I think it would be very useful to be able to use reasoning models in that context, personally it would be very useful to me
Feature Request: Integration of GPTs Directly Within Projects
Background
The current ChatGPT platform offers two powerful features:
- Projects, which help users organize workflows, files, and tasks over time.
- Custom GPTs, which allow users to tailor AI interactions for specific needs.
However, these features operate independently, which creates a disjointed user experience. Users must switch between Projects and GPTs, manually transferring content between the two. This lack of integration undermines the platform’s potential for seamless productivity and efficiency.
User Story
As a user, I want to be able to directly use GPTs within my Projects so that I can seamlessly integrate AI-generated outputs into my workflows without needing to switch contexts or manually transfer content.
Proposed Feature: Direct GPT Integration in Projects
Enable users to select and run GPTs directly within the Projects interface, allowing for streamlined AI assistance in content generation, brainstorming, and analysis, with outputs saved and organized within the project workspace.
Acceptance Criteria
- Direct GPT Access:
- Users can select and activate a GPT within a Project.
- GPT responses appear directly within the project workspace.
- Context Awareness:
- GPTs automatically adapt to the Project’s context (e.g., topic, tone, or purpose) based on metadata or initial user input.
- Users can refine the GPT’s behavior with additional prompts tailored to the Project.
- Embedded Outputs:
- GPT-generated outputs are automatically saved as files, notes, or attachments within the Project.
- Users can choose where and how the outputs are organized (e.g., by folder, as draft documents).
- Collaboration Features:
- Team members within a shared Project can view and collaborate on GPT-generated outputs in real-time.
- A history of GPT interactions is logged within the Project for reference.
- Seamless User Experience:
- No need to switch between the Projects and GPTs sections of the platform.
- A “Run GPT” button is available in the Project toolbar, allowing users to select from existing GPTs or create new ones.
- Performance and Usability:
- GPT integration does not slow down Project performance.
- Users can easily toggle between manual inputs and GPT assistance.
Expected Results
- Improved Productivity:
- Users complete tasks more efficiently by eliminating context switching and manual content transfer.
- Enhanced User Satisfaction:
- A more cohesive experience increases engagement with the platform.
- Broader Adoption:
- Professionals, teams, and businesses are more likely to adopt ChatGPT as a comprehensive tool for managing both workflows and AI-driven tasks.
- Innovation Leadership:
- By integrating GPTs into Projects, OpenAI strengthens its position as a leader in AI-enhanced productivity tools.
Examples of Use Cases
- Report Writing: A user can draft, revise, and finalize a report within a Project by using GPTs to generate sections, proofread content, and format the document.
- Video Production: Teams working on video scripts can use GPTs to brainstorm ideas, create scripts, and refine narratives within their shared Project.
- Research and Analysis: Users can ask GPTs to analyze data, summarize findings, or generate insights directly in their research Project workspace.
Conclusion
Integrating GPTs directly into Projects would transform the ChatGPT platform into a unified powerhouse for productivity, reducing friction and elevating the user experience. This feature would unlock greater value for users, fostering deeper engagement and broader adoption across industries.
We recommend prioritizing this feature to align with user needs and market expectations.
To enhance AI’s ability to answer NCLEX questions more accurately, it’s essential to focus on improving reasoning by integrating advanced reasoning models with custom GPTs.
Key Points Include:
-
Common Challenges:
- AI struggles to achieve high accuracy when it should be able to reason through 99% of questions accurately.
- The remarkable ability of AI to achieve correctness rates of 50% or higher suggests that training in reasoning through licensing exam questions and how to process the information could be improved.
- Custom GPTs enable more accurate answering of NCLEX questions
- Enhanced reasoning models answer NCLEX questions more accurately than GPT4o, but generally do not exceed 90%.
-
Need for Higher Reasoning Models:
- Integrating advanced reasoning models with custom GPTs can enhance AI’s decision-making capabilities, leading to more precise answers.
- Current models like GPT-4, despite improvements, are still limited in reasoning and typically achieve only 60-80% accuracy on NCLEX questions.
- Exploring options like “o3-mini high,” “o1,” or “o1 pro” specialized models might demonstrate accuracy rates of roughly 90%. Advanced training and/or integrating with customGPTs might enable more reliable answering of NCLEX, USMLE, and other licensing exam type questions.
-
Conclusion:
- The integration of higher reasoning models with custom GPT functionality holds promise for improving AI performance.
This approach emphasizes the synergistic use of advanced reasoning capabilities and tailored AI models to significantly boost performance, addressing both the reasoning errors of current models and limiting knowledge supplementation files to areas where the GPT makes mistakes.
I just made the experience that in my custom Gpt all of a sudden o3 Mini was definitely active. This happened when using chatgpt on my mobile phone. Crazy. And it definitely started to reason.
But, I wouldn’t know how to activate this myself, yet.
In my chatgpt suddenly o3 Mini appeared and it started to reason
O3 mini does not work with certain files. In my custom GPT, I have about 10 megabytes of .docx files. Do you have uploaded files as part of your custom GPT?