Grok 4.3, Grok Skills, and Grok Build land with new Vercel, Canva, Gamma, and S&P connectors

xAI's May releases form a coherent push to turn Grok from a chat product into a developer and productivity surface. Grok 4.3 (May 4) is positioned as a cost-efficient flagship with reasoning chains and video input. Grok Skills (May 18) lets users define persistent custom expertise that survives across sessions, addressing the long-standing complaint that chatbots forget context. Grok Build 0.1 is xAI's coding-agent model, going head-to-head with Claude Code and OpenAI Codex.

Mechanically the most interesting piece is the OpenClaw integration (May 19), which gives Grok subscribers access to chat and generation capabilities inside the open-source agent platform that crossed 300K GitHub stars before being overshadowed by Google Spark. The new third-party connectors (Vercel for deploys, Canva for design, Gamma for slides, S&P Global for finance data) are clearly an attempt to match Anthropic's MCP ecosystem reach.

Competitive context: xAI's productivity push lands the same week we learned Anthropic is paying xAI $1.5B/month for compute — Musk's lab is simultaneously a frontier developer, a coding-agent competitor, and a compute landlord. The connector strategy mirrors what every major lab is doing, but Grok's distribution advantage (X.com) is unique and the Tesla in-car deployment angle keeps Grok closer to consumer surface than its rivals.

What to watch: independent benchmarks for Grok 4.3 vs. GPT-5.5, Claude Opus 4.7, and Gemini 3.5 Flash; whether Grok Build can pull mindshare from Claude Code; and whether Skills' persistent-expertise model becomes a moat or gets copied within a quarter. Musk's '💯' tweet (329K likes) is content-free but signals he thinks the month went well.