I haven’t tried Instinct. What I’m wondering is whether Instinct is OpenClaw made easy to set up — prebuilt bundles of useful functionality, with someone else running the machine.
What’s public so far. TechCrunch reports Instinct is an AI personal assistant in private testing, built by a team led by a former Sierra research scientist. You reach it by text, WhatsApp or a phone call, connect it to your email, messages, calendar, location and screen, and it books the appointment and works the inbox. The same piece lays out the concerns: terms that grant the company a “perpetual and irrevocable” license to user materials for AI training, emails kept in plain text after disconnection, and the ability to send mail without approval.
That description is the shape of an OpenClaw, which runs on your own hardware and talks to you through your own chat apps. The difference is custody.
If that’s the case, I doubt I’d be in at even a $25M valuation.
I’ve heard favorable reports of Grok’s computer, though. That’s xAI’s Grok Bot, it seems, launched in beta on August 11: “your team of always-on agents” that “have their own computer, work inside tools and apps like you do, and keep working 24/7,” including on platforms with no clean API.
The other thing worth your hours, if you work in physical AI, is Anthropic’s new Model Hardware Standard. Anthropic opened a research preview yesterday and describes it as “a shared specification for AI agents to safely operate physical devices” — microscopes, liquid handlers, robotic arms — aimed at “scientific research labs and advanced manufacturers,” with device-level safety limits and a waitlist. It’s model-agnostic, and any agent harness can reach it through standard protocols such as MCP.
This is worth watching for your world. Anyone in physical AI with hours on it already?
