo recognition side won’t be perfect for everything, especially when it comes to food variants or portion weights.


The idea is that photo + user context (e.g. 'similar portion to last time' or ‘medium portion’) could make things faster rather than flawless. It wouldn’t replace a food diary for those who already track carefully, but might help those who don’t or want a quicker option.


Also yes – CGM integration is something I’m exploring. Some CGM systems like Dexcom do have APIs, but you're right it adds complexity. Ideally, the app would sync with CGM data to give tips or suggest insulin adjustments based on patterns rather than hard numbers.

While many people out there have brilliant control and systems that work for them, there are just as many who struggle. The aim of the app is to support better long-term control for everyone – by reducing the effort needed to make more informed decisions day-to-day.

Appreciate the thoughtful response – it really helps to work through what’s practical and what’s not.

 

[tgwynwa2024]

It wouldn’t be useful to me at all. I know how many carbs are in my meals and the bolus I take not only takes into account those carbs, but what I’m about to do, my blood sugar, any corrections, etc etc. I don’t use any apps like that because they’re time-consuming and a faff, as well as doing a worse job than me in calculating my dose.

Totally fair – and I really respect that level of control. You're absolutely right that for people who already know their routine and numbers inside out, something like this may not add much.


That said, not everyone feels that confident or consistent day-to-day – and the aim is to support those who want help making quicker, more informed decisions, especially if they’re overwhelmed or still learning.


Really appreciate your perspective.

 

[tgwynwa2024]

You have pretty well said exactly what I was thinking, my example being a sausage where the carbs can be way different in what would appear to be an identical one in a photograph. Certainly I would think not reliable enough to be dictating insulin dose.

Totally get where you're coming from – and I agree this kind of tool would never replace someone’s own judgement or the accuracy they get from weighing or knowing their usual brands.


The idea wouldn't be to dictate insulin doses, but to act more like a support tool – helping users make quicker estimates based on past meals, visual patterns, and their own data (like previous insulin responses to similar meals).


For example, someone could say ‘this is a similar portion to that pasta meal last week’, and it would remember how much insulin they used, what the results were, and offer that as a guide. Not perfect – but potentially useful for people still learning or those struggling with consistency.


Appreciate the honest feedback – the trust issue is 100% something I’d need to get right.

 

[Bruce Stephens]

For example, someone could say ‘this is a similar portion to that pasta meal last week’, and it would remember how much insulin they used, what the results were, and offer that as a guide. Not perfect – but potentially useful for people still learning or those struggling with consistency.

So in some sense it could be an automated food diary as much as (or rather than) an AI tool offering advice.

 

[tgwynwa2024]

So in some sense it could be an automated food diary as much as (or rather than) an AI tool offering advice.

Yes, I think you’ve sumed it up really well — it’s closer to an automate food diary that learns over time, rather than an AI tool trying to tell you what to do.


But it would still use AI — just in the background. Things like:

• Recognising foods from images to cut down manual input
• Spotting patterns in how your blood sugar responds to certain meals
• Offering helpful nudges based on your own past data

It wouldn’t replace anyone’s judgement — just aim to make things a bit easier and more consistent, especially for people who find food tracking or carb counting overwhelming

 

[martindt1606]

I’m reasonably happy with my bolus (and basal) ratios and how that impacts meal time dosage if my carb counting is correct. The issue I have is the impact of other factors eg pre and post meal exercise or stress especially if not anticipated at meal time.

 

[Bruce Stephens]

Yes, I think you’ve sumed it up really well — it’s closer to an automate food diary that learns over time, rather than an AI tool trying to tell you what to do.

I was imagining the interface as providing a suggested insulin dose.

But it sounds like it might instead be offering information about what I did and what happened when I had meals which it considers similar looking. So less an AI tool and more an automated food diary which uses a bit of AI to find photos of meals which look similar to each other.

I'm not sure that makes it much more practical to produce but it does sound like something that could plausibly exist.

 

[rebrascora]

I wouldn't be interested in an app as I store most of the information I need in my head and I also doubt it would cope with a low carb way of eating which also requires injecting for glucose release from protein.

I think this forum probably isn't necessarily representative of the Type 1 population as a whole, as many of the members here are likely more informed and switched on to their diabetes management due to exchanging thoughts and ideas and tips with other Type 1s here. What I am saying is that you might be canvassing views from people who would be less likely to find it helpful than the general population of Type 1s.

Personally I am not joined at the hip with my phone and I very, very rarely use apps. I even prefer a reader with my Libre to the phone app and I think many people have a mistrust of AI particularly, so I suspect downplaying that element might be a good idea.

It concerns me that people are becoming more and more reliant on technology and that reliance makes them very vulnerable if/when the technology fails as they don't have basic skills to fall back on.

 

[Proud to be erratic]

Personally, I'm increasingly wary of anything using AI. It's taking over my attempts at searches for discrete information and often generating nonsense (looks plausible, but isn't true).

Secondly, there is a risk of this app creating further confusion: we already frequently encounter Newbies not understanding the numbers and their significance, which they get from finger prick tests in comparison with CGM data. Never mind the limitations of CGM data. Obviously, without seeing what this app will display, I foresee the potential of a 3rd set of information adding to the confusion of diabetes and (boosted by the apparent legitimacy of AI) potentially giving an apparent sense that the app will be right! If the app is able to learn only about what the user is both experiencing and encountering AND NOT drawing from worldwide or even UK wide experience then I might feel more comfortable about an app steering me towards reliable numbers.

 

[Inka]

Totally fair – and I really respect that level of control. You're absolutely right that for people who already know their routine and numbers inside out, something like this may not add much.


That said, not everyone feels that confident or consistent day-to-day – and the aim is to support those who want help making quicker, more informed decisions, especially if they’re overwhelmed or still learning.


Really appreciate your perspective.

I get what you’re saying, but I think education would benefit those people more. So, if your app could incorporate some kind of learning/information modules, maybe leading to carb ratios or whatever that can be carried on by the app, then that would be more useful IMO.

However, an additional worry would be the uninformed/less confident/newly diagnosed blindly following an incorrect or incomplete app insulin dosage suggestion that, to another person, would obviously be wrong. They’d then take the wrong dose of insulin.