The Vergecast hosts discuss the underwhelming launch of GPT-5, highlighting its incremental improvements, coding challenges for non-experts, and user frustrations over sudden model access changes. They also cover recent tech industry news, AI reliance concerns, and the limitations of chatbots in explaining their own workings, urging critical evaluation of AI outputs.
The Vergecast episode opens with hosts Jake Kasernakis, Verge senior reviewer Vonger, and Verge senior AI reporter Hayden Field discussing the much-anticipated but underwhelming launch of GPT-5 by OpenAI. Despite the hype and expectations, users found GPT-5 to be only an incremental improvement over GPT-4, with some even feeling it was less friendly and worse at creative writing. The launch was marred by issues such as the sudden removal of access to older models without warning, leading to user backlash. The hosts also note the new colorful splash screen as one of the few positive changes, though it was seen as a superficial upgrade compared to the model’s actual performance.
The conversation then shifts to the challenges of “vibe coding” with GPT-5, where the hosts share their personal experiences trying to get the AI to build interactive applications. Despite the promise of AI-assisted coding, all three found the process frustrating and complicated, especially as non-coders. The AI often produced code that required users to have some coding knowledge to implement or debug, making it less accessible for beginners. While GPT-5 showed some improvements in coding capabilities, it still struggled with basic functionality and usability, highlighting that the technology is not yet ready to fully replace human developers or serve as a simple tool for casual users.
Next, the hosts dive into a “corporate shenanigans” segment, analyzing recent headline-grabbing corporate moves. They unanimously agree that Perplexity’s offer to buy Google Chrome for $34.5 billion is a blatant publicity stunt. Apple’s lawsuit against a small movie theater chain named Apple Cinemas is seen as serious but somewhat excessive, reflecting Apple’s intense brand protection. The discussion also covers Apple’s workaround to an import ban on the Apple Watch’s blood oxygen sensor, which the hosts find ridiculous but serious. Lastly, Elon Musk’s claims that Apple is rigging the App Store against his AI app Grok are dismissed as another attention-seeking shenanigan.
The episode continues with a “thunder round” rapid-fire segment covering various tech stories. Highlights include a Verge writer’s experiment living solely with an LTE Apple Watch, which proved challenging but insightful about device dependence. They also discuss a study showing doctors becoming worse at detecting cancer when AI assistance is removed, raising concerns about over-reliance on AI in critical fields. The hosts reflect on the shutdown of AOL dial-up internet after 34 years, evoking nostalgia for early internet days. Additionally, the return of the Pebble smartwatch brand sparks debate between fans of minimalist, open-source devices and those wanting modern smartwatch features.
Finally, the hosts explore the limitations of chatbots like GPT and Grok in revealing their own inner workings. They explain that these AI models often fabricate explanations when asked about their design or bans, as they lack true self-awareness and only generate plausible-sounding responses based on patterns. This leads to confusion and misplaced trust among users who treat chatbots as oracles. The episode closes with a reminder to approach AI outputs critically, especially in sensitive areas like healthcare, and encourages listeners to support The Verge through subscriptions and feedback.