xAI has begun publishing daily build updates for Grok, its AI coding agent, marking an unusual transparency move in an industry typically shrouded in secrecy. Elon Musk recently directed his 200 million-plus followers to a dedicated tracking page where developers can monitor real-time progress on Grok Build, the company's specialized coding assistant. This shift signals that xAI is treating Grok's development more like an open-source software project, with visible, timestamped changes, rather than following the opaque release cycles common among AI labs.

What Is Grok V9 Medium, and When Will It Arrive?

According to reporting from May 25, Musk confirmed that xAI has completed training on Grok V9 Medium, a foundational model containing 1.5 trillion parameters, which is three times larger than the current v8 small model at 0.5 trillion parameters. The company is currently running supervised fine-tuning, a process that teaches the model to follow specific instructions more reliably. Reinforcement learning training, which further refines the model's behavior based on human feedback, is expected to begin imminently. A public release is targeted for roughly two to three weeks from that May 25 announcement, placing the launch window in mid-June 2026.

To understand what these technical steps mean: fine-tuning is like teaching a student to answer questions in a specific way, while reinforcement learning is like rewarding correct answers to shape better future responses. Together, these processes transform a raw language model into a polished product ready for public use.

How to Track Grok's Development Progress

• Daily Build Page: Visit xAI's dedicated changelog to see timestamped updates on Grok Build improvements, allowing developers to monitor changes as they happen rather than waiting for major announcements.
• Real-Time Visibility: The public tracker offers a window into iterative improvements that coding agents typically ship far more frequently than consumer-facing models, giving power users insight into what's changing under the hood.
• Developer-Focused Updates: The daily build page tracks Grok Build specifically, the coding agent designed for developers, rather than the flagship chat model, meaning updates reflect practical improvements for coding workflows.

Why Is Public Transparency Unusual in AI Development?

Most AI labs, including competitors like OpenAI and Anthropic, share little between major version announcements. They typically release new models with a press release and benchmark results, then remain quiet until the next major update months later. xAI's decision to publish daily build updates breaks this pattern significantly. For developers and power users already relying on Grok for coding tasks, the public changelog offers unprecedented visibility into what's being fixed, optimized, or added to the system.

This transparency approach carries strategic implications. By making development progress visible, xAI can demonstrate momentum and responsiveness to user feedback, potentially building trust and adoption among developers who value seeing tangible improvements. It also puts competitive pressure on rivals to either match the cadence or publicly explain why they won't.

What Does This Mean for the AI Coding Agent Market?

Coding agents have become a critical battleground in AI development. Companies like GitHub, Anthropic, and OpenAI all offer specialized tools for developers, and the pace of improvement matters significantly to users who rely on these tools daily. By accelerating both its development pace and its communication around it, xAI appears to be signaling that Grok will compete aggressively in this space.

The combination of a larger model (1.5 trillion parameters versus 0.5 trillion in v8 small), visible progress tracking, and a near-term public release suggests xAI is betting on transparency and rapid iteration as competitive advantages. Whether this strategy resonates with developers will become clear once Grok V9 Medium launches in mid-June and users can evaluate its actual performance on real coding tasks.