A Beginner’s Guide to Codex

[embed]

With the rapid acceleration of GenAI, there are now an overwhelming amount of tools that claim to make our lives easier. This guide will introduce Codex, a general purpose AI assistant built by OpenAI, to a non-technical audience.

Introduction to Codex

Codex is a general purpose GenAI assistant that completes tasks, creates files, and automates work from natural language instructions. Unlike traditional AI chatbots, Codex interacts directly with files and folders on a computer, enabling it to perform more complex and structured tasks.

Codex runs locally on a desktop and can interact with local files. This allows users to work with documents, spreadsheets, images, PDFs, and other files without manually copying and pasting information between applications.

Despite what the name implies, Codex is general-purpose and can be used for tasks beyond coding, such as:

• Organizing files and folders
• Summarizing documents and reports
• Analyzing spreadsheets and datasets
• Creating presentations and reports
• Planning trips and generating itineraries
• Managing personal finances
• Automating repetitive computer tasks
• Building simple tools and applications

FAQs about Codex:

What types of tasks can Codex do? Codex excels at tasks that involve multiple steps or require working across multiple files. Examples include:

• Combining information from several documents into a summary
• Cleaning and organizing large collections of files
• Analyzing spending patterns from bank exports
• Generating reports from spreadsheets
• Creating presentations from research materials
• Building custom tools tailored to a specific workflow

In general, Codex can automate tasks that involve several manual steps that are performed repeatedly on a computer.

Why use Codex instead of ChatGPT? ChatGPT and Codex perform similar tasks, but the key difference is that Codex has access to local folders and files on a computer. This enables Codex to directly work across files and create projects or analyses that would be difficult on ChatGPT.

In general, ChatGPT is better suited for ad-hoc tasks such as quick questions and image creation, while Codex is better suited for structured tasks such as conducting analyses or creating software.

How can Codex be accessed? Codex can be accessed through a desktop application available for Windows and macOS. More advanced users can also access Codex through an IDE (programming environment) or CLI (command line interface).

Setup

This guide will focus on setting up the Codex desktop app, users interested in accessing Codex via an IDE or CLI can find the associated instructions here (IDE link; CLI link) .

Installing Codex is quick and straight-forward and available on both Mac and Windows. Codex is currently unavailable for Linux, but interested users can sign up to be notified for the Linux version here.

Step 1: Download Installer

The button will say “Download for macOS” on Macbook

Navigate to the following link and download Codex for your operating system.

Step 2: Login to Codex

First screen upon opening Codex

Upon completing the download, Codex will prompt to login. There are two ways to login:

• ChatGPT account (recommended)
• OpenAI API key

It is recommended to sign up with your existing ChatGPT as the API key option is more limited in functionality.

Step 3: Run a Test Prompt

Codex main screen

There are two types of prompts that can be run in Codex:

• Chats: These are similar to conversations in ChatGPT. Chats are useful for asking questions, brainstorming ideas, or getting help with a specific task.
• Projects: Projects give Codex access to a folder on your computer and allow it to work directly with files. This is where Codex becomes especially powerful, as it can analyze documents, organize information, generate reports, and build solutions that span multiple files.

Both are accessed through the same prompt interface on the main screen, with the key difference being whether a folder is selected. Submitting a prompt without selecting a folder creates a Chat, while selecting a folder first creates a Project. The images below illustrate the difference:

Submitting a prompt here will create a Chat as no folder is currently selected

Selecting a folder such as “CodexExperimentation” creates a Project and gives Codex access to the files within that folder

Example Use Case: Job Application Tracker

In this example, we will use Codex to create a simple job application workspace. The workspace will include a browser-based tracker that can be opened directly on your computer, plus folders where Codex can save tailored resumes and cover letters for each role.

Step 1: Setup a Project Folder

Create a new folder on your computer called JobTracker.

Add a relevant resume to the folder for this project.

Note: If selecting “Start from scratch”, the project folder will be saved in “Documents” for both Windows and MacOS.

Step 2: Create a Codex Project

Screenshot of project creation

Navigate to the main chat interface screen on Codex and select “Add new project”.

Select “Use an existing folder” if a desired folder for the project already exists, otherwise select “Start from scratch”.

Step 3: Ask Codex to Build the Workspace

Once the Project is selected, enter the following prompt:

Create a simple job application workspace that I can use on my own computer. Please create a browser-based job application tracker that opens by double-clicking a file, without requiring me to install anything, use Terminal, or run setup commands.

The tracker should let me view and manually update my job applications. For each application, I want to track the company, role, job link, application date, status, recruiter/contact, follow-up date, and notes. It should also show a simple dashboard with counts by status and highlight applications that need follow-up.

I will place my base resume in this folder. I do not need the tracker itself to generate cover letters or resumes. Instead, I want to use Codex for that part. In the future, whenever I find a new job, I will paste the job link into Codex as a new prompt. Codex should then use the resume in the Resume folder and the job posting to create a tailored cover letter and tailored resume draft. Save those files in a new folder inside Applications, using the company and role name, and update the tracker with the relevant job details.

If the job link cannot be accessed, ask me to paste the job description directly. Do not invent any experience, skills, credentials, or work history. Only tailor the wording based on my actual resume.

Codex should create the files needed for the tracker and organize the workspace.

Step 4: Review the Created Files

Sample output of running the prompt

Upon completion, Codex will return the created files as well as instructions on how to run the program.

The results may vary, but in this case, the job tracker can be accessed at tracker.html (name may vary), with jobs of interest being added in future chats.

Sample output for job applications tracker (results may vary)

Apart from this primary file of interest, Codex has also set up the folder structure for future customized resumes and cover letters and instructions for future prompts. Your results may vary, but should roughly align with such structure.

Step 5: Test the Job Applications Workspace

The next step is to test to ensure the tool works as intended. To test this, a sample job link will be inputted into Codex continuing in the same chat, as shown below:

Sample input and output continuing from previous conversation (results may vary)

Once Codex is done, review the job application tracker and created cover letter to ensure this worked as intended. To access the tracker, open tracker.html (or whichever file Codex mentions contains the tracker).

Updated job application tracker

As can be seen above, the job application tracker updated successfully with the job provided to Codex.

The next thing to check is the produced cover letter and resume. Here, it is created in a dedicated folder for this application, though this may vary on different runs of Codex.

Generated cover letter for inputted job

Step 6: Iterate on the Job Applications Workspace

Based on this initial testing, further edits can be made to the workspace depending on the desired outputs. This is one of the biggest advantages of building with Codex: the first version does not need to be perfect.

For instance, if the tracker is missing a field, the layout is confusing, or the resume and cover letter folders are not organized as desired, the user can simply describe the change they want and ask Codex to update the project.

In the example shown below, Codex is asked to improve the layout of the tracker for easy editability.

Example prompt to improve job application tracker

Conclusion

Congrats, you have successfully used Codex to help with job applications! This guide is intended to showcase a small use case and Codex can be used for much more powerful tasks, even by non-technical users. As you get more comfortable, start looking for repetitive workflows in your own life that Codex could turn into simple, personalized tools.