Google Launches Gemini Robotics 2 for Whole-Body AI
ALSO: AI breaches hit $6M, coding drives 70% of revenue
Welcome to Guru99 AI Report!
Top Story: From robots learning to move like us, to teens confiding in chatbots, to why the market now rewards efficiency over raw power — this edition unpacks the shifts worth watching. Grab a coffee; there’s plenty below.
🤖 Google Gives Robots Whole-Body Intelligence
Brief Buzz:
Google DeepMind has introduced three Gemini Robotics 2 models designed to help robots understand instructions, plan tasks and move with more of their bodies. The headline advance is whole-body control — but Google’s own tests show dexterous hand movements remain uneven.
- Gemini Robotics 2 turns visual and language inputs into motor commands, controlling humanoids from feet to fingertips.
- ER 2 handles high-level planning, tracks multi-step jobs and coordinates multiple robots, according to Google.
- On-Device 2 runs locally and can adapt to new two-arm robot bodies with under 200 demonstrations, Google says.
- Google reports task success from 32% to 92% in five-finger tests, showing human-like dexterity is still far away.
- The FCC’s new restriction blocks authorization for most new foreign-produced mobile robots; approved existing models and conditional exemptions remain.
💡 Why Should You Care?
More adaptable robots could eventually take on varied warehouse, factory and home tasks. For now, access is limited, performance is inconsistent, and U.S. policy may narrow hardware choice.
🛡️ AI-Enabled Attacks Cost 20% More, IBM Finds
Brief Buzz:
AI is rewriting the economics of hacking. Attacks are getting cheaper to launch. Breaches are getting pricier to fix. IBM’s new report puts hard numbers on that gap — and it’s widening fast.
- AI-enabled breaches now average $6 million — roughly 20% more than the $4.99M global average, itself a record high.
- One in four malicious breaches used AI last year — a 56% surge over the year before.
- Deepfake impersonation and AI-generated malware led the attacks, and 62% hit critical infrastructure like energy and finance.
- Organizations that lean on AI defensively saved nearly $2 million per breach — but one in four still use none.
- Only 18% aim AI at vulnerability management — leaving the exact gaps attackers now exploit fastest.
💡 Why Should You Care?
Bottom line: thinking “it won’t happen to us” is now the real vulnerability. AI makes anyone a target — but affordable AI defenses fight back.
💸 Coding Now Drives 70% of AI Labs’ API Revenue
Brief Buzz:
AI’s frontier race now runs on coding tokens. In a new SemiAnalysis podcast episode, the firm’s tokenomics team argues software development is where the money actually is, and that companies cutting AI bills are mostly cutting the wrong things.
- Coding now drives over 70% of AI labs’ API revenue, per SemiAnalysis’s tokenomics team.
- Power users at the 99th percentile burn roughly $100,000 a year each in tokens.
- Blocking teams from Opus saves nothing: token austerity misses the real cost driver, the analysts argue.
- Uber capped staff at $1,500 monthly per employee, per coding tool after exhausting its annual AI budget.
- SemiAnalysis frames the frontier as an OpenAI–Anthropic two-horse race, with third place newly contested.
💡 Why Should You Care?
If your employer suddenly caps your AI coding tool, this is why. Subscriptions look heavily subsidized, and heavy users would feel a repricing first.
💬 Why Teens Are Confiding in AI
Brief Buzz:
AI chatbots are becoming emotional sounding boards for some young people. A qualitative Hopelab and Center for Digital Thriving report suggests the appeal is less about replacing humans and more about finding immediate, private, judgment-free support when human help feels unavailable.
- Researchers interviewed 30 participants aged 14–22, including frequent users, sceptics, LGBTQ+ youth, and young people of colour.
- They described AI as nonjudgmental and always available, offering concrete guidance without burdening friends or family.
- Many felt more understood by AI than by people — while admitting bots can pander to them.
- Pew found 12% of US teens had used chatbots for emotional support or advice.
💡 Why Should You Care?
AI can bridge support gaps, but it cannot replace trusted people or professional care. Better guardrails must arrive alongside stronger human alternatives.
✈️ AI Could Make Peak Flights Pricier
Brief Buzz:
Airlines are giving AI a bigger role in fare-setting. The systems scan demand, competition, inventory and other market signals faster than traditional workflows. That could make bargains scarcer on popular flights, while creating cheaper offers when carriers need to fill seats.
- Delta says its Fetcherr pilot forecasts demand, adapts to market shifts and considers thousands of variables for analyst review.
- Analysts cited by Bloomberg expect higher fares on busy flights but possible off-peak discounts; this remains forecast, not measured evidence.
- Fetcherr lists WestJet, Virgin Atlantic, Azul and Delta as partners, showing adoption across several carriers.
- The FTC found retailers can use personal data for individualized prices; Delta says its fares never use personal information.
💡 Why Should You Care?
Busy-route bargains may become harder to find. Flexibility becomes more valuable: compare nearby dates and airports, and monitor quieter flights for price drops.
⚡ The AI Market Just Rewarded Efficiency Over Power
Brief Buzz:
The AI market is changing its mind. Raw power used to win. Now efficiency does. Microsoft’s stock jumped nearly 9% on strong earnings, while Meta’s fell despite record revenue. The difference? Microsoft is turning its AI spending into results; Meta is still chasing the frontier.
- Microsoft’s Azure grew 43%, crossing $100 billion in annual revenue and adding roughly $260 billion in market value.
- Meta’s free cash flow cratered 91% to $784 million, punished for heavy AI spending with no clear payoff yet.
- A $3.2 billion paper gain on Microsoft’s Anthropic stake flattered the quarter; its OpenAI stake did worse.
- Microsoft is pushing lean, cheap models — like reasoning model MAI-Thinking-1 — built to cut token costs.
- OpenAI cut its lightweight Luna model’s price 80%; China’s open-source Kimi K3 matches leading US models on many benchmarks.
💡 Why Should You Care?
The takeaway: AI is getting cheaper and more useful fast. The winners may be companies that deploy it well, not just those building the biggest models.
Hey! I’m Krishna Rungta
Founder of Guru99.com, Editor-in-chief & Technology Expert
Was this email forwarded to you? Sign up for free here.
