GPT-5’s launch is not going well… Are we entering the AI “Trough of Disillusionment"?

GPT-5 Launch Reception

• ⚠️ The launch of GPT-5 has been met with a tepid response, feeling more like an incremental upgrade rather than a revolutionary breakthrough.
• 💡 Many users and experts note that the improvements are less striking due to the progress made by competitors like Claude, which offer comparable performance.
• 💬 Some suggest the name "GPT-5" might have been reserved for a more spectacular model given its brand equity.

Performance Benchmarks and Initial Feedback

• 📊 GPT-5 shows improvement over previous GPT models, particularly in coding capabilities, scoring 74.9% on the software benchmark verified test.
• ⚡ However, anecdotal feedback from the engineering community is mixed, with some engineers finding GPT-5 too slow and preferring Claude Sonnet for agentic tasks.
• 📉 Comparisons on Agentic tasks show GPT-5 not performing as well as Claude Opus.

Expert Reviews and Capabilities

• 🧪 Reviews from teams like Every.to found GPT-5 good at everyday tasks, pair programming, and writing, but it fell flat in agentic engineering and editing.
• 🎨 In contrast, Latent Space's "vibe tests" highlighted GPT-5's ability to solve complex dependency conflicts and excel at "vibe coding apps," creating functional features with surprising detail.
• 🚀 Latent Space reviewers praised GPT-5 as "unequivocally the best coding model in the world," suggesting a significant leap in software engineering automation.

Criticisms and Limitations

• 🗣️ Former Uber exec and AI critic Gary Marcus described the launch as "shambolic," arguing GPT-5 still struggles with fundamental problems like errors, hallucinations, and a lack of reasoning and generalization.
• 🧠 Marcus contends that the "pure scaling approach" has reached its limits, and the core architecture is insufficient for achieving AGI, advocating for new approaches like neuro-symbolic AI.
• 🔬 A study from Arizona State University cited by Marcus indicates that even the latest models, including GPT-5, fail to generalize well outside their training data.

OpenAI's Market Position

• 📈 Despite the mixed reception, OpenAI still dominates the B2B space through its APIs, though reports show a slight decline in market share.
• 📱 For consumers, ChatGPT remains the number one AI product, with 700 million weekly active users, and has become a "verb" in daily vocabulary.
• ⚖️ Elon Musk's XAI plans to sue Apple over the "excessive unfair promotion" of ChatGPT, highlighting its strong market presence.