The creator shares a balanced review of GPT-5’s coding capabilities with the Cursor Vibe environment, noting initial promise but ultimately finding its performance underwhelming, especially with large codebases and UI generation compared to competitors. While cost efficiency and the new router system are promising innovations, GPT-5 doesn’t represent a revolutionary leap and requires users to adjust to its limitations and changes.
In this video, the creator shares their personal experience working with GPT-5 and the Cursor Vibe coding environment. Initially, they found the experience quite cool but were ultimately underwhelmed due to the high expectations set by early news coverage. While Cursor worked well at the start, issues arose when dealing with a large codebase. As the context window filled up to around 60-80%, the quality of GPT-5’s responses noticeably declined, with answers becoming short and less proactive, often suggesting trial-and-error approaches rather than providing direct solutions.
The creator found that resetting the context and relearning the codebase improved the situation, but the performance still wasn’t extraordinary. They acknowledged that GPT-5 is a great model but emphasized that it has limitations, much like other current language models. It wasn’t a revolutionary leap ahead of competitors, and other models also performed well. Despite this, the auto mode in Cursor was praised for being quite effective, offering a smooth coding experience without major complaints.
One notable observation was about GPT-5’s UI generation capabilities, which the creator found to be subpar compared to models from Anthropic. Although the UI generation wasn’t bad, it didn’t meet the higher standards set by some other AI tools. Another interesting feature introduced with GPT-5 was a router system that decides which model to use for specific tasks. While OpenAI admitted to having issues with this router, the creator believes this approach will become more common as it helps save costs and resources, making it a potentially valuable innovation if perfected.
Cost reduction was highlighted as the most significant improvement in GPT-5’s release. The creator sees this as a major win for OpenAI, as the model maintains performance without feeling worse than previous versions while significantly cutting resource usage like electricity. This efficiency is beneficial for both the company and users in the long run. However, the creator also expressed some nostalgia for the older models, which they had grown accustomed to using for specific tasks, noting that adapting to the new system requires some adjustment.
In conclusion, the video offers a balanced perspective on GPT-5’s capabilities and limitations. While the model and its integration with Cursor provide a solid coding assistant experience, it falls short of the hype and doesn’t dramatically outperform other models. The introduction of the router system and cost savings are promising developments, but users may need time to adapt to the changes. Overall, GPT-5 is a step forward in some respects but not a game-changer in the coding AI landscape.