Cheat Layer
Automate Your Business Using Machine Learning In The Browser
什麼是Cheat Layer？
Cheat Layer是由https://cheatlayer.com開發的Chrome擴展程式，該擴展的主要功能是“Automate Your Business Using Machine Learning In The Browser”。
擴展截圖
下載Cheat Layer擴展crx文件
下載Cheat Layer擴展crx格式的文件，手動將Chrome擴充功能安裝到瀏覽器中，也可以將crx文件分享給朋友，輕鬆安裝Chrome擴充功能。
擴展使用說明
Sign up at cheatlayer.com
Tutorials: https://www.youtube.com/cheatlayer
Twitter: https://twitter.com/CheatLayer
For the past year, we've been working towards the "Holy Grail" for automation, and we're excited to finally reveal it! 💪
During the pandemic, I donated my time to help people build online businesses. People started referring me, and it quickly grew overwhelming as I was helping strangers globally. 🌍
Around the same time, I got early access to GPT-3, so I figured I could train everything I was doing to scale up my efforts to a "personal software engineer" in language models. 💻
In October 2021, we were the first to get approved by openAI to sell GPT-3 for code generation, after months of developing security mitigation features to gain access, while the only other at the time was Microsoft with Codex. 🙌
The Holy Grail
We set out to solve that "Holy Grail," and it took a year to develop the framework that can finally solve it. We call it Project Atlas. 🗺️
Project Atlas can be trained to build entire automations from end-to-end of any complexity. We spent a year+ solving all the road blocks in automation, including desktop automations, cross origin iframes, changing selectors, uploading DALLE images, etc, and distilling this knowledge into GPT-3. This was all built on openAI Codex until 3 weeks ago, when we finally got access to the chatGPT API and GPT-4. 🔧
The "Android" To ChatGPT Plugins
"Plugins" seem like a step backwards related to a frame of reference starting with mobile apps, a vestige of an era that will pass in our opinion. It's possible for this tech to build the plugin/app for you instantly on-demand for new problems, and clearly that's the future that solves more problems
Our language-based automation format and library immediately enable each new language model to solve the "last mile" and execute, allowing users to build new "plugins" using entirely language alone. We enable everything possible manually across all websites and desktop programs, including automations that you won't find in chatGPT plugins, so users can discover and build using simple language rather than python coding.
Community-Discovered Cheat Codes
If Project Atlas doesn't yet natively support an automation, chatGPT actually attempts to build it based on context live! We taught chatGPT our internal library that allows sending data, linking APIs, opening tabs, and controlling the desktops across Mac/Windows/Linux and Chrome browsers. This enables users to discover new apps and automations using entirely natural language, while before our bottleneck was our team and freelance software engineers. We have live hackathons, free prompt classes, and an active community contribute to building our library and defensive moat. Join our active community of 5000+ global users building together!
Automations To Branded Tools Instantly
If the custom automations you discover save you time, they're potentially valuable to others as products. We're the first and only service that allows turning no-code automations into branded products instantly. You can even use chatGPT+Project Atlas to design your front end. Check out this tutorial on launching your own chatGPT apps in 5 minutes for free: https://youtu.be/NGlfGRpkd0Q
Forget Apps and Plugins
Rather than waiting for your services to build the features you need, or waiting for someone to build an app to solve your problem, access software engineering agents which build custom versions of services tailored to your needs.
Software engineering agents can build entire apps, frontends, backends, marketplaces, cryptocurrencies, games, and even chatGPT-powered products. We're training these agents like employees now.
We believe in the near future, as we train agents and the context window increases complexity, everyone will have a personal software engineer building solutions on-demand. Agents could use all the same underlying tech to clone popular services with custom features, and it would cost you a fraction of the larger service. Large services often can't build every feature for every user, and need to focus on what the majority needs. Join our community to help us build this future by leveraging your deep industry insight to share useful agents.
Not Another autoGPT Clone
If you tried and gave up on agents already, Project Atlas is a completely different no-code solution and not a wrapper to an open source project.
A year ago we were the only startup using GPT-3 to generate automations, and we published our agents design to facebook several months ago. We were the first startup to get approved to sell GPT-3 for code generation to automate tasks in summer 2021 by openAI.
We don't know if we were the first to start building agents, and Adept.ai likely started around the same time we did, but we've been working on the hard road blocks for over a year with thousands of global customers. What's important is the root problem had not been solved well yet, and even finishing 4th, 5th, etc. is worth billions.
Not Your Daddy's Agents
Project Atlas agents are self-healing, and can re-generate failed tasks to try again until they succeed. Agents can confirm when tasks generate expected results and then store the working code in a vector database. This allows Project Atlas to share a memory that constantly improves as users test it.
