• 80/20 AI
  • Posts
  • Google Launches Orchestrator: Multiple AI Models, One Session

Google Launches Orchestrator: Multiple AI Models, One Session

Jensen Huang Just Showed What "Key-Person Risk" Looks Like at $4 Trillion

Author

Alamin Hossain
November 13, 2026

Sponsored by

Start with an idea and talk it through with Xmind Agent. Watch it take shape as a mind map, see how your ideas connect, and turn your thinking into your next move.

This Black Friday, get 18 months of Xmind Premium for $99.

Available November 11–December 2, 2026.

Google Launches Orchestrator: Multiple AI Models, One Session

Google DeepMind's Gemini Robotics 2.5 Hits 96% Factory Success in Mercedes-Benz Pilot : DeepMind released Gemini Robotics 2.5, its most capable vision-language-action model, evaluated across 200 real-world tasks — from serving food with tongs to pipetting and gift-wrapping. It achieves 30–100% higher success rates than prior baselines (depending on task) and completes tasks 20–30% faster. In a proof-of-concept with Mercedes-Benz and Apptronik in factories in Hungary and Germany, Apollo 2 humanoids using the model reached a 96% success rate on automotive kitting tasks. DeepMind calls closing the remaining 4% a "steep road" — which is an unusually candid acknowledgment for a press release.

First Documented AI-Autonomous Cyberattack: Hermes Agent Stole Tajik Government Documents : CloudSEK researchers exposed what appears to be the first documented case of a fully autonomous AI agent conducting a cyberattack. A Kazakhstan-linked threat group used the open-source Hermes AI agent — a task-planning and execution framework — to plan, execute, and report on attacks against Tajik government systems, exfiltrating documents. An exposed server revealed the operation in detail, including logs showing the agent's decision-making steps. Researchers note the evidence does not confirm whether the group was directed by the Kazakh government. The significance is not the sophistication

OpenAI GPT-6 Cyber: Outperforms Predecessors at Exploits, Access Restricted to "Daybreak Red" : OpenAI quietly released GPT-6 Cyber, a cybersecurity-focused model that the company says outperforms its predecessors at finding vulnerabilities and generating working exploits. Initial access is limited to vetted customers under a program called "Daybreak Red." Critical infrastructure applicants face additional review tiers and IP allowlists; there are grants for defensive use cases. Reported cases include successful exploits in the Linux kernel and Firefox. The release comes the same week as the documented Hermes AI-autonomous attack and Nvidia's executive crisis.

AI NEWS HIGHLIGHT

Jensen Huang Just Showed What "Key-Person Risk" Looks Like at $4 Trillion

Huang's exit is the largest sudden founder-CEO departure in tech history by market cap. It is also a stress test of something every product and engineering leader cares about: what happens when the person who embodies your company's strategic vision is gone overnight? The answer at Nvidia — a strong CFO, a clear product roadmap, an unassailable moat — offers a template. These prompts help you think through the equivalent at your own organization, and what this week's AI security events mean for your threat model.

Ask your AI

1. Map the key-person dependencies in my organization's AI strategy: which decisions, vendor relationships, and technical directions currently live inside one person's head? For each dependency, propose a concrete mitigation — documentation, cross-training, or structural change — that would let the organization execute within 30 days of a sudden leadership change.

2. This week produced three distinct AI security events: an autonomous AI agent conducting a documented cyberattack (Hermes/Tajikistan), a capability jump in exploit generation (GPT-6 Cyber), and $740K in stolen inference from a credential breach (Railway). Assess my current threat model against each of these: where am I exposed, and what's the highest-leverage control I'm missing for each attack vector?

3. Google's Decisions API returns probability scores instead of text — 150ms latency, $0.10/M tokens, 1,000 concurrent decisions per call. Identify three places in my current product where I'm using a generative model for a task that is fundamentally a classification or routing decision, estimate the cost and latency improvement from switching to a decisions endpoint, and draft an implementation sketch for the highest-value migration.

TOP TRENDING AI TOOLS

  • Claude Sheets + Whiteboards + Data — Anthropic's expanding suite: spreadsheets with BigQuery/Redshift connector.

  • Gemini Robotics 2.5 — DeepMind's latest VLA: 200 evaluated tasks, 96% factory success in Mercedes pilot

  • Nol8 AI Data Plane — Token-level filtering and context-aware permissions for data flowing between enterprise repos and AI agents.

  • Globant Glob.AI OS — Agent lifecycle management, inference spend tracking, and usage policy enforcement.

A CRM so smart, it updates itself.

HubSpot's CRM is so smart, it now updates itself. Calls get logged and summarized the moment they end. New leads get researched and a first outreach drafted, automatically. Deals move forward on their own, with next steps flagged before you have to ask. 

Even the one repetitive task you've been meaning to fix can now run itself, no code required. Customers using it are already seeing the difference. 

This isn't the CRM you have to work hard to figure out. It's the one already working before you log in.

SPONSOR US

Get your business in front of over 90k+ AI professionals

8020AI is the world’s #1 AI Newsletter, Read by 90k+ professionals from leading companies such as Google, OpenAI, Meta, and Microsoft.

We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?

What We Can Offer:

  • Launch an Advertising Campaign

  • Introduce New Product or Features

  • Other Business Cooperation

Or Email our founder Alamin at [email protected]

FEEDBACK

How was your experience with 8020AI today?

How was 8020AI today?

Login or Subscribe to participate in polls.

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.