Hacker News Reader

I photographed everything in my fridge and everything in my cupboards and uploaded it to an LLM and photographed by shopping receipts and asked it to create weight loss meal plans based on my existing shopping and food buying habits.

Worked real nice.

Did I make it into a business? Should everything be a SAAS? Nah.

I'd pay you for that. Take my money you SaaS!

Why? Just drop the pics into CHatGPT and say "make me a weight loss meal plan based on what I have in the fridge and the cupboards"

You could make a website that does that in a couple of hours.

But then I need it to do the web order as well finding the correct products so things get delivered.

After that I need it to tell how to organize the bag into the fridge.. Ideally I'd have meta AI glasses or similar to that.

I can use ChatGPT once for doing all that, but then I need it all automated and a fluid experience.

My ADHD brain doesn't trigger dopamine for doing the same thing twice.

Wow that's a great idea thanks for sharing

What's a good LLM for this kind of task?

Gemini 2.5 Flash works quite well for it.

I mean the llm you used is a saas. But the question is whether a framework for doing this has value (I think it does)

I then added on an SMS message that tells me things I COULD eat at every meal time. It does not say what I must eat, only presents options - I don't like being told what I must eat, all I want to know is what's the best choices of what I have, then I can choose.

  Time for Lunch.
  Choose one of:
  ---
  Tuna salad: 1/2 cup tuna (drained) + 3 cups salad greens + 1/2 cup beetroot + vinegar/lemon (≈850 kJ)
  ---
  Egg salad: 2 boiled eggs (chopped) + 3 cups salad veg + 1/4 cup grated cheese (≈1000 kJ)
  ---
  Chicken salad bowl: 1/2 cup cooked chicken pieces + 3 cups salad veg (≈950 kJ)
  ---
  Minestrone soup: 2 cups (≈1060 kJ)
  ---
  Creamy chicken soup: 2 cups (≈1090 kJ)

I think it's kind of like AI coding, you can do an awful lot just in the chat with chatgpt but there's a lot of advantage of a framework like cursor/claude code to manage more complicated context and run things for you. For this case maybe recipe gen is something you want more control over and are ok managing the context of things in your pantry and other stuff in your head. But I think if the context of things like your pantry was indexed really well there'd be a huge advantage in having a framework beyond chatgpt messages.

It would be helpful to have a step-by-step walkthrough on the site (video or screenshots). The "How It Works" section is a little vague to suit that bill.

I'm looking for both an example of what to expect and inspiration for how I could be eating with your service.

I'd also recommend focusing on the API integrations, and/or whatever other features separate elevate this beyond asking $LLM for a weekly meal plan and shopping list taking into account dietary preferences.

Totally agree on a video walkthrough, and it's on my list. I've mostly been putting it off because I'm not sure if I'm capable of making a polished enough presentation, which is a bad excuse.

Also 100% on messaging to differentiate from just using an LLM to do this yourself. This one feels a bit harder to me because I think it perhaps is not obvious until you get into the product and use it for a bit, though the integrations are a pretty good concrete example.

Descript has some nice video/screen recording tools to help beginners make this kind of thing look moderately professional. I've used it for various walkthroughs in my previous place of employment.

We really need this as my partner hates the idea of a regular meal plan, but then struggles every week to decide on a plan.

So I threw one together myself, but never quite finished it off, so very keen to use yours!

But the setup for this is overwhelming, so many forms. And then the meal plan is taking ages to generate.

I'd have bounced long ago if I was a normal person.

So it finally finished and I mentioned we regularly have a roast on Sunday, Monday or Tuesday and it decided to suggest a roast on Sunday and Tuesday.

Also image loading is very slow.

Think it's a great idea, generally looks pretty good, etc, just definitely look at the setup. I do think it pretty bloody slow though, and I'd love to switch meals, on mine you could just swipe left or right on a meal, it had actually generated 10 auggestions per day that you could rotate through to sorts customize it yourself. I didn't use AI on the fly though, I just had AI generated 1,000s of options up front and then randomized it.

Ah, I hear you on the onboarding wizard. I did my best to keep it streamlined but understand it's still overwhelming. Hard to balance keeping it light with collecting enough data to provide reasonable recommendations.

The meal plan generation does take a bit, _typically_ it's around 90ish seconds but it does vary a lot. That's odd the images are loading slowly, I'll need to look into that.

