When OpenAI CEO Sam Altman declared that GPT-5 delivers interactions akin to "talking to a PhD-level expert in any topic," the AI community took notice. This assertion positions GPT-5 not as an incremental update, but as a potential paradigm shift in artificial intelligence capabilities. But does it hold up under scrutiny?

Article illustration 1

"GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD level expert,"
— Sam Altman, OpenAI CEO

WIRED's senior AI reporters—including Will Knight (author of AI Lab) and Kylie Robison (author of Model Behavior)—are rigorously testing GPT-5’s performance across critical dimensions:
- Reasoning depth: Does it handle nuanced, domain-specific queries with academic rigor?
- Coding proficiency: Can it debug complex systems or generate production-ready code?
- Creative execution: How does it balance originality with factual accuracy in content generation?

Early observations suggest tangible improvements in contextual understanding and reduced hallucination rates, though questions remain about its ability to consistently replicate expert judgment in high-stakes scenarios like medical diagnostics or legal analysis.

This evaluation isn't academic. For developers, GPT-5's architecture could redefine:
1. Coding workflows: Potential for real-time pair programming with "expert" AI
2. Enterprise adoption: Shifting from experimental chatbots to mission-critical knowledge systems
3. Accuracy thresholds: Raising expectations for error-free technical outputs

As Knight notes: "The leap from 'helpful assistant' to 'domain expert' demands unprecedented reliability—especially when incorrect AI-generated advice could have serious consequences."

WIRED will unveil comprehensive findings during an August 14 livestream, dissecting whether GPT-5 truly heralds a new era of specialist-grade AI or reveals persistent limitations in artificial general intelligence. For engineers and tech leaders, the answer will shape integration roadmaps for years to come.