How We All Underestimated Grok Build and Why Cursor May Soon Be Obsolete

The article suggests that xAI is testing a new model, possibly Grok 4.6, under the Grok 4.5 name on the LLM Arena, based on community observations. Simultaneously, xAI is assembling a suite of tools: Grok Build, Workflows, Grok Bot, and Grok Imagine Image 2.0. Grok Build is an open-source Rust TUI and agent runtime for coding, supporting interactive and headless modes, MCP, and permissions. Workflows allow parallel execution of up to 128 agents, or 1024 for large tasks, for tasks like PR review or deep research. Grok Bot is an always-on cloud-based agent that can work in browsers and SaaS interfaces, with routines that automate recurring tasks. Grok Imagine Image 2.0 offers precise editing tools like magic wand, segmentation, background removal, and multi-reference editing, enabling asset generation for video and app building. The author emphasizes that Cursor's data, including trillions of tokens of developer interactions, were used to train Grok 4.6, but notes the contamination risk. The article argues that time-to-verified-result is more important than tokens per second, and that Grok's low cost per task (around $2.49 per task versus $5.07 for GPT-5.5 in Codex) makes it competitive. The author concludes that this stack could disrupt tools like Cursor, potentially making them obsolete.