- Everyday AI
- Posts
- Ep 644: 5 Underrated ChatGPT Features You Should Be Using But Aren’t
Ep 644: 5 Underrated ChatGPT Features You Should Be Using But Aren’t
Canva goes big with AI announcements, Apple opens up AI Siri to third parties, unlikely AI music collab and more
Everyday AI
October 31, 2025
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
6 minute read
🎙 Daily Podcast Episode: Think you know ChatGPT? Think again. Discover 5 overlooked ChatGPT features that very few are using. Give today’s show a watch/listen.
🕵️♂️ Fresh Finds: Bill Gates says the AI is bubble likely to pop, OpenAI announces Sora pricing updates, Amazon’s big AI Capex project goes live. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: Canva goes big with AI announcements, Apple opens up AI Siri to third parties, unlikely AI music collab and more. Read on for Byte Sized News.
🧠 Learn & Leveraging AI: Here are 5 key ChatGPT features you may have missed. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We covered: Inside Amazon's 30K job cuts due to AI, OpenAI’s $1 trillion IPO, Fed Chair says no to AI bubble, Gemini’s huge user jump and more. Check it here!
5 New Overlooked ChatGPT Features You Should Be Using But Aren’t 👀
ChatGPT Agents and Atlas have taken all the spotlight.
But these 5 underrated ChatGPT features can instantly improve your results.
Also on the pod today:
• Advanced Voice Mode in ChatGPT & GPTs 🗣
• Gmail and Google Calendar Auto Connectors 📥
• Project Folders vs. Custom GPT Organization 🗃️
It’ll be worth your 33 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Base44 helps you build apps with your ideas, Airtop helps you create agents with natural language, Frabrica builds internal apps with prompts only
AI Acquisitions — Figma acquires Weavy — bringing node-based AI media creation directly into design.
Bill Gates on AI — Bill Gates warns AI is a dot-com style bubble, valuable long-term but many firms will fail. Want the details?
Data Centers — OpenAI and Oracle announced they are building a 1+ GW Stargate AI campus in Saline Township, Michigan
Atlas Browser — OpenAI revealed some details on how they built Atlas, and it’s not a simple Chromium fork.
Amazon AI — Amazon's Project Rainier is live, running nearly 500,000 Trainium2 chips and boosting Anthropic's compute fivefold.
OpenAI Pricing and Credits — OpenAI announced the availability to extend its limits by offering credits in Sora, as well as credits in Codex.
1. Canva unveils a design-aware AI model and a slew of creator tools 🪄
Canva launched its own foundational AI design model that generates editable, layered designs across formats like posts, presentations, whiteboards and websites, and it rolled the AI assistant throughout the app while adding 3D object generation and style-copying features.
The company also introduced a spreadsheet product tied to a new app-builder so users can turn live data into visualization widgets, plus Canva Grow for ad creation and analytics after acquiring MagicBrief, and permanent free access to Affinity as it integrates pro tools with Canva.
2. Apple opens Siri to more third‑party AI partners 🤖
According to CNBC, Tim Cook said Apple plans to integrate more third‑party AI models into its operating systems, with ChatGPT already in Siri and Google Gemini and other partnerships reportedly in the works.
Cook noted an AI‑upgraded Siri is on track for next year and that Apple remains open to acquisitions to speed development, comments made alongside Apple’s record Q4 earnings report. For professionals and startups, broader model access could mean easier voice‑driven integrations, new distribution channels, and faster time to market for AI services that plug into the Apple ecosystem.
3. Character.AI to block under-18s from chatbots ⛔
According to The Information, Character.AI will bar users under 18 from open-ended conversations and add age-assurance features by November 25 after lawsuits and regulatory scrutiny tied to teens forming harmful attachments to AI companions.
The move follows multiple wrongful-death suits and wider concern about mental-health risks, and comes as lawmakers and states push new rules for AI interactions with minors.
4. Anthropic shows Claude can sometimes "notice" its own thoughts 🤖
A new study from Anthropic reports that researchers injected concepts into Claude's neural activity and the model occasionally detected and described those internal changes, suggesting a limited form of introspection that emerged without explicit training.
The capability was real but fragile: top models detected injections about 20% of the time, often confabulating details and failing under many conditions, so the findings are scientifically intriguing but far from reliable for safety-critical use.
5. OpenAI unveils Aardvark, a GPT-5–powered security researcher agent. �*
OpenAI announced Aardvark, a private-beta agent built on GPT-5 that scans code repositories to find, explain, and help patch security vulnerabilities, marking a timely push into automated defensive tooling as software flaws continue to surge.
The agent inspects a codebase to understand its purpose and risks, hunts for issues by reviewing past and new commits, and annotates findings for human review, then attempts proof-of-concept triggers in a sandbox and tags results with searchable metadata. It can also generate candidate fixes by pairing its analysis with OpenAI’s Codex to produce patches for developers to vet and apply.
