Quick Overview
This short format showcase video features creator Matt Wolfe presenting prompt concepts for next-generation ChatGPT capabilities inside the Codex application. The demonstration follows recent updates in model tooling and explores how expanded agent features could streamline creative, administrative, and developer tasks.
Key Points
- 1.GPT-6 computer use capabilities allow the AI model to operate external desktop software like Blender to construct and render 3D scenes from photographs.
- 2.GPT Image 2.5 introduces precision image editing that modifies specific designated sections of an image while preserving the surrounding structure.
- 3.ChatGPT Work uses a persistent cloud browser to access online accounts and manage subscription billing without interacting with the user's personal browser.
- 4.The GPT-Live 1 API enables applications to maintain an uninterrupted voice dialogue while background agents perform live research.
- 5.The Agents API supports persistent multi-session memory to build coaching tools that track progress and resume from previous sessions.
Summary
- 1.Open Blender 3D diorama generation. The first prompt directs GPT-6 inside the Codex app to launch Blender and build a miniature 3D model of a workspace based on an uploaded photograph. The model handles scene construction, lighting setup, and final rendering directly within the software. Upgraded computer use features allow the system to manipulate complex third-party tools to transform two-dimensional reference images into complete digital dioramas.
- 2.Localized interior design editing with GPT Image 2.5. The second prompt asks the model to redesign an interior room by adding a cozy reading corner while keeping the remainder of the image unchanged. The prompt specifies maintaining the overall room structure and focusing solely on design modifications in the designated area. GPT Image 2.5 executes precise localized edits, altering only the requested section while leaving surrounding elements fully intact.
- 3.Subscription dashboard tracking through ChatGPT Work. The third prompt asks ChatGPT Work to review signed-in AI subscription plans, extract their upcoming renewal dates, and construct a unified billing dashboard. ChatGPT Work uses a persistent cloud browser, which accesses external web services independently in an online workspace so it can interact with user accounts without hijacking the local browser.
- 4.Background voice research using the GPT-Live 1 API. The fourth prompt outlines building a voice-powered travel planner that conducts online research while continuing an active spoken conversation. The GPT-Live 1 API enables applications to maintain continuous dialogue with the user while background agents simultaneously query sources and compile information.
- 5.Persistent skill tracking using the Agents API. The fifth prompt directs the model to build a guitar practice coach that records user performance and resumes training across distinct practice sessions. Using the Agents API, the application retains session history, tracks difficulties encountered, and picks up where the user previously stopped.
3D Modeling with Computer Use
The first prompt instructs GPT-6 to open Blender, construct a miniature 3D diorama of a workspace using a reference photograph, configure lighting, and execute the final render. This workflow leverages upgraded computer use capabilities that allow the model to interact directly with external creative software.
Targeted Inpainting with GPT Image 2.5
The second prompt demonstrates GPT Image 2.5 editing an interior room photograph by adding a cozy reading corner. The tool updates only the requested area, preserving the underlying architecture and visual details of the rest of the room.
Account Tracking via Cloud Browser
The third prompt uses ChatGPT Work to inspect signed-in AI subscriptions, retrieve upcoming renewal dates, and compile an overview dashboard. The persistent cloud browser executes these navigation tasks in an isolated workspace without taking over the user's desktop browser.
Voice Agents and Session Persistence
The final prompts highlight developer tools, showing the GPT-Live 1 API maintaining continuous voice conversation during live background research, and the Agents API retaining user progress across sessions to build a persistent guitar practice coach.
The Bottom Line
The video establishes five prompt workflows using model capabilities such as external software control, precise image editing, isolated cloud browsing, continuous conversational research, and session persistence. It demonstrates how these features function within the Codex application across design, administrative, and coaching use cases. It leaves open the wider availability, implementation specifics, and general release schedule for the demonstrated model versions.
FAQ
What is ChatGPT and what are ChatGPT new features showcased in the video?
ChatGPT is an AI assistant platform, and the video showcases advanced capabilities including GPT-6 computer use in Blender, GPT Image 2.5 targeted editing, ChatGPT Work cloud browsing, the GPT-Live 1 API, and the Agents API.
How does GPT-6 computer use automate 3D modeling inside Blender?
GPT-6 operates Blender directly on the desktop to convert a reference photograph into a structured 3D miniature diorama, complete with lighting and rendering.
How does GPT Image 2.5 handle targeted image editing in a room photo?
GPT Image 2.5 modifies only the designated area, such as inserting a reading corner, while preserving the structural layout and details of the surrounding room.
How does the ChatGPT Work cloud browser inspect AI subscription renewal dates?
ChatGPT Work connects to web accounts inside a persistent online cloud browser, allowing the model to gather subscription information and build a dashboard without interfering with the user's personal browser.
What capability does the GPT-Live 1 API provide for conversational agent applications?
The GPT-Live 1 API allows an application to carry on an uninterrupted voice conversation while an autonomous agent conducts background research and retrieves sources.
How does the Agents API support persistent sessions in coaching applications?
The Agents API saves session history and progress over time, enabling a coach assistant to track problem areas and pick up where the user left off in previous sessions.
Worth watching for
Developers, designers, and AI power users interested in prompt workflows for autonomous software control, targeted image manipulation, and persistent API agents.
- chatgpt
- codex
- gpt-6
- prompt-engineering
- agents-api
- computer-use