Our agents solve unique problems you can't find elsewhere, and we present it in a project management interface. A project management interface is the ideal way to manage and iterate functional agents in parallel to accomplish real work as your own AI team.
Our agents can scrape unlimited data to google sheets, perform scientific research, send emails, automate your browser directly, generate images, generate social posts, simulate keypresses, pull in any javascript CDN library, generate and download files, use any API up till 2021, execute arbitrary javascript, and even build chatGPT-powered products. We've trained it like an employee over the past few months.
Project Management interface
Our unique project management interface enables managing agents like a team. You can edit tasks, re-order priority, and give agents additional context in real time as they execute in parallel. The system emails you updates and results. Build your own custom AI teams using entirely simple language and your own deep industry insight.
Community
Since we designed Project Atlas to grant users full access to all building blocks, our "plugins" are discovered in simple language rather than in python coding. Join our mostly non-technical community building and discovering agents together in simple phrases. Our users distill deep industry insight into custom agents you can't find elsewhere.
Introducing Atlas-1
Atlas-1 is a new multi-modal model we’ve been training over a year–it enables LLMs like GPT-4 Vision to perform precise actions and build robust, future-proof, autonomous agents.
Semantic Targets
Semantic targets are the key that gives LLMs like GPT-4 vision the ability to execute precise actions. Atlas-1 was trained to detect UI elements that are likely to match the intent of the next target to enable “Semantic targets,” which distill targets down to their underlying intent in language rather than code.
This solves a big problem that breaks all other automation tools today, including UIPath and Microsoft, when services inevitably change their code or designs. Previously, automation tools used targeting strategies like CSS selectors, Xpath, and even computer vision, but all of these fail when services update their design.Since semantic targets still translate potential targets the same even if you completely change the design of the website, this finally enables building robust future-proof automations.
Since these semantic targets operate in language, this also enables LLMs like GPT-4 to directly use them to perform precise actions.
Product Agents
We built our entire agents dashboard using “Live mode,” which enables building any complexity of product step-by-step like speaking to an engineer. Desktop Live Mode enables generating the backend as well as front end and hosting the entire service locally on your PC. Users have generated Chrome Extensions, Ios Apps, Android Apps, Cryptocurrency contracts, games, and AI Saas services.
Atlas-1 takes Live mode on Desktop a step further and actually visually checks the output to iterate until it matches the intended goal.
Marketing Agents
The goal of most social networks is to get users to stay on their app to sell more ads. They accomplish this using algorithms that have gotten very good at finding good content and matching it with an audience for you.
Our marketing agents scrape your profile of previous posts(or competitor's posts) to automate a/b testing and generate video/images/text that are highly likely to get the most engagement based on your history. By scheduling this agent daily, it can constantly iterate better content and drive traffic to grow any brand on autopilot.
Twitter Agent: https://youtu.be/4LtetnKz0kI
TikTok Agent: https://youtu.be/XOEWjR52mOM
Sales Agents
Our sales agents are getting 2X the reply rates vs Apollo.io with personalized conversations and instant responses. Deploy sales agents that learn what is high converting copy over time across Email, SMS, Embedded Chat with Voice, and Phone. We’ve had leads replying to agents 5+ times and making deals entirely without input.
Similar phone agent services charge $5k upfront and $16/hour for worse technology, but all subscribers can easily set up their own GPT-4 powered phone sales agent servers with twilio for $.01/minute.
GPT-4 phone sales agent: https://youtu.be/3aNxyjasRRM
Desktop Cloud Agents
Scale up operations by cloning your marketing and sales teams across Desktop Cloud Instances in parallel. Stream them directly to your agents management dashboard to manage your teams of AI agents.
