User ceased subscription to ChatGPT after a span of more than two years
In a groundbreaking development, OpenAI has announced the release of GPT-5, a new and improved version of their language model. This latest iteration marks a significant leap forward in artificial intelligence, offering enhanced efficiency, longer context windows, stronger multimodal understanding, improved accuracy, and practical integrations for productivity.
Previously, the writer had been relying on ChatGPT, but found it wanting, particularly when it came to tasks requiring source links and explanations of technical concepts. The writer experienced inaccuracies and issues that they initially attributed to their own inability to prompt the bot effectively or outdated training data.
However, the emergence of GPT-5 promises to address these shortcomings. One of the key improvements of GPT-5 over previous versions like ChatGPT (GPT-4) is its hybrid multi-model architecture. GPT-5 dynamically selects from specialized sub-models in real time based on task complexity and speed needs, improving efficiency, reasoning depth, and response time.
Another significant advancement is GPT-5's ability to handle larger contexts. It can process more than 256,000 tokens, enabling much longer conversations and documents without losing context. This vast increase over GPT-4 is a game-changer, particularly for journalists like the writer who value the written word highly.
GPT-5 also boasts stronger multimodal capabilities, unifying text, images, voice, and live video understanding and generation. This means it can interpret and reason about a wide range of input types with higher accuracy, including photos, charts, and videos.
In terms of reasoning and accuracy, GPT-5 shows dramatically reduced hallucinations, with up to 78% fewer errors than prior versions. It excels in extended reasoning tasks such as math, coding, and complex problem-solving. For example, it scores 94.6% on the 2025 AIME math exam without tools.
Moreover, GPT-5 is OpenAI’s strongest coding model to date, excelling in generating and debugging complex front-end code and larger software repositories. It also demonstrates improved "vibe coding" that produces responsive apps and websites from a single prompt.
GPT-5 is also designed to perform multi-step tasks more reliably, integrate with tools like Gmail and Google Calendar to assist with scheduling and reminders, and adapt dynamically as instructions change.
The writer, who has started using Gemini more for tasks that require finding source links and explaining technical concepts, will be testing GPT-5 to see if it can provide a major improvement over previous models. The writer's decision to unsubscribe from ChatGPT Plus was not sudden, but a result of feeling that they were not getting what they needed from the service.
Meanwhile, the writer's visits to LinkedIn have become a game of spotting AI-generated content. The writer cross-checks responses from different chatbots, including Gemini, with other chatbots for verification.
Interestingly, a friend convinced the writer to try Kagi Assistant, and they have started using it instead of ChatGPT. Kagi Assistant lets the user use all search models in one place and does a good job of replacing Google Search.
ChatGPT still has "tells" that make it easy to identify as AI-generated, such as overuse of colons, forced metaphors, and certain sentence structures. However, the writing produced by ChatGPT has improved over time, with many people consuming AI-generated content without realizing it.
In conclusion, the release of GPT-5 represents a major step forward in artificial intelligence, combining greater efficiency, longer context windows, stronger multimodal understanding, improved accuracy, enhanced coding, and practical integrations for productivity. It moves beyond ChatGPT’s primarily text-based conversational AI into a unified, dynamic, multimodal system.
The writer, having found ChatGPT insufficient for tasks requiring source links and explanations of technical concepts, is eager to test GPT-5 for potential improvements in these areas. (A statement about the writer's intention)
With GPT-5's stronger multimodal capabilities that unify text, images, voice, and live video understanding, it shows great promise in handling a wide range of input types with higher accuracy. (A statement about GPT-5's potential capabilities)
