I’ve been following AI developments for years, but what OpenAI just unveiled with their o3 model has left me genuinely stunned. This isn’t just another incremental update—it’s a fundamental shift in what AI can do, and it’s happening right before our eyes.
Just a year ago, AI systems had IQs below the human average. Now, OpenAI claims o3 has reached genius-level intelligence. While I’m naturally skeptical of such claims, especially without peer-reviewed validation, the demonstrated capabilities speak for themselves.
Visual Reasoning That Defies Expectations
What makes o3 truly revolutionary is its ability to “think with images.” It doesn’t just recognize what’s in a picture—it can analyze, reason, and draw conclusions from visual information in ways that seem almost human.
In demonstrations, o3 identified ships in harbors and predicted their destinations. It read tiny signs on buildings by intelligently zooming in and enhancing details that were barely visible to the human eye. It even identified filming locations from photographs and—yes—found Waldo in those crowded puzzle pictures we all struggled with as kids.
The system can:
- Identify a restaurant from a photo of its menu
- Recognize guitar chords being played in images
- Mark up fabric samples to highlight manufacturing defects
- Follow visual reasoning chains while showing its “thinking” process
This visual reasoning capability transforms how we can interact with AI. It’s no longer limited to text—it can process and understand the world the way we do, through images.
Memory and Personalization That Feels Human
What truly impressed me was o3’s enhanced memory capabilities. It can recall years of previous conversations and build a profile of your interests and knowledge. Then—and this is the game-changing part—it can identify gaps in your knowledge and teach you new things you might find interesting.
In one demonstration, o3 recognized a user’s interest in scuba diving and taught them about how coral larvae detect reef sounds and prefer the natural soundscape of healthy reefs. It even explained how scientists are using underwater speakers playing healthy reef sounds to encourage coral settlement and growth.
This level of personalization creates an experience that feels remarkably human. The AI isn’t just answering questions—it’s building a relationship with you over time.
Breaking the Boundaries of Known Knowledge
The most significant advancement might be o3’s performance on “Humanity’s Last Exam”—a collection of questions from scientists worldwide that no AI could answer when the test was created.
Just months ago, the best AI systems scored around 8% on this test. O3 is approaching 25%—a threefold improvement in a matter of months. At this rate, we might see scores above 50% by year’s end, which would represent a historic breakthrough.
Why does this matter? Because it suggests AI is moving beyond the “convex hull” of human knowledge—beyond simply working with what we already know. It’s showing signs of being able to push knowledge forward, potentially helping solve problems in:
- Drug design and medical research
- Agricultural innovation
- Clean energy development
- Longevity science
In demonstrations, o3 was able to complete unfinished research work, showing its “thinking” process with images along the way. While it was re-inventing something already known, the implications are clear: we’re approaching AI systems that can make novel discoveries.
Tools for Everyone, Not Just Experts
OpenAI also announced Codex, a coding agent designed for everyone—not just programmers. With minimal prompting, it can create functional applications based on simple descriptions or references to projects seen on social media.
This democratization of technology creation could transform how we interact with computers. When anyone can describe what they want and have AI build it, we enter a new era of human-computer collaboration.
I’ve started using a simple trick with ChatGPT that I recommend: customize it to look for peer-reviewed sources and rate their credibility for every answer. This helps distinguish between speculation and verified information—a crucial distinction as AI becomes more integrated into our information ecosystem.
The pace of AI advancement is breathtaking. Just when we think we understand the limits of what’s possible, systems like o3 come along and completely redefine our expectations. We’re witnessing history in the making—a transformation that will likely change how we work, learn, and solve problems for generations to come.
Frequently Asked Questions
Q: How does o3’s visual reasoning capability work?
O3 processes images by analyzing visual elements, zooming in on important details, and connecting visual information with its knowledge base. It can enhance image quality when needed and follow logical reasoning chains while displaying its “thinking” process visually. This allows it to read tiny text, identify objects, recognize locations, and even understand contextual information within images.
Q: What makes o3’s memory capabilities different from previous AI models?
Unlike earlier models that treated each interaction as isolated, o3 can maintain context across extended conversations—even spanning years of interactions. It builds a profile of your interests and knowledge over time, allowing it to provide personalized information that fills gaps in your understanding. This creates a more natural, human-like interaction where the AI remembers your preferences and avoids repeating information you already know.
Q: What is “Humanity’s Last Exam” and why is it significant?
Humanity’s Last Exam is a collection of challenging questions created by scientists worldwide that were specifically designed to test the limits of AI capabilities. These questions couldn’t be answered by AI systems at the time the test was created. O3’s performance jump from 8% to nearly 25% in just months suggests AI is rapidly approaching the ability to tackle novel scientific problems and potentially make discoveries beyond existing human knowledge—a watershed moment in AI development.
Q: How can everyday users benefit from these AI advancements?
These advancements make AI more accessible and useful for non-technical users. You can now interact with AI using images, receive personalized learning experiences based on your interests, and even create software applications through simple descriptions rather than coding. For research, work, and education, these tools offer unprecedented support by understanding your specific needs and knowledge gaps. The key benefit is that complex technology is becoming available to everyone, not just specialists.
