Google’s Gemini Robotics ER 2 Gives Robots a Real-Time Brain

FEATURED INSIGHT💡

Gemini Robotics ER 2 Gives Robots a Real-Time Brain

Google DeepMind has released Gemini Robotics ER 2, an embodied-reasoning model designed to act as a high-level brain for robots. Instead of directly controlling every motor, it interprets streaming video, audio, and text; plans multistep work; calls tools such as navigation systems or lower-level vision-language-action models; and continues reasoning while the machine moves. That separation lets one model coordinate different robot bodies without replacing the specialized controllers that handle precise motion.

The practical advance is temporal awareness. ER 2 classifies progress throughout a task, identifies the exact moment a critical event occurs, and can retry or correct a step when the video feed shows that something went wrong. Google also demonstrated two robots using a shared semantic understanding to hand work back and forth. The model is in public preview through the Gemini API and Google AI Studio, with sample code available for developers.

For builders, this points toward physical agents assembled much like software agents: one reasoning layer orchestrates specialized tools, checks whether actions succeeded, and asks for help when needed. Google reports stronger safety instruction-following and human-proximity performance, but those results and the demonstrations are company-reported; reliability in messy homes, warehouses, and public spaces still needs real-world proof. Even so, giving robots a better sense of when a task is actually complete may be as important as teaching them the task itself.

One Account. Every Market. No Closing Bell.

Markets don't wait for Monday. News breaks on a Saturday morning, and most traders can do nothing but watch.

Not on Liquid. Trade domestic and international equities, commodities, forex, crypto, and prediction markets — all from one account, 24 hours a day, 365 days a year. Liquid gives you access to any market, from anywhere, anytime. To us, access is arbitrage.

Getting started takes under 10 minutes: log in with Google, deposit with Apple Pay or a bank transfer, and trade from your phone or desktop — wherever you are in the world.

While everyone else is refreshing headlines and waiting for the open, you're already positioned. That's the difference between reacting to markets and actually trading them.

Trade 24/7 on Liquid Today

ON THE HORIZON 🌅

AI Insiders Want a Global Brake Pedal Built Before It Is Needed

A statement called Pacing the Frontier now lists 1,324 employees of frontier AI companies among its signatories, including senior figures from OpenAI, Anthropic, Google DeepMind, and Meta. It asks the U.S. government to support an international effort to build technical and governance tools that could deliberately slow automated frontier-AI development if capabilities begin advancing faster than society can understand or control them.

This is not a demand for an immediate freeze. The argument is that governments and labs may eventually need the option to buy time, while competitive pressure makes any single company or country unlikely to slow down alone. The live tally rose from roughly 1,100 when the statement drew major coverage on July 28 to 1,324 during this research pass.

The hard part is turning a shared concern into a credible mechanism. Any pacing system would need broad participation, verifiable signals, and rules that do not simply reward actors willing to ignore them. For leaders, the nearer-term lesson is practical: decide in advance which capability jumps or incidents would trigger tighter access, more evaluation, or a temporary stop, because designing those controls during a crisis will be much harder.

LATEST IMPORTANT NEWS 📰

OpenAI Offers Free Frontier Access to 100,000 Researchers

OpenAI is launching ChatGPT for Academic Researchers, beginning with 10,000 participants this summer and planning to reach 100,000 scientists, mathematicians, and engineers through 2027. Eligible researchers at selected institutions can apply for free access to frontier models, Codex, larger context windows, research connectors, and training; OpenAI says program data is not used to train its models by default. The initiative could widen access beyond wealthy labs, though the impact will depend on which institutions and projects qualify.

German Court Rules Against Suno in Copyright Case

A Munich regional court ruled that AI music company Suno lacked the rights to process songs represented by German licensing group GEMA. The court ordered Suno to disclose revenue tied to the infringement and pay damages that have not yet been quantified. Suno said it disagrees and is considering an appeal, so the decision is not final, but it gives creators and publishers a significant early win in the fight over whether training an AI music system requires permission.

Chime Cuts Staff as AI Reshapes Its Organization

Fintech company Chime is cutting 10% of its workforce, affecting nearly 150 employees, as it moves toward smaller teams and fewer management layers. CEO Chris Britt told employees that AI is changing what is possible while requiring new skills, according to a memo reviewed by Reuters. The move adds a concrete workforce-reduction case to last week's broader evidence that most workplace AI use still augments individual tasks rather than automating entire jobs.

The best voice models, now with full orchestration. Build real-time voice and chat agents on one low-latency stack: any LLM, your tools and knowledge, testing, Guardrails, and omnichannel deployment.

Start building free

FOR THE TECHNICALLY INCLINED 🛠️

eGoT Turns Retrieval Into a Graph of Reasoning Paths

The enhanced graph-of-thoughts method, or eGoT, combines knowledge graphs, vector search, and iterative LLM reasoning for questions that require evidence from several sources. It first extracts entities and relationships from a document collection into graph and vector databases. For each question, the model generates, evaluates, expands, and merges multiple reasoning paths instead of relying on one similarity search or a fixed chain of thought. The peer-reviewed Bioinformatics paper reports stronger results than several graph-RAG baselines on MultiHopRAG, HotpotQA, and Ultradomain, then applies the method to 1,046 biomedical papers on small-cell lung cancer and to possible links between lupus and ultraviolet exposure. The code is open source, making eGoT a useful pattern for teams that need multi-hop answers with traceable evidence rather than fluent summaries alone.

AI TOOL OF THE DAY 🚀

Granola turns meetings into private AI-enhanced notes, action items, and searchable memory without inviting a bot, with free apps available for macOS, Windows, iOS, and Android.

See Why HubSpot Chose Mintlify for Docs

HubSpot switched to Mintlify and saw 3x faster builds with 50% fewer eng resources. Beautiful, AI-native documentation that scales with your product — no custom infrastructure required.

Simplify Your Docs Today