01What is GPT 5.6 and How It Works

GPT 5.6 is the latest and most intelligent AI model from OpenAI, released on July 9. It offers three versions: Sol (the flagship, most powerful), Terra (the balanced workhorse), and Luna (fast and cheap for high-volume tasks). You can access it at chatgpt.com by selecting the model from a dropdown menu.

In this episode, Julian Goldie explains that GPT 5.6 can handle chat, create images, do deep research, and even control your computer when used with Codex. It works by reasoning through tasks, and you can see its thought process on the side. You can also choose effort levels—pro, extra high, high, medium, instant—to control speed and quality.

02Key Features and New Improvements

GPT 5.6 brings several key features:

  • Three models: Sol, Terra, and Luna, each optimized for different tasks.
  • Effort levels: Adjust thinking effort from pro to instant to manage speed and token usage.
  • Web access: It can search the web for the latest research to include in your outputs.
  • Image generation: Uses ChatGPT Image 2.0 but with GPT 5.6 reasoning for better results.
  • Deep research: Creates detailed reports with source citations after planning.
  • Projects and custom GPTs: Organize files and instructions for repeated use.
  • Voice mode: Improved, with smarter and faster responses.
  • Plugins: Connect to tools like Google Drive to automate real workflows.

Julian notes that even Luna, the lowest-spec model, outperforms GPT 5.5 in benchmarks.

03How to Set Up GPT 5.6

  1. Go to chatgpt.com and sign in or create an account.
  2. Click the model dropdown at the top of the chat window.
  3. Select GPT 5.6 from the list. You can also choose Sol, Terra, or Luna depending on your need.
  4. Pick an effort level: pro, extra high, high, medium, or instant. For most tasks, medium is a good starting point.
  5. If you want to use Codex for advanced tasks, download and install the Codex app from the provided link.
  6. In Codex, choose between local or cloud mode, and select your access permission (ask for approval, approve for me, or full access).

Once set up, you can start chatting or creating projects. Julian recommends using extra high for demanding tasks like coding, as pro can be slow (sometimes 10-20 minutes per response).

04How to Use GPT 5.6 for Real Tasks

Start with simple chat for quick questions. For content creation, paste your best-performing X (Twitter) posts into the chat, then prompt: "Create an X post about [topic] in a similar style." GPT 5.6 will analyze your style and generate a post automatically. You can edit or ask for changes.

For images, use a detailed prompt like the one Julian shows (e.g., "create an image of a cat eating cheese" with added style words). You can generate multiple variations with slightly different colors or layouts. To use deep research, type "deep research" and a topic like "best methods for AI automation." It will create a plan you can edit before it starts, then run for 10-15 minutes, searching many sources.

For building, use Codex. In the chat, describe your project, e.g., "create a beautiful website for the AI Profit Boardroom." Codex will plan, code, and save files locally. You can also use Codex to create videos by installing skills like Remotion from GitHub.

05Results and Comparisons from the Episode

Julian compared GPT 5.6 with rival models like Claude's Fable 5 and Google's Gemini. He says Fable 5 has been considered the best, but GPT 5.6 Sol is right up there, almost on par. He notes Gemini is "way behind."

On benchmarks, Sol performs best among the GPT 5.6 lineup, with Terra close and Luna lower but still better than GPT 5.5. Julian warns that benchmarks and marketing hype often differ from real-world results, so he advises testing for yourself.

In his own tests, he found GPT 5.6 Sol particularly strong for coding and content creation. He used Codex to build an entire website for his community, a habit tracker, and even a video using Remotion—all without being a coder. The video looked better than what his editors could create.

He also automated a blog post with Google Drive, which created a polished doc with SEO structure in minutes.

06Tips, Limits, and Mistakes to Avoid

Julian shares practical advice:

  • Avoid pro effort for everyday tasks; it’s slow. Extra high is his preference for the best balance.
  • Don’t waste time re-prompting. Use projects and custom GPTs to store instructions and files.
  • Monitor token usage. Higher effort levels and faster responses eat tokens quickly; check your usage often.
  • Always test outputs. Even with AI, the first version is rarely perfect—iterate daily.
  • Use approvals in Codex if you’re worried about the AI accessing your files or browser; you control permissions.
  • Keep prompts fresh. Every time the AI makes a mistake, update your prompt and save it (e.g., with text expander) for better results later.

Julian also notes that deep research isn’t useful for everything—only for one-off reports needing many sources.

07Who GPT 5.6 Is For and Why It Matters

GPT 5.6 is for anyone looking to automate tasks, create content, build websites, or learn AI skills. Whether you’re a beginner or advanced, you can start with simple chat and progress to complex Codex projects.

Julian emphasizes that AI can replace marketing teams, designers, developers, and video editors. He’s reduced his own team, replacing menial admin tasks with AI that never sleeps, doesn’t complain, and works 24/7.

In 2026, using AI effectively will be the most important skill. By building with GPT 5.6, you can solve real problems for yourself or your customers. Julian’s advice: analyze your frustrations and start building solutions with Codex.

He gives examples from his community—lead generation, video creation, and memory systems—all built with AI.

The future favors those who implement AI now. The idea is everything, and GPT 5.6 gives you the power to turn ideas into reality without coding.

08Advanced Workflows with Codex and Plugins

Codex takes GPT 5.6 beyond chat. You can create projects, build websites, run tasks in the background, and even control your computer with computer use. For example, you can ask Codex to go to google.com and type "hello"—it does it for you.

Plugins connect to apps like Google Drive. You can schedule tasks in Codex, such as publishing a blog post daily for SEO. Codex can edit files locally, code, and run 24/7 as an AI agent.

Skills from GitHub, like Remotion, teach Codex to create videos. You can even add your AI avatar from platforms like HeyGen for voiceover. Julian shows a fully automated video he made this way.

To publish websites, Codex can host on Netlify with your personal access token, even deploying to custom domains. For most people, starting with local mode and medium effort is enough; save time by using ready-made templates.

09How to Get the Most Out of GPT 5.6 in Daily Use

Pin important chats so you can return to them later. Use the pet icon in Codex to see background tasks and navigate back to conversations. Leverage the library to store files and images.

Create custom GPTs for specific purposes, like social media posting. You can train them with your own content, add web search and image generation, then share them with a link or on the GPT store.

Use scheduled tasks to have the AI do daily work, like deep research at 2:45 AM, so it’s ready when you wake up. For efficiency, set tasks like "write a blog post and save it to Google Drive"—it’s that easy.

Always check the plan in deep research before starting, because it auto-runs after 30 seconds. And remember to test different models (Sol, Terra, Luna) and effort levels to find what works best for each task, balancing speed, quality, and token usage.