The year 2026 opens with a subtle yet significant fracture: nostalgia for digital craftsmanship. While in 2023 attention was focused on the infinite scalability of generative models, today we witness a return to the aesthetic of ‘handmade’ in the virtual world, an unexpected parallelism with the boom of emotional luxury in China. Zhu Wenqian’s report on the Chinese market reveals a preference for mascots and zodiac objects with a ‘soft’ and playful design, an implicit rejection of hyper-rationalization and algorithmic perfection. This is not just a simple consumption trend; it is a symptom of a deeper dissatisfaction with a world increasingly mediated and predictable, a desire to reconnect with imperfection and authenticity.
The Architecture of Digital Empathy
China’s quest for ’emotional fulfillment’ in consumption is not an isolated phenomenon. It intersects with the push towards personal and contextual artificial intelligence (AI) promised by Apple with iOS 26.4. The evolution of Siri, with its ability to understand context and act across multiple apps, represents a bid to infuse ‘humanity’ into an interface traditionally cold and functional. However, this quest for personalization raises fundamental questions about the architecture itself of AI. If the goal is to create machines that understand and respond to our emotions, we must confront the problem of representing subjectivity. Large language models (LLMs) are inherently based on statistical probability, predicting the next token in a sequence. How can we translate emotional experience, with its complexity and ambiguity, into a format comprehensible for an algorithm? The answer may lie in the development of ‘World Models,’ internal representations of the world that incorporate not only factual data but also causal inferences and emotional simulations. However, these models are subject to intrinsic biases, reflecting the preferences and prejudices of their creators. The challenge is not just building intelligent machines, but ensuring they are intelligent in a way aligned with our values and aspirations.
The Paradox of Efficiency and Indiana’s New Strategy
The debate on the optimal size of AI models, raised by Sridhar Vembu, highlights a crucial tension between efficiency and capability. While large tech companies invest in ever-larger LLMs, Vembu argues that India should focus on smaller, more specialized models tailored to its specific needs. This strategy reflects an increasing awareness of the limitations of ‘general purpose’ models and the need for a more pragmatic and sustainable approach. The race to giant LLMs risks creating an even wider digital divide, concentrating technological power in the hands of a few companies with the resources to train and maintain these models. An alternative approach based on decentralization and specialization could democratize access to AI and promote local innovation. This vision aligns with the idea of embedded AI, integrated into specific systems and adapted to particular contexts. As highlighted by Microsoft’s CEO, ‘vibe coding’ represents a step in this direction, lowering the barrier to entry for developers and enabling the creation of more personalized applications with greater ease.
"Vibe coding is making apps easier to build."
This could lead to a proliferation of ‘micro-apps’ and specialized services, fostering a more dynamic and diversified innovation ecosystem.
Beyond Enthusiasm: The Fragility of the New Order
In the coming six months, we will witness increasing polarization in the AI sector. On one hand, large tech companies will continue to invest in increasingly powerful LLMs, seeking to maintain their dominant position. On the other hand, new enterprises and open-source communities will emerge, focusing on smaller, more specialized, and accessible models. The competition between these two approaches will define the future of AI and its impact on society. It seems clear that initial enthusiasm for ‘generative’ AI is giving way to a greater awareness of its limitations and risks. The real challenge is not just building intelligent machines, but ensuring they are used responsibly and sustainably, for the benefit of all. We are entering an era more mature and less euphoric, where innovation will be guided not only by technology, but also by ethical, social, and environmental considerations.
Photo by Amos K on Unsplash
Texts are autonomously processed by AI models