In partnership with
Auto-Generate Free SEO Audit Report
Most e-commerce sellers have no idea why their listings aren't ranking. Wrong keywords, missing metadata, weak product descriptions — the problems are there, but no one's told you where to look.
StoreClaw runs a full SEO audit across your Amazon and Shopify stores automatically. In minutes, you get a clear score, a breakdown of what's hurting your rankings, and exactly what to fix.
No manual review. No SEO agency. No guesswork.
Connect your store and StoreClaw surfaces every issue that's costing you search visibility — then tells you how to fix it.
Free to start. No credit card required.
TODAY IN AI
Click here to see Google’s new Gemini Robotics ER 2 model in action. Photo: Google
3 things that happened while you were busy
1. Google DeepMind launched the intelligence layer for robots.
On July 30, DeepMind released Gemini Robotics 2, which it calls "the intelligence layer" for the next generation of robots. It moves past tabletop arm demos into whole-body control: walking, crouching, stretching, and manipulating objects to clean a cluttered room, and robots can even team up to finish a job faster. The reveal ran on Apptronik's Apollo 2 humanoid.
2. It ships as three models, and one is available to you today.
The release is split into three: a vision-language-action model that turns sight and speech into motion, an on-device version that runs offline, and Gemini Robotics ER 2, the high-level planning brain, now in public preview via the Gemini API and AI Studio. The action models stay gated for partners, but the reasoning model is open for developers to start building physical-AI agents right now.
3. The killer feature is portability across robot bodies.
The quiet breakthrough: the on-device model adapts to an entirely new two-arm robot with fewer than 200 examples and a few hours of tuning, even across drastically different shapes and sensors. A skill learned on one robot can transfer to another. That is the difference between programming each machine and teaching a brain that any machine can borrow.
FROM THE FRONTIER
Two weeks ago we showed you the robot bodies. Google just shipped the brain.
The full stack. Our last robotics special covered the hardware: the centaur, the transformable humanoids, the IPO wave. The obvious missing piece was software smart enough to run them all. Gemini Robotics 2 is Google's answer, a general model that adapts to new hardware in hours instead of a specialist trained per machine. Bodies plus a portable brain is a functioning stack, not a demo reel.
The honest wall. Here is what the hype videos skip, and Google admitted it. Multi-finger dexterity is still the weak axis, ranging from 32% to 92% depending on the task, and Bloomberg noted the system still performs better with simple two-finger grippers than with humanlike hands. The brain is racing ahead; the fingers have not caught up. That gap is exactly why the laundry-folding future keeps slipping.
The business shock. The strategic tremor is for startups. By treating robotics as just another surface for the same Gemini model rather than a separate product, Google threatens to commoditize the very layer several well-funded robotics startups built their moat around. If general reasoning transfers to whole-body control, those startups get pushed to compete on hardware or deployment, not on the AI.
The takeaway. The pattern to hold onto: in robotics, the brain is arriving faster than the hands. Expect capable planning and navigation in commercial settings well before reliable fine manipulation in your home. Google also shipped a safety benchmark, ASIMOV-Agentic, that tests whether a robot will refuse unsafe actions and ask a human for help, the physical-world version of the agent guardrails we keep returning to. If your work touches logistics, inspection, or warehousing, the useful robots are coming to you first.
IN THE KNOW
What people are actually watching and sharing
The watering-can demo. The clip everyone is sharing shows Apollo 2 reasoning through a multi-step chore with full-body coordination. Watch the motion speed rather than the polish: smooth is easy to fake in a demo, fast and reliable is the real test.
A robot that says no. The overlooked release is ASIMOV-Agentic, a safety benchmark on Hugging Face that measures whether robots reject dangerous instructions. Naming it after Asimov's laws is on the nose, but the idea, a robot that declines, is the right one.
Robot Park. Apptronik quietly built Robot Park, a facility where Apollo 2 units generate the training data for the next model generation. Robots teaching robots is no longer a metaphor. It is a physical building in Texas.
The startup squeeze. Investors are reading Gemini Robotics 2 as a signal that big labs see robotics as a natural extension of frontier models, not a separate discipline. Expect robotics startups to lean harder on hardware and real-world deployment as their defensible edge.
PROMPT STATION
Separate robot hype from what will actually work for you.
The lesson of this launch is that robot brains are outrunning robot hands. That single fact should shape any automation decision you make this year. Paste this into Claude or ChatGPT and get a grounded read on which of your physical tasks are near-term realistic and which are still science fiction, so you plan around the real timeline instead of the demo.
You are a pragmatic automation strategist who separates hype from reality. My operation is: [DESCRIBE YOUR WORK OR BUSINESS]. Robots are advancing fast at whole-body movement and planning, but still weak at fine finger dexterity. Given that specific reality, split the physical tasks in my operation into three lists: tasks robots could plausibly do well within 1 to 2 years (movement, transport, navigation, inspection, simple gripping), tasks that need fine manipulation and will take much longer, and tasks that should stay human regardless. For the single best near-term candidate, describe what a realistic first deployment looks like, the rough cost range to explore it, and the one question I should ask any robotics vendor to cut through their sales pitchDescribe your operation plainly: "a mid-size warehouse", "a 30-table restaurant", "a poultry farm", "a hardware retail shop". Advanced tip: end with "now tell me the cheapest non-robot automation that solves 80% of the same problem today", because often software or a simple machine beats a humanoid on cost and reliability right now.