Put Your AI on Autopilot
Your AI agent shouldn't just chat; it should do. Granting Codex full control over your browser and computer unlocks genuine hands-free automation. Imagine delegating tedious tasks: it can scour the web for competitor research, dutifully enter data into spreadsheets, or even negotiate with customer service bots on your behalf. This frees you from manual, repetitive work, and Codex also organizes your files, deletes bloat, and clears memory, keeping your machine running smoothly without your intervention.
Automate daily routines with Scheduled Tasks. This feature allows Codex to perform recurring actions on a set schedule. Need to automate:
- Summarizing unread emails
- Reviewing project logs for errors
- Checking website health
For these routine, lighter tasks, leverage the efficient Luna model. It completes actions quickly without draining your processing quota, ideal for consistent, low-overhead automation.
Further extend Codex's native power with Plugins. These integrations connect your AI directly to the tools you already use, granting access to platforms like:
- Google Drive
- GitHub
- Notion
Codex can then work directly with your files and data, eliminating guesswork, preventing broken workflows, and making it a true digital assistant within your existing ecosystem.
Stop Typing, Start Commanding
Typing to your AI feels clunky once you experience Voice Mode. ChatGPT Codex now lets you manage multiple agents and threads simply by talking. Instead of clicking and typing, you can fluidly spin up new tasks, switch between projects, and get updates, all through natural voice commands. Imagine telling your AI, "Okay, spin up a new thread and write a poem that's a hundred words long," and watching it instantly comply, freeing your hands for other work. It’s an incredible leap in hands-free interaction, making your AI truly conversational.
Stop wasting time re-typing complex instructions. You can build a library of custom Skills for your most common, multi-step workflows. If you find yourself repeatedly entering the same elaborate prompt, save it. Then, invoke it instantly with a simple / command to execute intricate processes in seconds. This feature transforms repetitive prompting into a single, efficient action, ensuring consistency and significantly cutting down on manual input for frequent tasks.
Your sidebar can quickly become a chaotic mess of conversations. Tame that digital clutter by using Pinning. This simple organization hack lets you pin your most critical or active chats right to the top of your sidebar. No more endlessly scrolling through history to find that crucial project. Pinning ensures high-priority work remains visible and just one click away, keeping you focused and preventing important projects from getting buried and forgotten.
Execute and Ship Complex Goals
For truly ambitious projects, the standard chat interface often falls short. That's where the /goal command comes in. This advanced feature directs an agent to work relentlessly towards a single, measurable outcome over hours or even days. Imagine instructing your AI to "increase website speed by 50%," then letting it run autonomously until the task is complete. It frees you from constant oversight on complex, long-running objectives.
Need to share your work instantly? ChatGPT Sites offers an incredibly convenient solution. Publish almost anything to the public web directly from your chat. Turn a simple poem, a detailed spreadsheet, or an entire professional portfolio into a live, shareable URL in moments. The AI handles all the design and hosting, making it effortless to distribute your content to collaborators or even other AI agents. This capability significantly streamlines your publishing workflow.
Combining these features unlocks end-to-end project completion. Start by deploying Browser Control to research a niche topic, then leverage a powerful model like Sol to write a comprehensive report. Finally, use ChatGPT Sites to publish that report as a polished web page. This unified approach transforms fragmented tasks into a seamless, cohesive workflow, showcasing how deeply integrated these tools are. To learn more about the platform's foundation, read Introducing ChatGPT - OpenAI.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
The Pro's Resource Playbook
Don't just throw compute power at every task and hope for the best. Smart users practice strategic model selection to maximize efficiency and save quota. For 80% of your daily grind—think formatting documents, summarizing long articles, or simple text edits—the fast, inexpensive Luna model is your go-to. It handles routine tasks without burning through resources.
Reserve the powerful, quota-intensive Sol model for the 20% that demands deep reasoning and complex generation. If you're building a website from scratch or tackling intricate data analysis, Sol is your powerhouse. Using Luna for fine-tuning a website’s fonts or colors, after Sol built the core, illustrates this balance perfectly.
Treat your weekly usage quota like a precious resource. Actively manage it by regularly checking the usage and billing dashboard. This shows precisely how much compute power you have left and when your allocation resets. Don't let valuable compute go to waste, or worse, hit a wall mid-project.
Finally, keep an eye out for "banked resets." The platform often rewards active users with these temporary quota boosts. Use them strategically to power through heavy workloads, especially when you've hit your weekly limit but still have days before the official reset. These extra allowances can be a lifesaver for urgent, demanding projects.
Frequently Asked Questions
What's the difference between the Codex Luna and Sol models?
Luna is a fast, cost-effective model ideal for simple, routine tasks like text formatting or email summaries. Sol is a larger, more powerful model designed for complex, high-stakes generation and reasoning, such as building an entire website from scratch.
How does the /goal command work in Codex?
The /goal command instructs a Codex agent to run continuously until it achieves a specific, measurable objective. You can set a verifiable target (e.g., 'improve website speed by 50%') or let the LLM be the judge (e.g., 'make the website as fast as possible').
Can Codex control my computer?
Yes, with user permission, Codex can perform tasks on your local machine. This includes organizing files, deleting bloatware, and managing memory to improve computer performance, extending its capabilities beyond the browser.
What are Codex Skills?
Skills are reusable custom prompts or workflows. You can create a Skill from any successful chat thread and invoke it later by typing / followed by the skill name, saving you from re-typing complex instructions for common tasks.