I recently watched a video by Jeff Su that completely changed how I think about using AI tools like ChatGPT. With over 1.1 updates per week this year, it’s nearly impossible to keep track of all the new features—let alone figure out which ones are actually worth our time.
What struck me most was how Jeff cut through the meaningless jargon to show when to use specific features in real-world situations. As someone who helps businesses create superfans, I’m always looking for tools that can help streamline processes and create better experiences. Here’s what I learned and why it matters for anyone using AI tools.
Picking the Right Model: It’s About Complexity, Not Task Type
The most important takeaway? Choose your AI model based on the complexity of what you’re asking—not the type of task. This contradicts what many of us have been told.
Jeff recommends defaulting to the latest reasoning model (the one without the “extra baggage” at the end of its name). Reasoning models shine when:
- Your query is important or complex
- You’re willing to wait a bit longer for a better answer
- You need nuanced analysis or detailed thinking
Basic chat models work best when you need quick, straightforward answers to low-stakes questions. For example, if you’re asking “Which fruits have the most fiber?” a basic chat model works fine. But if you need a vegetarian breakfast plan with specific protein and fiber requirements, a reasoning model will give you much better results.
Search Features: Know When to Use Google vs. AI Search
I found Jeff’s advice on when to use ChatGPT’s web search versus Google particularly helpful. The rule of thumb is simple:
- Need a single fact? Use Google
- Need a fact with context or explanation? Use ChatGPT Search
- Want information formatted in a specific way? AI search excels
For instance, if I need to know today’s weather in Chicago, Google is faster. But if I’m planning a December trip to Chicago and want to know what clothes to pack based on typical weather patterns, ChatGPT Search provides a more useful, contextual answer.
Instead of just asking for stock prices, ask “When was NVIDIA’s latest earnings call? Did the stock price go up or down and why?” to get both data and valuable context.
Deep Research: Your Personal Research Assistant
As someone who regularly needs to research industry trends for my keynotes, I was fascinated by the Deep Research feature. This powerful tool can analyze hundreds of sources and produce detailed reports while you focus on other tasks.
For example, instead of manually comparing different companies’ earnings reports, you can ask the AI to analyze and compare AI chip roadmaps based on the latest earnings calls from NVIDIA, AMD, and Intel. Or you could request an analysis of high-yield savings accounts, including hidden fees and projections for someone saving $1,000 monthly.
The real game-changer is connecting your own sources like Google Drive, allowing Deep Research to incorporate your private or proprietary information. This opens up powerful business applications, such as generating reports that compare your company’s performance against competitors using both internal data and external industry reports.
Canvas: The Editor’s Best Friend
The Canvas feature is perfect when you know you’ll need to edit and build upon AI responses multiple times. I can see myself using this for preparing presentation outlines, drafting articles, or working on performance reviews.
What makes Canvas particularly useful is the ability to:
- Jump between different versions using back and forth buttons
- Use built-in shortcuts to make changes to the entire document or highlighted sections
- Get AI suggestions for improving specific parts of your writing
- Download the final output in markdown format for easy formatting in other programs
For those of us who write regularly, this feature alone could save hours of editing time each week.
Practical Commands That Make a Difference
I appreciated Jeff’s three favorite commands for text-to-text models:
- Elaborate – To add more detail to bullet points or brief sections
- Critique – To spot problems early before presenting ideas
- Rewrite – To improve previous content with specific tone adjustments
These simple commands can dramatically improve the quality of AI outputs without requiring complex prompts.
The AI landscape is evolving rapidly, but understanding when to use specific features based on your actual needs—rather than following generic advice—will help you get the most value from these tools. I’m excited to incorporate these strategies into my workflow and see how they can help me create better content for my audience.
What’s your experience been with these AI features? Have you found certain approaches work better for specific tasks? The more we share our practical experiences, the better we all become at leveraging these powerful tools.
