This week on X, OpenClaw's community served up three standout moments — a Mac computer-use breakthrough, a GitHub milestone that rewrote the record books, and a feature that finally makes your AI agent feel like a real conversation partner. Here's what you need to know.

1. Peekaboo 3.0: Your OpenClaw Agent Just Got Eyes AND Hands on Mac 👁️

Peter Steinberger (@steipete) shipped Peekaboo 3.0 on May 9 with a deceptively simple announcement: "I started this last year, but the models just weren't good enough. Now they are." The implications are enormous. Peekaboo v3 turns any OpenClaw agent into a full macOS computer-use operator — it captures pixel-perfect Retina screenshots, reads the full UI element tree (buttons, text fields, menus, window relationships), and performs real actions: clicks, typing, scrolling, dragging, window management, even navigating the dock and menu bar. It runs as either a CLI tool or an MCP server (npx -y @steipete/peekaboo), so it plugs natively into OpenClaw as a Skill and works with any underlying model — Claude, GPT-5, Gemini, or local Ollama. Within six days of v3.0, Peter shipped v3.2.0 adding MiniMax M2.7 and LM Studio vision support. Community reaction has called it the "missing piece" that completes OpenClaw's Mac stack: your agent could already think, talk, and act on your data — now it can see and operate your actual screen.

https://x.com/steipete/status/2053114837698249190

2. 371,000 Stars and Four Big Tech Sponsors: OpenClaw's GitHub Milestone is Staggering 🌟

@JulianGoldieSEO's thread on May 16 put the numbers in perspective: OpenClaw now sits at 371,000+ GitHub stars, making it the most-starred non-aggregator software repository on the platform — a title it claimed from React in March 2026 after blazing past the Linux kernel in February. The star history reads like a fever dream: 100,000 within 48 hours of its January 2026 relaunch, 224,000 by February, 250,000 in March, and now 371,000 in May. But what lit up the replies even more than the star count was the sponsorship news: OpenAI, NVIDIA, GitHub, and Vercel are now officially backing the project. With NVIDIA's NemoClaw enterprise layer, OpenAI's Peter Steinberger connection, and GitHub's platform infrastructure lining up behind it, OpenClaw is no longer just a community phenomenon — it has the institutional weight to match its viral momentum.

https://x.com/JulianGoldieSEO/status/2055580385489395924

3. Your OpenClaw Agent Can Now Join Your Discord Voice Channel and Actually Talk Back 🎙️

The v2026.5.9 beta (released May 9) shipped something the community had been requesting for months: real-time AI voice participation in Discord channels. The agent joins your voice channel, listens via speech-to-text, reasons with your chosen LLM, and responds out loud via TTS — with proper barge-in support so you can interrupt it mid-sentence and it immediately stops and listens. Voice Activity Detection means no wake word is required; just speak naturally. Three modes are available — STT/TTS, a realtime talk buffer with the OpenClaw brain, and full bidirectional realtime session — so you can tune the experience from "smart voice assistant" to "real conversation partner." Community reaction ranged from "this changes team standups forever" to detailed accessibility walkthroughs. The same release also added WeChat support, iMessage catch-up, and updated model support for Mistral Medium 3.5 and Gemini 3.1 Pro Preview — but the voice feature is what's getting all the screenshots.

https://newreleases.io/project/github/openclaw/openclaw/release/v2026.5.9-beta.1


Spotted something great on X about OpenClaw? Share it — the community is the best part.