Conference highlights loops, software factories, and forward deployed engineers as agentic workflows take center stage

On the second day of the AI Engineer World’s Fair, the dominant motif was the loop—iterative workflows that chain agents into persistent, goal-directed automations. Keynote speaker swyx titled his talk “Loopcraft: The Art of Stacking Loops,” tracing the evolution of AI engineering from chat interfaces in 2022 to tooling, goal-seeking agents, and now automations built from layered loops.

Talks repeatedly tied loops to the emerging concept of software factories, defined by Tereza Tížková as “the whole loop, the whole lifecycle of developing software with autonomy.” These factories are framed as end-to-end systems that ingest signals, react to user feedback and logs, prioritize tasks, and orchestrate agents across coding, review, and deployment. Zach Lloyd of Warp argued that “software engineering will become factory engineering,” predicting that developers will increasingly “build the thing that builds the product.”

A parallel theme was the rise of forward deployed engineers (FDEs), sometimes called agent engineers, who specialize in integrating agentic systems into organizations. Natalie Meurer, Head of Agent Engineering at Sierra, described FDEs as focused on orchestration layers rather than model-level work, noting that “most customer-specific work takes place at the orchestration layer.” Cursor’s Pauline Brunet positioned FDEs as partners in co-designing and co-building AI software factories to transform how organizations design, develop, and maintain software across the lifecycle.

Open models and local AI deployment were cited as foundational to scalable agent ecosystems. Zixuan Li of Z.ai highlighted the company’s open LLM GLM-5.2 as a “flagship model for long-horizon tasks” and introduced ZCode, a harness positioned as a counterpart to OpenAI’s Codex. Ahmad Osman of Osmantic argued that open models have improved rapidly, with “architectures becoming more efficient and many small improvements compounding,” enabling local AI systems to reproduce frontier capabilities more efficiently.