You can ask for meal replacements (with or without input instructions) from the meal plan view. The replacements are not particularly snappy though. Trying to balance efficient usage of AI with a snappy UX is tricky.

I'm in the UK, if you're hosting US that's probably part of it, it's easy to forget that mainstream websites have a ton of CDN.

But yeah, they were gradual rendering over 1-2 seconds.

I was on mobile, not sure if you've used responsive image sizing or not as it's a faff to inspect source, but if you haven't it's a game changer for making image heavy websites snappy.

Then again, depends where you're getting the images from, could just be you were getting throttled because you got a bit of a HN hug of death.

Yeah that tracks. I'm currently only hosting out of us-central1 in GCP. Haven't got around to setting up multi-region or a proper CDN yet.

How is this any different from say ollie.ai? This looks alot more like a funnel for grocery delivery combined with recipe database functionality not dissimilar from apps like Paprika.

The big push on the website from my impression is that its a funnel to Instacart or Kroger. Does it limit its recipes to whats in stock around me? How accurate is the stock levels shown in the app?

Are you only limited to Kroger or Instacart to see what groceries are required for a recipe?

I didn't see any outward showing of family sharing either, such as a meal calendar.

Can you speak to how it handles dealing with preferences, e.g. no gluten or seafood? This is critical in surfacing recipes, to me accuracy on this is very critical.

You're right that Meals You Love integrates with Instacart and Kroger for ordering meal plan groceries but it's not primarily a grocery funnel and this has no impact on meal plan construction. Really these are just value-add integrations IMO. For our family, the real value is actually providing a means for us to save our favorite recipes and get AI-recommended recipes peppered in. The grocery ordering is secondary, though quite useful for us.

The second big value prop for us has actually been the new "Quick Bites" feature which generates recipes based on ingredients you have on hand (either tracked in your pantry in the app or simply input manually). For instance, we'll often have some ingredients we want to use up but need suggestions on what to make. In this case, there's no grocery integration at all since we're using entirely stuff we already have.

Good question on the meal calendar. You CAN share meal plans (as well as recipes), though there isn't a calendar as such beyond the weekly meal plan view.

Preferences are primarily handled via the AI (uses Gemini on the backend) and this is the primary driver in how recipes are constructed and recommended.

I made something similar to scratch my own itch. What I really wanted was to plan meals for the week with as little thinking as possible.

The result is a very interruptible grocery list process, where I know I haven’t forgotten anything.

https://supperstock.ca/

It would be great if the meals had macros per serving listed, total calories, protein, carbs (+ fiber), fat per serving.

Edit: I see there is a nutrition facts button.

Is there a button I can press that says "suggest meals with more protein per serving"?

You can specify this in "Preferences & Requests" in the meal planning wizard or "customize with instructions" on an existing meal plan recipe.

I want to thank you for telling us Kroger has an API. This excites me.

Albertsons (and also Safeway) has a kind of hidden API that I use to pull all my grocery purchases into a Postgres DB in order to see what's what.

https://bsky.app/profile/davidwelton.bsky.social/post/3m34al...

Is there a way to know when chuck roast is on sale?

Looking at the weekly ads is probably the easiest way. Not a part of the API I've looked at.

It's actually a pretty decent API. I've hit some quirks with it, but for the most part works reasonably well overall. The best part about it was there were no hoops to jump through to start using it.

I think I might have the wrong concept of meal planning. For me, it should be simple meals using fewer ingredients and most importantly reusing ingredients you already have. During the meal creation, I asked to keep it to 1 protein source, but it gave me recipes for 4 different protein types. I cannot possibly be buying salmon, chicken and 2 different cuts of meat each week. The other ingredients follow the same pattern.

I'm sorry you hit that as it should definitely respect instructions around things like protein sources and ingredient selection. I'll take a look at that.

Almost everything I eat is prepared from scratch by me at home. I cook mostly be feeling these days. I don't find rando recipes online that useful. They tend to take more time just because you're trying to work out what they're talking about only to find out after several precise instructions, oh, they're making a béchamel sauce, which I could make with my eyes closed.

What I've been working on for my own personal uses is a curated database that keeps track of ratings and numbers like perceived difficulty and actual time taken (these vary wildly from what they tell you). Then an algorithm that generates a plan based on given parameters for that week, mainly how much time you have and whether you want to try new things.

