In the video, the creator tests GPT-5’s ability to build GTA 6 and assist with finding a suitable car seat, but finds the AI’s outputs to be slow, inaccurate, and lacking coherence or reliability. While GPT-5 shows some potential in generating quick answers and useful information, it ultimately falls short of expectations, prompting the creator to continue exploring its capabilities.
In the video, the creator tests GPT-5’s capabilities by challenging it to build GTA 6, a highly anticipated game. The code was written in React and took about 28 seconds to run on a virtual machine. However, the resulting game was very poor in quality, confusing to navigate, and far from what anyone would expect from a GTA 6 experience. The creator and his co-host Misha express disappointment, highlighting that the AI-generated game was essentially unusable and lacked coherence.
Following the GTA 6 test, the creator shifts focus to a more practical query, asking GPT-5 to help find a suitable car seat for a child with specific requirements. The car seat needed to be compatible with ISOFIX and also work with a seat belt. The AI took some time to process the request, and while it attempted to provide information, the results were mixed. The AI gave some inaccurate details, such as mentioning car seats without ISOFIX despite the requirement, and failed to provide clear pricing or direct purchase links, which was frustrating.
Despite the shortcomings, the AI did manage to pull up some useful information, including real mom reviews and some product details. The creator found this aspect somewhat promising, noting that the AI could potentially integrate affiliate links or monetization strategies in the future. However, the lack of clarity on whether the recommendations were influenced by affiliate marketing raised some concerns about impartiality.
Overall, the creator felt that GPT-5, while showing some progress, still felt like a patched-together version of previous models rather than a significant leap forward. The responses were often slow, sometimes inaccurate, and lacked the depth and reliability expected from a next-generation AI. The creator expressed a need to continue testing to see if improvements could be made or if the AI could deliver more practical and trustworthy results.
In conclusion, the video presents a candid and critical look at GPT-5’s current capabilities. While it shows potential in areas like generating quick answers and possibly integrating monetization, it falls short in delivering complex tasks like building a game or providing fully reliable product recommendations. The creator invites viewers to share their thoughts and indicates a willingness to keep experimenting with the AI to better understand its strengths and limitations.