# I was wrong... **URL:** \ **Category:** Content Creators\ **Tags:** gemini-4, theo, deepmind, google, tech, google-deepmind\ **Created:** [13 November 2026 23:45 UTC](https://www.artofsm.art/t/i-was-wrong/28407 "2026-11-13T23:45:53Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [13 November 2026 23:45 UTC](https://www.artofsm.art/t/i-was-wrong/28407/1 "2026-11-13T23:45:49Z")
[![](https://www.artofsm.art/uploads/default/original/3X/3/8/386ae3a2b1258c71ef8d270a76b78de7d2c1b481.jpeg "I was wrong...") ](https://www.youtube.com/watch?v=5BrgFY5jIn4) The creator admits to underestimating the severe performance issues of the Gemini 4 Carbon model, revealing it to be significantly slower and less efficient than competitors, making it impractical for real-world use. They criticize DeepMind’s leadership and infrastructure, urging major improvements while recommending users avoid Gemini in favor of better alternatives until these problems are resolved. ---
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [14 November 2026 00:10 UTC](https://www.artofsm.art/t/i-was-wrong/28407/3 "2026-11-14T00:10:12Z")
In this video, the creator admits to being wrong about the Gemini 4 Carbon model, initially expecting to be highly critical but finding some value in its lack of restrictions. However, after deeper real-world testing, they discovered that the model is significantly slower than anticipated—about five times slower than other comparable models. This severe lag, especially in multi-step tasks common in agent coding tools, makes the model practically unusable. The creator expresses frustration with Google’s apparent disregard for the user experience and infrastructure quality surrounding Gemini, emphasizing that until these issues are addressed, the model should not be taken seriously. The video highlights the extreme latency issues with Gemini 4 Carbon, noting that it takes over 11 seconds just to start processing a request, which is far longer than other models take to complete entire tasks. This delay compounds with each step in multi-turn interactions, making the overall experience frustrating and inefficient. The creator also points out poor caching behavior and inefficient file reading, which further degrade performance. Despite some initial hope that Gemini could be a viable option, the reality of its sluggishness and high cost—especially with upcoming price increases—makes it a poor choice for real-world engineering work. The creator also criticizes the leadership and communication from DeepMind, particularly calling out a figure named Verun for poor public engagement and leadership. While many individuals at DeepMind have been responsive and helpful, the overall management and product quality fall short of expectations. The video contrasts Gemini’s issues with other AI models and infrastructure improvements seen at companies like Microsoft and Meta, where similar problems were addressed more effectively. The creator urges DeepMind to use this critique as leverage to improve their models and infrastructure or risk losing credibility and talent. Throughout the video, the creator expresses deep disappointment and anger, not only at the technical shortcomings but also at the broader organizational failures that allow such a product to be released and maintained in this state. They emphasize that the problems are not just with the model’s intelligence but with the entire ecosystem around it, including hosting, caching, and API design. The creator warns against using Gemini models in their current form and suggests that users would be better off subscribing to alternatives like Claude, which offer better performance and value. In conclusion, the video serves as a candid and blunt critique of Gemini 4 Carbon and DeepMind’s handling of the product. The creator hopes this transparency will push for necessary improvements and help users make informed decisions. They also encourage those frustrated with Gemini’s shortcomings to consider opportunities at other companies that prioritize quality and user experience. Until significant changes are made, the creator strongly advises against using Gemini models for serious development work, highlighting the model’s current state as a “dumpster fire” that fails on speed, cost, and usability.