I also essentially rewrite all recipes in my own shorthand format which is based on how I think and execute the recipes in the kitchen. For example, I do mise en place for some things, but not others. I do chop all vegetables first, for example, but I don't pre-measure out all my spices because that's pointless.

Most off the shelf software has the wrong priorities for experienced cooks IMO. Like I don't need or care about the exact quantity required of tomatoes for the week. I just know this. I also don't need it to tell me I need e.g. turmeric. I cook mainly Indian food, I always have turmeric. I also haven't found one that understands meal prep or base ingredients that still need preparation (like cooked dal, dosa batter, masalas etc.)

My main problem is the analysis paralysis when deciding what to eat for the week. But using someone else's database for recipes is useless to me.

Nice, I think you did a nice job on the website and it'll be interesting to see how it goes. I use Claude for generating recipes and have started to curate my own recipe list for myself, nice idea to offer a more managed service for folks.

You can do that for free with Claude or ChatGPT or Gemini or grok or deepseek

How is this $10 / month?!

this is surreal

it imagines that all food ingredients are generic and interchangable, that there is no significant seasonal variability in variety and quality the harshest part is that all of these schemes start out with a loss leader special arangement with supliers that sets people up with higher quality products at great prices, that are soon swaped out for whatever mass product at as high a price as the market will bear, the whole thing designed to create predictability for retailers and drive profits up. grossmart with delivery, and a chatbot foodporning in your ear. blech

I’ve been wanting to build something similar using a combination of Mealie/Tandoor/Recipya, n8n/Node-RED, and the Kroger API for personal use. Has anyone else explored this setup?

Gonna try this out - I love the cook mode, the shopping list is helpful and that I can change out recipes for the days is great. I also like adding my own. Cool stuff

Let me know what you think or if there are areas that can be improved!

Once you get some recipes back this is pretty cool, but it took like 5 minutes to generate the recipes which doesn't super work for a sign up flow.

Sorry it’s a bit of a nitpick.

Not at all a nitpick, it is one of the things that is painful in the app IMO. I think the length of time for some of the operations is one of the bigger reasons for people to bounce after initial sign up. I need to think if there are some clever ways to improve the onboarding experience such as trying to pre-generate some recipes during/after the onboarding flow.

I think your sign up flow is just too long also, all that information might be necessary to get the app to function how you want but it feels more like filing your taxes. Maybe you could build an agent that the user talks to during sign up? And it will fill out the form in the background? Also you could definitely cache some high quality content (kind of like how TikTok has a core set of high quality videos it uses during new user onboarding)

This looks great. Was just talking about how an app with these exact features would really help me out. Will give it a try.

Please let me know what you think. I really value any and all feedback!

Cookbooks are already a thing. I am baffled that you think this is something we need AI for.

It's surprising how easy it is to stare at an entire bookshelf of cookbooks and not feel inspired. Ask me how I know.

I'm content to just keep healthy ingredients on hand and improvise, but I get that some people could want something with more structure.

How many people own a cookbook they haven't opened in a year? I think the value proposition is the same as Pandora: send me something I'll like that I wasn't expecting.

AI slop enjoyers want to delegate every single aspects of their lives to AI. For them cooking is an insult to their intellect, writing a personal message to someone is a waste of time, &c.

Why bother with cookbooks when you can cook the statistical average of spaghetti carbonara ?

Why bother spending 10 minutes of brain power when you can just pay 9.99 a month for a machine to tell you what to do ?

I get the skepticism as I shared similar thoughts on what recipe quality would be like. I'm also quite sensitive to AI slop. My wife, who really is an outstanding cook, was even more skeptical, and she has a _very_ high bar for recipes. To be clear, we don't use it to generate recipes for every meal. She has a collection of go-to favorites and then peppers in 1-2 AI recipes throughout the week, often to explore new things. She also uses it to create "Quick Bite" recipes for when we need something fast with whatever we have on hand, outside of an actual meal plan.

Honestly, I wasn't sure if the AI recipes were going to meet her standards when it comes to food, but so far she's been quite happy overall with the recipe quality. She often will request replacements or customizations, so the first pass suggestions are not always what she's looking for.

I realize it's definitely not for everyone though.

llooks nice, but also way too vibecoded. maybe fix the page before launching

Sorry, fix what page? I'm not sure I understand.

I wish you best of luck

Put pork as an allergy and first recipe suggested has bacon :)

Show HN: Meals You Love – AI-powered meal planning and grocery shopping