AIE Europe Day 2: ft Google Deepmind, Anthropic, Cursor, Factory, Linear, HF, Cerebras & more

AI Engineer Europe Day 2 highlights the shift from simple AI assistance to complex agentic workflows and the resulting engineering challenges. Google’s Gemma 4 demonstrates the power of open, on-device models, while Anthropic’s Model Context Protocol (MCP) establishes a standard for agent connectivity across diverse applications. However, rapid code generation introduces significant risks, including "slop" and compounding technical debt. Experts emphasize that "friction"—human judgment and slow, deliberate review—is essential to maintain software quality. Innovations like Cursor’s use of Markdown as code and Cerebras’s 1,200 token-per-second inference require developers to evolve into "agent choreographers." Ultimately, the bottleneck in software development has shifted from writing code to human attention and product taste, necessitating a move toward "agent-legible" codebases and zero-bug policies to ensure long-term system reliability.