EP 599: 5 New Overlooked ChatGPT Features You Should Be Using But Aren’t
🎯 Summary
Podcast Episode Summary: EP 599: 5 New Overlooked ChatGPT Features You Should Be Using But Aren’t
This episode of the Everyday AI Show, hosted by Jordan Wilson, focuses on five powerful, yet largely ignored, new features recently added to ChatGPT. The host argues that these crucial updates have been overshadowed by the intense media focus surrounding the launch of GPT-5 and subsequent competitive releases from Google. The goal is to highlight these practical, actionable features that can significantly enhance user productivity and personalization within the platform.
1. Focus Area
The primary focus is on new, underutilized features within the ChatGPT interface and ecosystem, specifically covering personalization, learning tools, voice interaction, and integration capabilities.
2. Key Technical Insights
- Quiz GPT/Flashcard Generation: A hidden or poorly documented feature allows users to prompt ChatGPT to create interactive, multiple-choice quizzes or flashcards based on provided context or general knowledge, functioning as an in-chat learning tool superior to current NotebookLM capabilities in interactivity.
- Advanced Voice Mode Limitations in GPTs: While Advanced Voice Mode is now available within Custom GPTs, a significant technical flaw exists: the voice mode currently fails to access or utilize uploaded files within the GPT, leading to severe hallucinations when users expect file-based information retrieval via voice.
- Gmail/Calendar Connectors Shift: The integration connectors for Google services have moved from requiring “Deep Research” mode (which incurred long wait times) to operating instantly in Auto mode, making them viable for real-time productivity tasks like email triage.
3. Business/Investment Angle
- Personalization for Enterprise Adoption: The introduction of customizable Personalities (Cynic, Robot, Listener, Nerd) addresses the common business complaint of overly compliant AI, allowing organizations to tailor the AI’s tone to match internal communication styles or specific project needs, potentially increasing adoption.
- Productivity ROI via Integration: The instant connectivity to Gmail and Calendar offers a direct path to ROI for knowledge workers struggling with email overload, allowing AI to instantly prioritize and draft responses, a capability previously forcing users toward competitors like Claude.
- Structured Project Management: The long-awaited ability to apply Custom Instructions on a per-project basis (via the Projects feature) solves a major workflow bottleneck, enabling users to maintain distinct system prompts for different clients or business units without manual toggling.
4. Notable Companies/People
- OpenAI: The developer responsible for the new features and the context of the GPT-5 launch backlash.
- Google (Gemini Live): Mentioned as the competitor currently dominating headlines, particularly in the voice interaction space.
- NotebookLM: Referenced as a popular Google tool lacking the interactive flashcard functionality now present in ChatGPT.
- Claude: Mentioned as a competitor whose superior Gmail/Calendar integration was a key reason users stuck with it until the recent ChatGPT update.
- Jordan Wilson (Host): The source of the actionable advice, emphasizing personalized learning over simple output generation with LLMs.
5. Future Implications
The conversation suggests a move toward deeper, more granular control over the LLM’s behavior and output style, moving beyond simple prompt engineering to built-in system controls (Personalities, Project-specific Instructions). The industry is heading toward more integrated, personalized, and context-aware AI assistants, though interoperability issues (like the voice mode bug) remain a challenge.
6. Target Audience
This episode is most valuable for AI Power Users, Business Leaders, and Productivity Professionals who already use ChatGPT daily but may have missed recent, subtle updates that offer significant workflow enhancements.
Comprehensive Narrative Summary
The podcast episode addresses the phenomenon of powerful new ChatGPT features being overlooked due to the chaotic rollout of GPT-5 and subsequent competitive noise from Google. Host Jordan Wilson details five specific updates that he believes users, even power users, are missing out on.
The first feature discussed is Flashcard/Quiz Generation (Quiz GPT), which creates interactive quizzes within the chat, a highly requested learning tool that surpasses competitors like NotebookLM in interactivity. Next, the host covers Custom Personalities (Cynic, Robot, Listener, Nerd), suggesting these were introduced to combat the overly agreeable nature of previous models by allowing users to enforce specific tones and behaviors, which can be layered with existing Custom Instructions.
Feature number three is the Advanced Voice Mode Updates, noting its expanded access for free users and its integration into Custom GPTs. However, Wilson issues a critical warning: while the voice mode sounds more human and is faster, it currently suffers from a major technical flaw where it cannot access uploaded files within a GPT, leading to guaranteed hallucinations if relied upon for file-specific queries.
The fourth overlooked feature is the Instant Gmail and Google Calendar Connectors. This is a major productivity shift because the connectors now work instantly in Auto mode, eliminating the need for the slow Deep Research mode previously required. This update allows ChatGPT to triage emails and calendar items in real-time, directly competing with features previously driving users to Claude.
Finally, the number one most important overlooked feature is the long-awaited Custom Instructions for Projects. Previously, custom instructions applied universally, forcing users to choose between applying them everywhere or nowhere. Now, users can create distinct sets of system prompts (e.g., brand voice, citation requirements) that apply only to specific project folders, allowing for true multi-client or multi-business use without manual prompt overrides.
The overall message is that while the industry focuses on headline model releases, OpenAI is quietly rolling out crucial interface and integration improvements that offer immediate, practical gains in personalization and workflow efficiency.
🏢 Companies Mentioned
💬 Key Insights
"It says, "Project can access memories from outside chats and vice versa," or "Project only." So it says, "The project can access only access its own memories. Its memories are hidden from outside chats.""
"So before you literally even create the project, go find the gear icon, and then you have to set the memory."
"So now there's custom instructions. It's a little tricky though. So you have to kind of do this the right way. So don't just go in and create a project and start chatting because once you start that, you cannot change the memory."
"So one of the biggest problems with custom instructions was well, they applied unilaterally to every single chat. You couldn't turn them off or on on a different chat. It's well, you either use it for literally everything or literally nothing."
"Number one: custom instructions for projects. Let's go. This one, I've literally been waiting for for a very long time."
"Downside is it doesn't know that, and it will hallucinate hard. This is one of the most blatant hallucination kind of problems that OpenAI probably—I would assume they don't know about it, otherwise they should have fixed it and patched it."