I tested Grok Bot, Grok 4.6, and Cursor Origin - here’s my honest take

The host discusses the shifting landscape in AI and collaboration tools, highlighting the recent acquisitions and product releases from Cursor, SpaceX, and other key AI players. The focus is on the growing interest and improvements in models like Grok, especially with Cursor's integration into the SpaceX ecosystem, and on the new products designed for enterprise AI workflows.

GROK Bot is introduced as a user-friendly, multi-agent desktop and mobile chat assistant that integrates multiple connectors and accounts, such as Gmail and Slack, into a single interface. The host emphasizes its plugin architecture, ease of setup, and the virtual machine feature that allows each bot to operate with a degree of independence. The host shares personal use cases for various GROK bots and discusses its straightforward onboarding process, highlighting its simplicity and the plugin ecosystem that makes it adaptable.

While praising GROK Bot’s ease of use and connector capabilities, the host criticizes its limited customizability, lack of control over underlying models and configurations, and the generic, less engaging 'vibe' of its AI personality. These limitations are seen as barriers for users who prefer highly tailored or chaotic setups like open Claws.

Cursor Origin is presented as an early-stage, agent-native GitHub replacement that aims to integrate code repositories, pull requests, and CI/CD extensions within the Cursor environment. The host experiments with its features, noting that while it offers promising primitives like importing repositories and managing pull requests, it currently feels like a wrapper over GitHub rather than a revolutionary native experience. The discussion reflects on the potential of Cursor to become a sort of central code command center, but also notes current limitations and the need for more compelling features.

The host tests the GROK 4.6 model against various benchmarks, including a proprietary Claire Index and design/evaluation tasks, comparing it to models like GPT-5.6, Claude, Opus, and Sonnet. Results show GROK 4.6 performing well in certain design and UI tasks, especially when following explicit instructions, and ranking highly on the Claire Index alongside GPT-5.6. The host discusses the nuanced differences in performance depending on the task and how personal taste influences these ratings, ultimately viewing GROK 4.6 as a competitive and promising model.

The host shares their overall impressions, endorsing GROK models as worth trying and highlighting their potential, especially when used with Cursor's ecosystem. Despite current limitations, the host is excited about the future updates and sees GROK as a serious competitor in the AI model space. The episode concludes with reflections on commitment to certain workflows, and an invitation for listeners to share their opinions about GROK and Cursor’s evolving ecosystem.