OpenAI Dev Day featured some promising innovations like the “Sign in with ChatGPT” token-sharing feature and an upcoming Decision API, but was largely overshadowed by controversial usage cuts, high pricing, technical glitches, and product offerings that closely resembled competitors’. The event left the community with mixed feelings, balancing excitement for new capabilities against concerns over transparency, originality, and value.
The OpenAI Dev Day event promised exciting new developments but ended up being somewhat underwhelming and controversial. Prior to the event, OpenAI announced a significant reduction in usage for their $200/month API plan, cutting it in half without lowering the price. This move frustrated many users who felt it was a bait-and-switch, especially since the company claimed the models were twice as efficient to justify the change, though no clear evidence was provided. To compensate, OpenAI introduced a $500/month plan with 25x usage and an “ultrafast mode” that is eight times faster but six times more expensive, allowing users to burn through their token limits very quickly.
The new product announcements largely consisted of features and applications that closely resembled existing tools from competitors, such as Muse (similar to Grok Bot) and Spaces (akin to Notion or Google Docs). The always-on agent concept was touted as a breakthrough, but the demonstration was marred by technical glitches and an unconvincing Wi-Fi excuse. Additionally, OpenAI lost the domain for “dots.com” to SpaceX, which redirected users to a competitor’s product, adding to the perception that OpenAI was trailing behind in some areas.
One genuinely innovative feature introduced was the “Sign in with ChatGPT” functionality, allowing users to share their API tokens with third-party applications securely. This approach could streamline the use of multiple AI-powered apps without the need to share sensitive API keys directly, potentially becoming a standard in the industry. This token-sharing model was seen as a forward-looking step that could foster a more integrated AI ecosystem, despite the other disappointments of the event.
OpenAI also announced the upcoming Decision API, a general-purpose classifier capable of handling images, but it was not immediately available, frustrating those eager to try it out. The company emphasized openness by integrating models like GLM 5.3 Flash and Kimmy K3 into their Codex platform, but there was skepticism about how “open” this truly was, given OpenAI’s control over which models are included and how they are managed. This raised questions about the balance between openness and proprietary control in the AI space.
Overall, the Dev Day revealed a mix of incremental improvements, questionable business decisions, and some promising innovations. While the ultrafast mode and token-sharing features showed potential, the event was overshadowed by usage cuts, product mimicry, and technical hiccups. The community’s reaction was mixed, with excitement tempered by concerns over pricing, transparency, and originality. The future of AI development with OpenAI remains intriguing but uncertain, especially as users weigh cost against capability and innovation.