This is a repost from our old blog originally published Dec 27, 2022: https://web.archive.org/web/20221228111840/https://blog.cheatlayer.com/language-as-a-universal-interface-e54ec841e273?gi=711cde8021e1

When we first launched Cheat Layer, the original vision was to use GPT-3 to build the entire automation from end to end. Users may remember this early version which had a kernal of that original vision.

GPT-3 at the time wasn’t powerful enough to scale that to all automations, but we’ve finally built a framework which can do it using the latest Codex and GPT-3. After a year of learning all the automation road-blocks, by meeting with hundreds of users personally, we’ve trained Project Atlas to understand how to solve all the hardest technical limitations.

We’re finally nearing that original vision with the latest Project Atlas update:

Project Atlas now builds the whole automation for you in the background, just like having a conversation with our office hours staff.

New Features

• Schedule automations using natural language, for example “post motivational quotes to twitter everyday at 10am”

• Automate posting Dalle images to social platforms. Next, we can extend this to a/b test the content generation against actual engagement rates and grow brands on autopilot.
• Auto-posting social posts(Facebook, Twitter, Instagram) text using GPT-3 davinci.

• Automated blog posts to wordpress using GPT-3.
• Our own internal sales team automations, including completely automating Apollo.io. We’ve set it up such that our sales team constantly gets demos filling into their calendars on auto-pilot.

• General navigation like clicks, scrolls, and search. For example, “Go to amazon and search for greenies” or “Go to google and search for news”
• General automations like scraping emails/phones/links/text/html/meta data/images/video and filling out forms. These general automations work on all websites, immediately supporting millions of websites out of the box.
• Desktop automations like opening programs, and controlling the mouse/keyboard between programs.

Automation Road Blocks

As you can see in the videos, literally all the technical road-blocks are down now after a year+ working on all of them, and we can build basically anything next with the only limitations being our imaginations. These include the browser security sandbox, cross origin iframes, and automating outside the browser.

The coolest, most striking, part for me is that the interface is the same for everything — language, so one day even super complex APIs or any other website can use this same interface.

Users no longer have to learn another interface again, which reduces complexity and brain fog. A simple universal interface likely helps you reduce aging itself, and it seems like “language” is the intuitive best choice for this interface. That’s just the beginning, and now we’ll start releasing all the new user-trained automations daily.

Intuitively, it makes sense that language would work to solve this problem, since it’s the same interface we use with office hours staff. This is how we know Project Atlas can eventually replace office hours as your personal software engineer for automation.

We can now scale up training across freelancers in big batches using these videos to accelerate what Project Atlas can do. This framework is everything we need to train even the most complex automations, and over time our unique data set will be our defensible moat.

Help Us Build The Future of Automation

Project Atlas is still in Beta, but users have access to help us iterate and train before the launch next month.

If you believe in this vision, then you understand very soon most people won’t be building their own automations. The market of people who would never consider building is much larger, worth billions per year in the US alone, and Project Atlas allows us to reach them first.

This framework finally allows users to help us accelerate training now.

Our big advantage over competitors, who have raised millions, is we’ve had thousands of users globally helping us train and build Project Atlas for the last year. We’re finally nearing the end of this race worth billions in B2B contracts(even if we win 2nd), and our solution is currently ahead of the competition now.

Next month, we’re holding a Hackathon to finally launch Project Atlas to the world. We’re flying users out from our Discord as well to make this Hackathon event special, including from other countries.

You can help us make the launch epic and earn free codes by helping us train Project Atlas.

How to help:

1. Record a screen capture video of you MANUALLY performing an action from end to end without automation at all. It can be as complex as you want across multiple programs. Users can earn up to 3 free stacked codes doing this 3 times right now.
2. Submit the original video file (not links) to support@cheatlayer.com
3. Choose automations which are valuable to your business and on this bounty list:

Cheat Layer Bounty List

Sheet1 academia.edu accuweather.com addthis.com adf.ly adobe.com adp.com air…

docs.google.com

Here’s a great example of a submission:

ZGi Brands

Watch this video I made for you

sendspark.com

We can train basically anything next, so please point us in the most valuable directions and help us finish this race.