• The AI Brief
  • Posts
  • 5 ChatGPT Settings That Put You Ahead of 99% of Users

5 ChatGPT Settings That Put You Ahead of 99% of Users

And GPT-5.2 is now outperforming 74% of human experts on their job

Hey, here's this week's AI Brief!

  • Your AI Toolkit: 5 ChatGPT settings most people never touch

  • Prompt of the Day: Find Your Best AI Opportunities on Auto-Pilot

  • Last Week in AI: GPT-5.2, ChatGPT Images upgrade, Zoom AI Companion 3.0

  • What I'm Learning: Claude Skills might be the feature I've been waiting for

🛠️ Your AI Toolkit

5 ChatGPT Settings That Actually Matter

Prefer video? I made a 60-second walkthrough showing exactly where to find each setting.

@thaddeusai

Comment "prompt" & follow to get all my favorite prompts. These 5 settings make ChatGPT 10x smarter. Most people never touch them. #ai #ch... See more

Most people use ChatGPT like a fancy search engine. They open it, ask a question, get an answer, and close it. That's maybe 10% of what the tool can do.

Here are five settings that transform ChatGPT from a basic Q&A tool into something that actually knows you and works on your behalf:

1. Connect Your Apps Go to Settings > Connectors. You can link Gmail, Google Drive, Calendar, even GitHub. When you use agent mode or deep research, ChatGPT pulls directly from your actual tools instead of guessing. Only connect what you're comfortable sharing, and you can also limit what connections you allow it to take.

2. Turn On Memory Settings > Personalization. Enable both toggles: saved memories and chat history. Now ChatGPT remembers what you've told it and weaves that context into every conversation automatically.

3. Switch to GPT-5 Thinking or Pro Mode Click the model selector at the top. This forces ChatGPT to reason longer before responding. You get deeper, more thorough answers to complex questions. Essential if you're using it for real work.

4. Set Up Custom Instructions Stay in Personalization settings and toggle on "Enable customization." You get two boxes: one tells ChatGPT who you are and what you do, the other controls how it writes. This makes every response cleaner and tailored to how you actually work.

5. Set Up Schedules This one gets almost no attention. Settings > Schedules lets you have ChatGPT run tasks automatically. I have one that reviews my recent conversations every Monday and surfaces new AI workflow opportunities. It runs whether I remember or not. (See this week's Prompt of the Day for a preview of the prompt I use.)

If you do all these, I’m confident you’re at least ahead of 95% (if not 99%) of other ChatGPT users :)

đź’­ Prompt of the Week

Find Your Best AI Opportunities on Auto-Pilot

This is the scheduled prompt I mentioned in the Toolkit section. I have it run every Monday at 10am, and it surfaces AI workflow opportunities I'd otherwise miss.

This is just a preview. The full prompt includes ICE scoring criteria and output formatting. If you want the full prompt, just reply to this email and I'll send it over.

Important setup note: Schedules within a ChatGPT Project can only read conversations and files within that project. Make sure you're setting this up in the project where you do your most important work, not a random side project.

Why it works: ChatGPT has context from your conversations that you've already forgotten. This prompt mines that context for patterns you can't see yourself. It's like having a productivity consultant review your week and hand you a list of improvements every Monday.

Prompt Preview:

Find Your Best AI Opportunities on Auto-Pilot

Purpose: Continuously mine real work to surface high-leverage AI workflows worth templatizing, prioritized by ROI.

Scope of Analysis:

- Analyze all chats and files since the last run

- Include: strategy discussions, client work, coaching sessions, workflow design, prompt experimentation

- Exclude: throwaway questions, non-actionable chats

Core Task:

1. Review all new material since the last run

2. Identify: repeated tasks, pain points, manual synthesis work, decision bottlenecks, reusable patterns

3. For each opportunity, propose one concrete AI use case suitable for templating

đź’ˇ Last Week In AI

GPT-5.2 is Here (Dec 11) 

  • Last week, OpenAI was calling "code red" after Google's Gemini 3 became the top AI model. Then they dropped GPT-5.2. For the first time, an AI is outperforming a majority of human experts on real work tasks, beating professionals 74% of the time on a new evaluation called GDPval.

  • I think the biggest thing to pay attention to is the improved document creation. Look at the GPT-5.1 vs GPT-5.2 comparison when making a slide given the SAME prompt… it’s night and day.

ChatGPT Images Gets a Major Upgrade (Dec 16) 

  • OpenAI also released GPT Image 1.5, making image generation 4x faster with much better editing. You can now make specific changes (swap a shirt, adjust lighting) without the whole image changing.

  • There's a new dedicated Images tab with preset styles and trending prompts too. Try it for your marketing materials!

Zoom AI Companion 3.0 Goes Agentic (Dec 15) 

  • Zoom launched AI Companion 3.0 with "personal workflows" that turn meeting conversations into action items, documents, and follow-ups automatically. The tools you already pay for are getting smarter. Check your Zoom plan to see if you have access before signing up for new AI subscriptions.

đź“– What I’m Learning

Claude Skills Might Be the Feature I've Been Waiting For

Anthropic just released an interview explaining Claude Skills and it's worth 15 minutes of your time. Skills let you upload documents that Claude remembers across every chat in a project. It's breaking your business processes down for Claude to then be able to access across every chat or project and be able to continuously update when it detects you've changed your way of working. This feels like the future of context management.

I'm testing it with my agency’s SOPs and the early results are promising. My full breakdown video is coming soon!

Final Thoughts

Hope this was helpful. My top two recommendations from all of these developments are:

  1. Try pushing GPT 5.2 to its limits by asking it to do more for you (especially creating powerpoint decks and excel analyses) & see where it breaks.

  2. Start to A-B test yourself anytime there's a new feature (Claude Skills) or model released (GPT-5.2) with an actual work task that's important to you, so you can decide if it's worth changing your workflow.

-Thaddeus

Reply

or to participate.