CNET published a feature-length explainer today positioning “claw” not as a synonym for OpenClaw but as a new computing category in its own right, with multiple companies now building, shipping, or wrapping their own agent variants. The piece quotes NVIDIA CEO Jensen Huang at GTC 2026 calling claws “the new computer” and argues the category has moved past the single-product phase.
“First there was chat, then there was code, now there is claw,” CNET opens with an Andrej Karpathy post from February. The framing is significant: the largest consumer tech publication in the US is no longer treating OpenClaw as a novelty. It’s treating it as a platform class.
Beyond OpenClaw: The Vendor Ecosystem
The CNET piece names multiple “claw” vendors beyond Peter Steinberger’s original OpenClaw. Gavriel Cohen, creator of NanoClaw and CEO of NanoCo, told CNET that these are “general-purpose computer agents. Anything that a person can do with a computer, an agent can do.” Cohen credited Steinberger with connecting models to tools in a way that unlocked “YOLO mode — do anything.”
The distinction CNET draws between agents and claws is specific: a standard AI agent waits for a prompt, while a claw can wake itself up autonomously, react to environmental triggers, maintain memory across sessions, and operate tools like browsers and terminals without step-by-step scripting.
GTC 2026 Validation
Jensen Huang’s GTC statement — “Every company in the world today needs to have an OpenClaw strategy, an agentic system strategy” — has been widely cited since March. What’s new in the CNET piece is the editorial decision to frame Huang’s endorsement as validating a category, not just a product. When the CEO of the company supplying the hardware layer publicly declares something “the new computer,” the downstream effects ripple through procurement, strategy, and investment decisions.
Category vs. Product
The shift from product coverage to category coverage matters for the competitive landscape. When CNET wrote about ChatGPT in early 2023, the coverage was about one product. Within six months, “chatbot” and “LLM” had become categories with dozens of entrants. The same arc appears to be happening with claws: OpenClaw originated the concept, but NanoClaw, NVIDIA’s NemoClaw, and others are now defining their own implementations.
For builders, the implication is that “claw compatibility” may become a requirement the way “ChatGPT API compatibility” drove the model-wrapper wave of 2023-2024. For enterprises evaluating agent platforms, the question is no longer “should we use OpenClaw?” but “which claw architecture fits our stack?”