Welcome

Below you’ll find written documentation and video demos of all of the commands/shortcuts/functions DATA has and how to use them.

If you prefer the accessible bite-sized version of this page, check out DATA Wiki

This page is super long and detailed and totally optional!

You shouldn’t need to read or watch any of this to get the most out of DATA, DATA should just walk you through different actions it can take and talk to you naturally about what it can do.

But for those of you who love to dive deep into how things work, here are the resources to do that. The purpose of this page is so that you can know everything I know about how DATA works so you can make informed choices about how to use it.

This page in combination with Building Your Own Commands will allow you to fully understand how DATA works & build anything with it.

You can use the search icon in the corner to skip to any part of this page.

Or you can drop down this table of contents and click anywhere to skip to the section you’re interested in.

Welcome

DATA is a tool that is meant for you to get into the prompts and edit them to your liking. You can duplicate ANY of the shortcuts on this page and make them your own, and that is intended. I expect you have your own taste and way of doing work, and DATA makes it easy for you to program software with just your words. The reason I’ve gone into detail about how they work here is so that you can easily audit what parts of them you’d like to change and jump right in.

DATA as a power tool, here’s how to know when to use easy mode

The most important thing to know about DATA is that it depends on AI products from other companies to work. DATA is like a car — you sort of have to choose what quality of gas you put in the engine, and that will determine how efficiently it runs.

GPT-3 & other affordable models require TIME and testing to get them working reliably for each use case. I’ve tried to do a lot of that work for you, but inevitably when you decide you want to use DATA for something unique, it will require effort on your part to test the outputs, evaluate how DATA gets them wrong, and then re-write the prompts to get GPT-3 to do something new.

Whereas GPT-4 and smarter models like Claude just use money instead and usually get very detailed instructions right on the first try. So if you’re prototyping something new or you’re asking DATA to do something unique, it’s easier if you start with GPT-4 and then tune from there.

I expect over the next year AI models will come out that easily get it right on the first try very affordably, but we don’t quite have them yet.

So if you like to tinker and play around, use “gpt-3.5-turbo-16k” in the “model” field in DATA settings.

If you want to get some work done and you just want DATA to be brilliant every time, use “gpt-4”