xAI's AI coding tool Grok Build has released 11 versions in just 15 days, jumping from v0.1 to v0.2.11 at a pace far outpacing most competitors' half-year update cycles. The CLI-first Agentic Coding tool launched as an early beta on May 14, powered by the in-house grok-code-fast-1 model with a 256K token context window, supporting up to eight sub-agents running in parallel on separate Git worktrees.
Built-in Plan Mode is enabled by default, requiring user approval before code editing. Grok Build is the only AI coding tool currently integrating X platform real-time search, allowing developers to query community discussions and track API controversies directly within their workflow. The latest round of updates focuses on four areas: agent capabilities now allow sub-agents to share terminal backends and task scheduling across sessions, plus a new "laziness detector" and proactive reminders; compatibility adds Windows ARM64 and macOS x86_64 support, fixing Linux Wayland and WSL copy-paste issues; user experience sees terminal video playback boosted to 30FPS, multi-image paste and macOS screenshot shortcuts; stability increases default retry budget to about five minutes with better timeout handling for background tools.
Poaching Two Core People from Cursor
Version bumps alone weren't enough — xAI also targeted talent. The company hired Andrew Milich, engineering lead, and Jason Ginsberg, product lead, from Cursor parent Anysphere (valued at $50 billion), plus Devendra Singh Chaplot from Mistral AI. More striking: xAI signed an option agreement with Anysphere in April, allowing it to acquire the entire Cursor team and product for $60 billion in late 2026, with a $10 billion penalty if it backs out. Cursor already uses xAI's Colossus supercomputer in Tennessee for model training. Elon Musk admitted in March that "xAI didn't get it right the first time and is rebuilding from the foundation," and all 11 founding members have left.
SWE-bench Gap and AI Coding Market
Grok Build scored 70.8% on SWE-bench Verified, while Anthropic's Claude Code hit 87.6% and OpenAI's Codex CLI reached 85-88.7% — a gap of roughly one generation. However, xAI's strategy isn't just about benchmark scores: eight sub-agent parallelism is the highest among similar tools, and Grok Build can directly read Claude Code users' skills, plugins, and hooks configuration files, offering zero migration cost. One Medium tester found Grok Build automatically loaded 47 skills on startup, 12 of which came from Claude Code's configuration directory.
The AI coding market in 2026 has reached $12.8 billion, with 84% of developers using or planning to use AI tools, averaging 3.1 tools per person. Cursor leads with $2 billion annual revenue and 1 million paid users; Claude Code holds 28% primary tool market share and 46% "favorite tool" votes; GitHub Copilot has 4.7 million paid subscribers.
Grok Build pricing follows xAI subscriptions: SuperGrok at $30/month, X Premium+ at $40/month provide access, while SuperGrok Heavy at $300/month ($99 for the first six months) unlocks full capacity. Arena Mode (multi-agent competitive auto-scoring) is touted as a flagship feature but remains unreleased as of v0.2.11.