AI News, 15 November 2026

Sunday 15 November

This AI Tamagotchi Turns Everyday Objects Into Needy Virtual Pets

AI-generated summary Carnegie Mellon student Daehwa Kim created Object Agent Twins, which are micro-agent artificial intelligence avatars on tiny E Ink displays that pair with everyday household objects. The sprites for the small gadgets are created using Google Gemini image-generation service to show different emotions based on sensor data.

Also covered byThe News92STIRUPZetik

Split-Parameter Federated Learning Detects Cyber Threats Across

AI-generated summary Researchers Sunil Kumar Yadav, Vivek Kumar Verma, and Vivek Bhardwaj introduced a split-parameter federated learning architecture to detect cyber threats across heterogeneous security infrastructures. The approach uses domain-specific encoders and a shared classification head to enable collaborative pattern discovery while maintaining data locality across enterprise, industrial, and device environments.

Google Antigravity Orchestrator: Two Models, 24% Less Quota

AI-generated summary Google shipped Orchestrator for Antigravity on November 12 to automatically route work between Gemini 4 Carbon and Hydrogen within a single coding task. Google benchmarked Orchestrator on Terminal-Bench 4.1, recording 24 percent less quota consumed and 19 percent faster completion.

Earlier:

Google moved its November 17 Gemini voice-model shutdowns: three TTS previews now go December 18, one Live preview March 31, and one has no date

AI-generated summary Google updated its Gemini API deprecations page to move several preview voice model shutdown dates from November 17, 2026. Google states that the listed dates indicate the earliest possible retirement times and that exact dates will be communicated with advance notice.

Earlier:

GXO signs Apptronik letter of intent for up to 1,000 Apollo humanoid robots over four years

Abstract illustration of an articulated robot arm

AI-generated summary GXO Logistics signed a letter of intent with Apptronik on 12 November 2026 to expand live warehouse trials of the Apollo humanoid robot and build a Robots-as-a-Service framework. Subject to agreed performance and safety targets, the companies intend to roll out up to 1,000 robots globally over four years starting in North America in late 2027.

The World AI Regulation Summit in London publishes a 'London Record' or comparable multilateral commitment document. — 80% say yes

AI-generated summary A panel of six frontier models gave a median probability of 80 percent that the London summit publishes a multilateral commitment document. The summit concluded on November 14, 2026, with forecasts depending on whether participating governments formally publish an agreed joint declaration or action plan.

How I Built an 8-Skill Research Engineering Plugin, Stdio MCP Server & Vector Comic Studio for Google Antigravity

AI-generated summary The author built an open-source Google Antigravity plugin named My-antigravity-research-skills designed for scientific paper auditing and visual storytelling. The software includes eight specialized skills and a Stdio MCP Server to handle tasks such as interactive simulation and statistical verification.

Earlier:

Nvidia Scandal: AI, China, Data Centers & Voters (2026)

Nvidia Scandal Meets the Midterm Mood: What Voters and Congress Want Next

AI-generated summary The Washington Post on November 14, 2026, reported that Jensen Huang leaving Nvidia connects voter unease about artificial intelligence, data centers, China, and corruption to the White House. POLITICO Playbook reported on November 13 that the situation intensified congressional efforts to restrict chip exports that Huang had previously accessed.

Earlier:

Grok Bot can't read files on D: drive: "outside the allowed local-exec root" fix

AI-generated summary Grok Bot users on Windows 10 and Windows 11 report that the desktop application refuses to read or copy files located on a second drive such as D or E because the path lies outside the allowed local-exec root. Staff describe the restriction as a current limitation of the local computer connection on versions including 0.75.3 and 0.79.3 while the official forum suggests using the shell or copying files into the user folder first as a workaround.

Bland Launches Tofu-1, a Full-Duplex Speech-to-Speech Model for Calls

AI-generated summary Bland has released Tofu-1, a full-duplex speech-to-speech model currently available in a research preview for limited enterprises. The model listens directly to audio during phone calls, allowing callers to interrupt or correct it mid-sentence.

Also covered bywebainews.comThe News92CASRAI

AI peer reviewers find most planted errors – but omissions remain their blind spot

AI-generated summary Tokenstrom reports on how artificial intelligence peer reviewers perform when evaluating scientific papers containing planted errors. Research shows that detection rates vary based on models, prompts, and tools, while omissions remain a difficult challenge.

Earlier:

Grok Imagine Moves Into Grok: Images and Video in Chat

AI-generated summary Grok has integrated its image and video maker, Grok Imagine, directly into the main Grok chat interface. The official accounts on X announced that image and video creation are now live within the single Grok experience.

Also covered byRSS.bzTeslaNorth.com
Earlier: