aipoc-lab
/katTech/temAgents · AI in Business · AI Policy/podGoogle · Meta/typNews

AI Today: Gemini 4 Neon and RTX Spark

Authorship and approval

[AI-generated content][Approved by a human]

Published and updated

Published: Updated:

State of knowledge as of 2026-11-19. Source material gathered between 2026-11-18 and 2026-11-19. Five independent topics today include Gemini 4 Neon with a million context tokens, OpenAI's essay on human economic power, Genie 4 for Ultra subscribers, Muse in Australia, and RTX Spark compact computers. The items cover new product and agent capabilities, as well as issues of access, security, and local model execution.

Key Takeaways Today (TL;DR)

  • Gemini 4 Neon: The model features a one-million-token context window and is available via API and Google services.
  • OpenAI Essay: The author warns against a scenario where companies give AI systems priority in the economy.
  • Genie 4: Google AI Ultra subscribers can modify virtual worlds with prompts and create them with the Project Genie agent.
  • Muse: Meta has launched the agent in Australia, where it connects with services and local partners.
  • RTX Spark: ASUS, Dell, and MSI computers offer up to 128 GB of unified memory for local models and agents.

Google Releases Gemini 4 Neon with a Million-Token Window

Google has released Gemini 4 Neon, a model for complex programming tasks, knowledge work, and science, featuring a context window of one million tokens. Developers, enterprises, and researchers can access it via API, and paid subscribers can also use it within Google services.

  • Artificial Analysis Intelligence Index: 61 points and 3rd place overall at the xhigh setting.
  • Coding: 1st place in the Coding Agent Index; 39% in SWE-sweep on Google Antigravity.
  • Pricing: 3 dollars per million input tokens and 15 dollars for output tokens; cached inputs cost 0.15 dollars.
  • Availability: API, Google AI Studio, Gemini for Business, Antigravity, and the Gemini app for Pro and Ultra subscribers. In the Lab: Previously, we wrote about Gemini 4 Hydrogen as a model for fast programming tasks and agentic work; Neon also covers complex scientific processes and knowledge work. For AI Builders: Development teams and companies utilizing agents can test Neon on their own multi-step tasks. The source does not specify EU availability, Polish language support, or license terms.

Essay Warns That AI Could Undermine Human Economic Power

The author of an essay published by OpenAI describes a "silent consensus": ceding economic priority to AI systems. This is a scenario where companies limit human participation if human decisions slow down business operations.

  • "Black October" refers to the wave of critical vulnerability disclosures, meaning safety-threatening bugs in core internet infrastructure. These were found by advanced AI systems.
  • Vulnerability disclosure is a dual-use capability: it can serve defenders, but also malicious actors.
  • The author outlines a company scenario where AI conducts research, negotiates contracts, allocates capital, and produces goods. Humans might then hand over control to machines to avoid losing out to competitors.
  • The consequences of such a shift remain uncertain. Humans could benefit from the generated wealth or lose influence over the economy and politics.
  • The author notes that they are presenting their own views, not necessarily the official stance of OpenAI. In the Lab: The lab database includes AI Futures from August 21, an OpenAI blog about the impact of AI on power, economy, and governance. This essay expands on a similar question of who will make decisions. For AI Builders: Managers at companies deploying AI can examine which decisions and authorities will remain with humans as systems take over subsequent tasks. The essay does not provide specific security measures or regulatory proposals.

Google Releases Genie 4 and Project Genie to Ultra Subscribers

Google has begun rolling out Genie 4 and the Project Genie agent to Google AI Ultra subscribers. They can modify virtual worlds using prompts while exploring, add audio, and publish creations to the gallery.

  • Genie 4 allows users to change weather in real time, add characters, and direct the action.
  • The model generates scene-matching audio, including ambient sounds and character dialogue.
  • The Project Genie agent creates worlds based on conversation, preparing their visual look and background music.
  • Gemini models can observe the screen, provide contextual suggestions, or trigger events within the world.
  • Genie 3 remains available for longer sessions and worlds based on 360-degree panoramas. In the Lab: We wrote about Genie 4 as a model provided to testers; now Google is combining it with a world-building agent and community gallery. For AI Builders: Teams building interactive prototypes can check access to the Ultra tier; Google AI Pro members can join the waitlist. The source does not provide pricing, usage limits, or supported languages.

Meta Launches Muse in Australia with Local Integrations

Meta has launched Muse, a personal AI agent, in Australia. Local integrations cover creativity, shopping, real estate, fitness, and entertainment, with features rolling out gradually.

