In partnership with
News of the day
1. Gemini Robotics 2 showcases advanced AI robot dexterity, tackling complex tasks like packing, sorting, and even tying knots with precision and adaptability. → Read more
2. New open-source Token Saver extension for Claude Desktop drastically cuts PDF token costs by 90-99% using local hybrid RAG. → Read more
3. Lilian Weng, co-founder of Thinking Machines, steps down citing health issues but rejoins OpenAI to lead a team on recursive self-improvement. → Read more
4. Microsoft AI CEO Mustafa Suleyman reveals a strategy shift towards cost-effective specialist AI models over expensive general-purpose ones. → Read more
Our take
Hi Dotikers!
Google DeepMind just released Gemini Robotics ER 2, and the announcement should be read for what it is: Google isn't building robots, Google is building the brain that drives them.
The model doesn't move a single motor. It watches a video feed, understands the space, plans the steps, then hands off to a low-level action model for execution. In practice, the motor layer becomes just another tool, called the same way you'd call Google Search or a custom function. All of it wired into the Live API with bidirectional streaming, so there are no "hold on, thinking" pauses in the middle of a gesture. Add multi-robot collaboration, a Boston Dynamics Spot fetching popcorn on a voice command, an Apptronik Apollo 2 coordinating its whole body, and the picture gets clear: robotics is being absorbed into the same agentic pattern we already know from software. An orchestrator, tools, a verification loop.
Where it gets interesting is the numbers Google publishes without dressing them up. 91.3 % for pinpointing the exact moment an event happens in a video, solid. But 57.4 % on tracking task progress, and that's presented as the best score on the market. In other words, knowing whether the job is done is still close to a coin flip. A robot flatmate with a one in two chance of knowing it's finished screwing in the light bulb: let's just say we're keeping the stepladder within reach.
That's exactly what was missing from yesterday's story. Jensen Huang promises 100 billion agents and billions of robots, and bets the constraint will be physical: memory, data centers, gigawatts. These benchmarks tell a different story. The bottleneck isn't the foundry, it's temporal understanding. We know how to build a robot that acts. We don't yet know how to build a robot that knows it's done.
Alex.
How owning AI deployment expands your career
Across product, ops, and CX teams, a new kind of role is taking shape: the person responsible for making AI actually work, day to day. In this roundtable, three people living this shift share what it's really like: Simone Santiago Broad (Yoco), Yelva Espinoza (Zumba Fitness), and Fin's Dave Lynch. You'll hear how they carved out these roles, what the job looks like across industries, the skills they'd hire for, and the challenges they're tackling right now.
Watch the full conversation on demand.
Meme of the day
