Smol AI tutorial in 5 mins | Build ENTIRE codebase with a single prompt

Key Takeaways

ah what is this ah tight tight tight yeah ah blue yellow pink whatever man just keep bringing me that every once in a while there's a project that made me feel we're just scrapping the surface with generative Ai and small AI is definitely one of those projects it got more than 8 000 stars on GitHub in just five days and it is one of the fattest growing open source AI purchase so what is small AI it is basically an AI that right codes for you it's different from asking chat topt to write some simple code snip it actually built out entire code base for even very complicated apps it's almost like having the whole engineer team in your pocket and here's quick example I ask gpd4 to write a product requirement DOC for a classic snake game and then pass on those requirements to the small Ai and ask it to build this app and boom I just got a snake game that is fully functional from small AI if I have more requirements I can just come back and add more stuff and I can ask it to build multiple things for example a simple calculator apps and it will also automatically build a app for me so so how does this small AI work when looking through the code base it's actually very straightforward first they pass the product requirement into a prompt they will generate architecture of the apps to what are different files we need to create and then it will pass this list of files as well as original product requirement to another prompt that will generate a list of dependencies so that GPD can understand what are the different functions and variables in different files and in the end it will run this for Loop where it will pass on both the product requirements architecture and the dependency to a new prompt that will actually write the code for each individual file that's pretty much it but to be honest at this moment it's still quite buggy most of the time you will have some bugs or generate some code that require modification to actually run properly and I'm going to show you how can you run small AI without those box step by step firstly let's set it up we can go to small AI GitHub which I post Link in the description and your first idea clone this Ripple on your local machine and then your open folder in your visual studio and there should be a DOT example.env file you'll rename this to VMV exporting your open AI API key here and before you actually run small AI this is one last step you should do is install model modal is basically a cloud computing platform where it's gonna run small AI once that's done it's pretty much all set you can open the terminal and run this command line model and prop after that you can add your own prompt like write a web app for to do but on the other side what I would recommend you to do is actually create an MD file which is basically a documentation and put all the product requirements here so that it'll be easier for you to update the requirement and you can just run prd.md if you just start using small AI like this quite often you will see some arrows and those arrows are meaning due to the rate limit of open AI I think that's because when small AI is running it will call open AI API multiple times simultaneously that's why it exceeds the rate limit but thanks to another YouTuber called Anubis AI he actually modified the main.p code and add some delay between every API call so it won't trigger this Arrow I'll put a modified code in the description below so that you can copy and paste in here just override the whole main.py file now we finish the setup the next step is get the product requirement you can write simple product requirement by yourself but to make sure small AI actually produce the results you want you actually need to add a lot more details to it what I normally do is I will go to chatgpt and ask gpd4 to help me write down the whole product requirement doc which including the core functionalities interface descriptions as well as the file structure and I can just copy those requirements put into this MD file that we just created and then run this model around ming.py prompt prd.md you can see here it already identified what kind of files they need to create as well as break down all the shared dependencies including the variables data schema Dom element and function names this whole process is going to take a while for some complicated apps it can take like 15 to even 40 minutes but here is also another tips I want to share what I noticed that at this stage is still pretty buggy in terms of making sure the function writing in the file a is also compatible with file B so for 90 of time for the whole code base it generated it didn't work out a box however if I asked it to write everything into just one file then it'll work magically like 80 to 90 of time the code I generate it just work out of box so in the requirement I will often say that it should only have one file that include everything and that's exactly what I did here so now since it is finished the file is generated will be under this generated folder if we run this it is a snake game out of box it has all the features that keep a snake move and it has a function to generate food and also have a way to keep a score and the last thing I want to mention is that they not only provides a way to generate a code they actually provide a way to debug the code as well so if I add another window here I can run that command line called Moto run debugger.py at the prompt here's where I can add any Arrow I saw when I'm running the code and the same thing I can put the arrows here in prompt directly but I can also create debug.md and then you can add the arrows you saw and we can run this model.round debugger py prompt debug.md thank you and then you can see it will start giving me analysis of the potential arrows and how can I fix it so it's like having both Engineers who generate code and also a debugger at the same time that's pretty cool and this is small AI developer as I mentioned it is a little bit buggy it's not capable to produce a very complicated apps out of box yet so I don't think it's going to replace a real engineer with its current format however as a product designer and product manager I'm really excited about this because if I want to user test a certain product ideas I can ask it to build a prototype very very quickly and tasks with real users for almost zero cost but on the other side you can think about how this whole ecosystem start composed on top of each other we already see AI that can design the user interface and it's very easy for AI to produce product requirement Doc and do research so I think it's very possible that in future we can have the whole AI product team that including both product manager product designer engineer QA out of box and care around lots of different experiments for you autonomously so I'm very excited to see what other interesting apps that you asked small AI to do and comment below for any ideas you have thank you and see you next time