Paying for GPT Plus, but it feels like I’m testing an unfinished beta product.
Experiencing Frustration with GPT Plus: Is It Still a Premium Service or Just an Unfinished Beta?
As an active user of ChatGPT Plus, I’ve relied on the service heavily for my daily professional tasks. However, recent updates and changes have left me questioning whether I’m genuinely receiving the premium experience I’ve paid for, or if I’ve become a beta tester for an incomplete product. Here, I’d like to share my observations and solicit insights from others experiencing similar frustrations.
The Impact of GPT-5 Release on User Experience
Since the rollout of GPT-5, my interaction quality has noticeably declined. Initially, I was pleased that the older GPT-4o version was reintroduced, hoping it would offer a more refined experience. Unfortunately, it mirrors GPT-5 in performance—characterless responses, sluggish thinking, and a lack of depth. The key difference I’ve observed is that GPT-5’s search capabilities seem more precise, producing fewer hallucinations. Aside from that, the improvements are minimal at best.
Loss of Contextual Memory and Reduced Competence
Previously, GPT-4o effectively utilized stored information, allowing for more coherent and context-aware conversations. Now, it appears to ignore internal memory, promptly dismissing stored data even when explicitly asked to recall it. While the model may list saved items, it doesn’t incorporate that information into subsequent responses—this issue is apparent from the very first exchanges.
Decline in Cognitive Capabilities
I’ve also noticed a significant drop in the model’s ability to handle complex, multi-threaded thinking. Early versions of GPT could manage extensive conversations or brainstorming sessions with relative ease. Presently, I must explicitly spell out each step because the model struggles to connect ideas logically or follow nuanced threads.
Technical Failures and Usability Issues
Functional issues have become routine. Tasks such as reading text from images or PDF files now frequently fail—something that previously worked reliably. Additionally, the voice mode feature, which was introduced with high hopes, has been virtually unusable since its launch. It frequently cuts out or interrupts mid-sentence, turning what should be a seamless experience into a frustrating ordeal.
Increasingly Concerning Hallucinations
Alarmingly, GPT-5 seems more prone to “hallucinations” than GPT-4o. Instead of improving, it has become more stubborn about sticking to inaccuracies, repeatedly returning to prior mistakes despite correction attempts.
Post Comment