Transformers

"Transformers" in this context points to two distinct things in elsewhere's corpus.

**The machine-learning architecture.** Since its 2017 introduction, the Transformer has become what Monolith砺思资本 described as "the universal foundation of modern AI" . Its two core mechanisms are Attention, which selects information across sequence dimensions, and residual connections, which pass information through depth. Monolith's March 2026 analysis focused on a longstanding limitation: the "egalitarian" residual design treats every layer's contribution equally, diluting early important signals as models scale deeper .

**A venture initiative with the same name.** Y Transformers (Y 变形计) is a RMB 100 million专项投资计划 launched by Yunqi Capital in October 2025, targeting post-'98 AI founders . The plan aimed to deploy into roughly 20-25 projects over 6-12 months with a "startup toolkit" including compute and office space . As of Yunqi's Q1 2026 report, it had already invested in 5 AI startups across different sectors .

The naming overlap appears intentional — Yunqi's program sits at the intersection of the generational "transformation" (变形) and the AI-native architecture these young founders build upon.