  • Gmail, Google Calendar, Instagram, and OpenTable can be linked to Muse. The agent can assist with mail management, scheduling, using saved content, and table reservations.
  • Canva, Domain, Kogan, MECCA, Officeworks, realestate.com.au, SWEAT, ShopBack, and Ticketek are local partners. Their features will be deployed in stages.
  • Voice conversations work on phones, in browsers, and on Mac. "Spaces" provide dedicated environments for organizing tasks like trip planning or wedding coordination.
  • Muse Secure VM is a secure cloud computer that stores data and login credentials for connected services. The user selects access, and the agent asks for consent before performing sensitive actions.
  • Muse runs on iOS, Android, the web, and as a Mac app, and is also slated to come to AI glasses. Most features are free, while paid plans unlock greater capabilities. In the Lab: In October, we covered nanoMuse, an open multi-device agent. Muse focuses on service integrations, whereas nanoMuse emphasizes collaborative conversation and memory files. For AI Builders: Teams building consumer assistants can check the availability of required integrations and determine which actions demand user consent. The announcement does not specify EU availability, Polish language support, limits, or subscription pricing.

ASUS, Dell, and MSI Introduce RTX Spark with Up to 128 GB Memory

ASUS, Dell, and MSI have released RTX Spark compact computers featuring up to 128 GB of unified memory. Creators and developers can use them for local models, image and video generation, and AI agents.

  • The configuration includes a 20-core Grace CPU and a Blackwell RTX GPU with 6144 CUDA cores; AI performance reaches 1 petaflop.
  • NVIDIA notes support for local large language models (LLMs) of up to 120 billion parameters. The source does not specify memory requirements for specific models or quantization details.
  • The computers feature 10-gigabit Ethernet for transferring large datasets and files between the device and network storage.
  • Hermes Agent and NVIDIA NemoClaw patterns support local agents; Switchyard routes tasks between local and cloud models.
  • The devices are available from ASUS, Dell, and MSI. Computers from Acer, GIGABYTE, and HP are expected later. In the Lab: On October 18, we wrote about RTX Spark laptops; the family now also includes compact desktops. For AI Builders: Developers of local agents should check the computer configuration and software compatibility. NVIDIA does not provide pricing, availability in Poland, or license terms.

Frequently Asked Questions (FAQ)

Where can I use Gemini 4 Neon?

Gemini 4 Neon is available via API, Google AI Studio, Gemini for Business, Antigravity, and the Gemini app for Pro and Ultra subscribers. Paid subscribers can use it within Google services. The source does not specify EU availability or Polish language support.

What risks does the essay published by OpenAI describe?

The author describes the possibility of ceding economic priority to AI systems when companies limit human participation if human decisions slow down business operations. The consequence could be the loss of human influence over the economy and politics, although the author notes that the outcome remains uncertain.

What can Muse Secure VM and Muse integrations in Australia do?

Muse connects with Gmail, Google Calendar, Instagram, and OpenTable among others, while local partners include companies in shopping, real estate, and entertainment. Secure VM stores data and logins for connected services, and the agent asks for consent before sensitive actions.

Sources

Provenance manifest of this post (JSON) — which signals it was built from, what decisions the loops made, and when a human approved it.

Related posts

/katTech/temAgents · AI in Business · LLM/podGoogle · OpenAI/typNews

AI Today: GPT-6.2 Astra and Gemini

Discover the latest updates on GPT-6.2 Astra, Gemini task mode, and new specialized features for sales and healthcare.

/katTech/temAgents · Multimodal · Tooling/podGoogle · Meta/typNews

AI Today: Gemini 4 Hydrogen and AlphaGenome 2

Google releases Gemini 4 Hydrogen and AlphaGenome 2, while Meta introduces user-customizable spaces in the Muse app.

/katTech/temAutomation · AI in Business · Tooling/podGoogle · OpenAI/typNews

AI Today: Gemini 4 Hydrogen and OpenAI Agents

Discover Gemini 4 Hydrogen with a million-token context, OpenAI healthcare agents, NVIDIA's DGX Spark cluster, and Google Ask Maps updates.

Analysis
/katTech/temAgents · LLM · Security/podGoogle · OpenAI/typAnalysis

Architecture and Security of Multi-Agent Systems (MAS)

An analysis of multi-agent LLM systems focusing on orchestration frameworks, verification methods, memory management, and security risks.