We’re releasing our Mac/Linux build in the coming months, but until then all users can access Desktop Cloud Agents directly through our agents dashboard. 擴展基本資訊
| 名稱 |  Cheat Layer |
| ID | oolmdenpaebkcokkccakmlmhcpnogalc |
| 官方網址 | https://chromewebstore.google.com/detail/cheat-layer/oolmdenpaebkcokkccakmlmhcpnogalc |
| 簡介 | Automate Your Business Using Machine Learning In The Browser |
| 檔案大小 | 18.85 MB |
| 安裝次數 | 27,316 |
| 目前版本 | 12.9.7 |
| 更新時間 | 2024-03-03 |
| 上架時間 | 2020-07-25 |
| 評分 | 4.74/5 共 47 次評分 |
| 開發者 | https://cheatlayer.com |
| 電子郵箱 | [email protected] |
| 付費類型 | free |
| 擴展官網 | https://cheatlayer.com |
| 說明頁面URL | https://cheatlayer.com/billing |
| 隱私政策頁面URL | https://instoo.com/eula.html |
| 支援的語言 | en-US |
| manifest.json | |
{
"update_url": "https:\/\/clients2.google.com\/service\/update2\/crx",
"manifest_version": 2,
"name": "Cheat Layer",
"description": "Automate Your Business Using Machine Learning In The Browser",
"version": "12.9.7",
"icons": {
"512": "icon.png"
},
"browser_action": {
"default_icon": "icon.png"
},
"key": "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAgXQD4PZTCoKYKac3GGptPJsUjeDqxgQQGYVt+AOUvlCzvNb1sAy1DiZQpQXT9FXlMDNdBbfUux0zvgACwqbT6GwsyF+Mb0Tzg9Nuj9Ae6r2O3nI6ltilgF4C5k25SVs9y\/2hu8Df5+WgqMB2aVbk\/+vMDn5hDQ8rbP4vu57UVYKRuOdkEdS1RxaJfmYAAmqZ1wBniaJccYoIJ8VtCjev7aXCSvzVNHoBo0C+Kx8r0HouabC4a8pUK+cz6D\/hOeZENPDvWkLdKGR10Zw1g2we4Kw4u7x1aagvduZSAjWvbaw\/jSqPyPMOYz7B2iLzoGrJxorfaNWp08\/Ql1j2pMjnVwIDAQAB",
"content_security_policy": "worker-src blob:;script-src script-src *.jsdelivr.net\/ *.unpkg.com\/ *.google.com *.instagram.com 'self' 'unsafe-eval' https:\/\/jsdelivr.net https:\/\/unpkg.com https:\/\/apis.google.com https:\/\/cheatlayer.com\/ https:\/\/www.gstatic.com\/ https:\/\/*.firebaseio.com https:\/\/www.googleapis.com; object-src 'self'",
"content_scripts": [
{
"matches": [
"http:\/\/*\/*",
"https:\/\/*\/*"
],
"css": [
"bootstrap-grid.min.css",
"style.css",
"css\/jquery-ui.min.css",
"css\/jquery-ui.structure.css",
"css\/jquery-ui.theme.css",
"css\/style.css"
],
"js": [
"progressbar.js",
"supabase.js",
"cronstrue.min.js",
"pdf.min.js",
"ml5.min.js",
"chart.js",
"HtmlSanitizer.js",
"FileSaver.min.js",
"exceljs.min.js",
"jquery-3.5.1.min.js",
"js\/jquery-ui.min.js",
"html2canvas.min.js",
"firebase-app-compat.min.js",
"firebase-database-compat.min.js",
"popper.min.js",
"tippy.all.min.js",
"intro.min.js",
"js\/main.js",
"js\/custom-picker.js",
"leader-line.min.js",
"interact.min.js",
"screenshot.js",
"ace.js",
"theme-tomorrow.js",
"mode-javascript.js",
"ext-language_tools.js",
"cheat.js"
]
}
],
"commands": {
"open-atlas": {
"suggested_key": "Alt+S",
"description": "Run Project Atlas on any website"
}
},
"background": {
"scripts": [
"jquery-3.5.1.min.js",
"peer.js",
"supabase.js",
"sse.js",
"pdf.min.js",
"html2canvas.min.js",
"tesseract.min.js",
"firebase-app-compat.min.js",
"firebase-database-compat.min.js",
"background.js",
"gapi-client.js"
]
},
"oauth2": {
"client_id": "839545279735-0b4c3gs6sgds88brcthb7itebp4b8gf4.apps.googleusercontent.com",
"scopes": [
"https:\/\/www.googleapis.com\/auth\/spreadsheets",
"https:\/\/www.googleapis.com\/auth\/spreadsheets.readonly",
"https:\/\/www.googleapis.com\/auth\/documents",
"https:\/\/www.googleapis.com\/auth\/documents.readonly"
]
},
"web_accessible_resources": [
"img\/l.svg",
"img\/avatar.png",
"atlas.gif",
"openai.png",
"gmail.png",
"excel.png",
"filter.png",
"rss.png",
"logo.png",
"logo.gif",
"folder.png",
"backroot.png",
"redx.png",
"pdf.png"
],
"permissions": [
"downloads",
"debugger",
"webNavigation",
"*:\/\/*.twitter.com\/*",
"*:\/\/*.facebook.com\/*",
"*:\/\/*.instagram.com\/*",
"*:\/\/*.cloudflare.com\/*",
"*:\/\/*.cheatlayer.com\/*",
"identity",
"https:\/\/docs.google.com\/*",
"https:\/\/docs.google.com\/spreadsheets\/",
"https:\/\/*.googleapis.com\/*",
"tabs",
"*:\/\/*\/*",
"storage",
"https:\/\/instoo.com\/"
]
} | |