6. UMG settles with Udio, will build licensed AI music service �*
Universal Music Group has settled its copyright lawsuit against AI generator Udio and agreed to collaborate on a licensed AI music generation and streaming platform due in 2026.
The partnership ends UMG’s part in a high-profile infringement fight and promises a subscription product trained on UMG’s catalog, while Udio’s current product will operate in a restricted “walled garden” until the new service launches. The deal signals major labels are moving from litigation to licensing and product-building as the industry sets rules for generative AI.
🦾How You Can Leverage:
Between agents and Atlas, so many crazy useful features inside ChatGPT are sitting there unused like forgotten apps on your phone that you got giddy about when you downloaded them.
We're talking hidden QuizGPT flashcards that slap harder than Will Smith. Personality modes that fix ChatGPT's annoying yes-man problem that's been driving everyone nuts.
Gmail integration that FINALLY works instantly instead of making you wait 12 painful minutes. Yasssss.
These ain't minor updates, shorties.
They're workflow game-changers that even power users completely missed because the timing was absolutely terrible.
But you've been missing them.
So on today's show, we're exposing five buried ChatGPT features that'll save you hours weekly while your competitors stay distracted by launch drama.
Oh, and we snuck in even more goodies when/if you share this post.
Let's mine these productivity gems, y'all.
1 – Project Custom Instructions Finally Work 🎯
Custom instructions were hot garbage until three weeks ago.
Get this - they applied to literally every conversation you ever had.
Client work got the same tone as weekend meal planning. Professional emails used identical formatting as personal brainstorming sessions.
Complete nightmare for anyone juggling multiple contexts.
The new custom instructions feature for projects just fixed this mess forever.
Now, each folder gets separate rules AND isolated memory that never contaminates other workspaces.
Your work project demands citations and bullet points. Personal stuff stays conversational and fun.Client folders follow their specific brand voice without bleeding into your grocery list chats.
Most people don't even know this exists because OpenAI barely whispered about it in their change log.
Try This:
Create three new projects right now with names that match how you actually use AI - work, personal, clients. Click the gear icon before any chats and set memory to "project only" so contexts never mix.
Write instructions like "Always cite sources, 200 words max, clear next steps" for work and "Be conversational, help me think" for personal projects. This five-second setup saves explaining context every single time.
2 – Gmail Autopilot Crushes Inbox Anxiety 📧
Email overwhelm just got demolished without the endless inbox boolean searches.
The Gmail connector existed for months but was stuck in that painfully slow deep research mode that made you want to throw your laptop.
Need inbox priorities? Wait forever.
Want those confusing twenty-person email threads explained? Another brutal deep research delay that killed your momentum.
Connect once and ChatGPT immediately sorts urgent emails from spam, drafts intelligent replies, and translates those group conversations that make absolutely zero sense.
Here's what's crazy about this update.
If you train it a bit, it understands email importance based on actual context, not just desperate subject line keywords like "URGENT!!!"
Startup pitches automatically get filtered as noise. Client urgency gets flagged immediately.
Meeting requests get proper priority ranking instead of fighting for attention with newsletter garbage.
Your inbox anxiety disappears because you're finally seeing what actually demands your brain space.
Try This:
Connect your Gmail in ChatGPT settings under "Data controls" right now. Ask "Which emails from the past 48 hours need my immediate response and why?" You'll get a smart priority list instead of chronological chaos.
Request draft replies for the top three, add personal touches, and paste into Gmail - thirty minutes of inbox anxiety becomes three minutes of focused action.
3 – Voice Mode in GPTs Unlock Different Thinkings 🧠
Talking to specialized AI completely changes how ideas flow through your brain.
Advanced voice mode now works inside every custom GPT you've built.
Voice brainstorming with your marketing assistant feels totally different than typing the same boring questions.
Ideas surface faster. Strategic connections happen naturally.
Real-time conversation reveals insights that stay buried in text exchanges.
But hold up - there's a massive gotcha waiting to wreck your day.
Voice mode can't access uploaded files in GPTs yet, and it doesn't know this limitation exists. So be careful… voice mode in custom GPTs will hallucinate HARD if you’re asking it about uploaded files in your GPT’s memory.
(Obviously works fine in text-only mode though.)
Still, being able to use Advanced Voice Mode in custom GPTs is a commute game changer for busy execs.
Try This:
Pick your knowledge-based custom GPT - marketing strategist, coding assistant, whatever you use most.
Start a voice conversation about your biggest current challenge and actually speak your messy thoughts instead of typing perfectly structured sentences.
Let it flow naturally for ten minutes without worrying about sounding professional - verbal brainstorming unlocks completely different neural pathways than typing.
Reply