Auto-Soylent Top 5 Remix


Over the last week I’ve been working on applying nonlinear optimization techniques to producing/optimizing Soylent recipes. ( The idea/process was discussed here: although my method doesn’t use genetic algorithms). Today I’m ready to unveil the first automatically generated Soylent recipe, which was made by running my algorithm on all of the ingredients from the top 5 most favorited recipes on the DIY Soylent site. So, without further ado here’s the Auto-Soylent Top 5 Remix:

The code I used to generate this is up on GitHub here (it’s built on the skeleton courtesy of @nickp ) :

It’s a work in progress with the ultimate goal to add functionality to the DIY soylent site to assist with optimizing ingredient amounts for nutrition and price. In the above recipe I put a lot of emphasis on lowering the price. The recipes this was built from range from $2.22 to $5.50, but the final result here is just over a dollar. However, this emphasis results in carbohydrates at 118%, which is obviously less than ideal. Tweaking some control variables will fix that but it’s for a price. I honestly don’t know much about nutrition, so I would really appreciate if someone more knowledgeable were to take a look at this recipe and see if it seems reasonable.

I’ll be generating quite a few more recipes as I work the kinks out of the application, so if you’d like me to try yours just post it here. There isn’t functionality to find ingredients, so any ingredients you want to try out have to already be in the recipe.

Also, while I have your attention, I’m an independent consultant with an artificial intelligence background looking for work. I run and this whole thing is basically a publicity stunt. Well, that and it’s fun. Ok, you may resume your regularly scheduled soylenting now.


A few comments.

Ignore sulphur entirely. If the proteins are correct, there is no need to supplement it. It has been erroneously left in the RDI schedule for historical reasons. There is much discussion on this forum about the topic.

Secondly, while being mathematically “correct”, nobody is going to attempt to take 1.21 pills or 0.55 pills. You need to constrain granularity to 1 unit for those items. I suppose you could take a pill every second day and get close to 0.55 pills but most people would find that too complicated .

Otherwise, wow. Wonder how it tastes.

EDIT: after thinking about it, I suppose a mortar and pestle to grind up a weeks worth of pills into powder and then weighing it out with scales would not be that hard. Maybe fractional pills would be ok.

Maybe you can run this one through the sausage grinder and see what it spits out:


Great first attempt. I don’t think it would be very palatable at this point, but for the price it’s a nice attempt. Flavorings would go a long way toward getting someone to continue using.

The abundance of carbs won’t help either. Hard for palatability and probably rough on digestion. You may consider restricting the macro percentages. 95-105% for each?

Second on removing the MSM, it’s unnecessary, save a few pennies.

Also, you may look into restricting the omega 6 and omega 3. Plus, many are getting away from the RDI in favor of balancing the 6/3 ratio, anywhere from a 1/1 to a 4/1 ratio.

What’s keeping the potassium from hitting the last 1%?

The taste would be hard to cover.


I ran that recipe through, and the changes are actually pretty dramatic.One of the supplements didn’t have a cost listed, so it wasn’t used, which probably screwed things up a bit. You can take a look for yourself and see if this is useful:

I removed sulfur as a targeted nutrient, fixed the macro-nutrients to be a little tighter, and updated my recipe. I’m looking into the ratio thing, but that’s going to take some work.


I have updated the recipe for the missing price

I must admit I mainly crafted it for taste and palatability - I feel the obsession to shave every last cent off the price a little counter productive at times, if understandable.

I’d be curious to see how much that changes the micronutrients balance with that price added.

The macronutrients mix it came up with is interesting, I’m a bit concerned about the amount of RPI I had reduced it to a minimum as it makes the drink very chalky. Rather than take it out altogether, I planned on just slowly using it up, until I can get some finer mesh RPI, or the official Soylent. Maybe try setting the maximum on that to its current value and see how it turns out.


Upon further inspection many of the ingredients in my previous blend did not factor in shipping costs or were generally impractical to obtain at the listed prices. I mixed US and European recipes which made for a theoretically cheap but practically impossible to produce recipe.

So I took another crack at it. This time I limited the ingredients to only things which could be purchased on US Amazon. I still pulled ingredients from other recipes, so I’m still relying on the nutrition information others have entered. However, this time I went to Amazon, checked the prices, and added in the shipping costs for any items that couldn’t get free shipping. I also took the liberty of rounding off the computer numbers to amounts that would be reasonable to measure. Unfortunately, the price for this more realistic recipe is about 5$ per day, but on the bright side you could actually go buy it at that price right now on Amazon.

@unsynchronized Running your recipe with the Fenugreek supplement priced didn’t really change anything. The optimizer took that completely out as well. Alas, I can’t optimize anything I can’t quantify (like taste or texture), but I’ll work on making the process more tweakable. I don’t have a mechanism for setting a min or max on an ingredient at the moment. I’ll look into that. I may not be able to get it to optimize taste, but I think I can let you fix taste-based ingredients and then optimize nutrition and price around them.