AI News Brief: Morning roundup, 2026-11-14
Summary
The 06:00–12:00 CET window was dominated by agents, memory and the business side of AI: new tooling for building and controlling AI systems, a flagship model swap at Google, and a growing backlash over AI's effect on writing, schools and young workers. Several threads run through the list. Companies are packaging AI as ready-made agent products, from Nectar's creator-outreach agent to Google's upgraded Search, while infrastructure write-ups show how teams can meter and justify those deployments. The other thread is the cost of the technology: publishing disputes, school pilots drawing criticism and a leadership change at Nvidia that raises open-source questions. Readers building with or deciding on AI will find the practical tooling useful and the backlash stories a sign of where scrutiny is heading.
- Stanford researchers open-sourced Memorilla, which compresses a document collection into 16 memory tokens for a frozen Qwen3-8B. It scores 84.33% on FactKG while using 108 times fewer document tokens than full-context prompting.
- Nectar launched an AI agent that automates influencer marketing, covering creator discovery, outreach, partnership management and campaign result tracking for brands.
- Google faces calls for AI moratoriums in schools after a teen's lawsuit labelled its Gemini a "suicide coach," with families confronting the company over a pilot programme.
- The Guardian's new series follows young job seekers, who describe endless cycles of applications and interviews and companies that never send rejections, arguing the social contract is broken.
- Runway released Praxis-1 under Apache 2.0 with open weights, using knowledge from video training to control robots and leading the World Action Model category on Nvidia's RoboLab-120 benchmark, though it cannot yet plan complex action sequences alone.
- The Hugging Face blog introduced LLMON, an LLM-native markup language now available in IBM's Granite 4.1 3B and 8B models. The source text offers no further detail.
- Google rolled out Gemini 4 Hydrogen to replace 3.8 Flash, skipping the slower, security-conscious release cycle used for its Carbon and Argon models.
- Google's AI Mode in Search now handles complex queries better, surfacing relevant links and richer visual answers, powered by Gemini 4 Hydrogen.
- Authors are losing deals and being removed from prize shortlists over suspected AI use, with romance novelist Coral Hart describing 370 novels written under 21 pen names using AI and earning "high six figures."
- Vercel Labs released Vercel UI SDK, an Apache-2.0 beta of 50+ UI behaviours that coding agents compose into accessible components across React, Vue, Svelte, Solid and Web Components.
- The "No AI Slop Manifesto" from AIE speakers, published in Latent Space's AINews, sets out how speakers plan to push back against low-quality AI-generated content.
- Jensen Huang's resignation as Nvidia's leader raises questions about the company's commitment to open-source AI, including its Hugging Face acquisition, open-weight models and advocacy against restrictions.
- A new AWS guide shows how to run a self-service LLM gateway on Fargate using LiteLLM, giving developers one API key for Amazon Bedrock models while platform teams gain per-user cost controls and quota isolation.
- An AWS post explains how to calculate true cost per task on Amazon Bedrock across four hidden cost drivers, then cut it with caching, routing and output limits, and prove agent ROI with human baselines and holdout tests.
- Kiro, the AI coding tool, is now available on Android through Google Play, extending its agent-assisted development to mobile.
Summaries written with AI from the linked source articles.
Sources in this roundup
| AI News & Artificial Intelligence | TechCrunch | 3 article(s) | |
| AI (artificial intelligence) | The Guardian | 2 article(s) | |
| AI – Ars Technica | 2 article(s) | |
| Artificial Intelligence | 2 article(s) | |
| MarkTechPost | 2 article(s) | |
| Hugging Face – Blog | 1 article(s) | |
| Kiro Blog | 1 article(s) | |
| Latent Space | 1 article(s) | |
| The Decoder | 1 article(s) |
Most-mentioned keywords
| 3 mention(s) | ||
| open | 3 mention(s) | |
| agent | 2 mention(s) | |
| amazon | 2 mention(s) | |
| bedrock | 2 mention(s) | |
| gemini | 2 mention(s) | |
| llm | 2 mention(s) | |
| source | 2 mention(s) |
Sources
- Stanford Research Open-Sources Memorilla, a Latent Memory Module for Frozen LLMs
- Nectar launches an AI agent to automate creator discovery and outreach for brands
- Google faces calls for AI moratoriums in schools after Gemini pilot backlash
- ‘The social contract is broken’: young people on feeling shut out of careers
- Runway's open-source Praxis-1 turns video AI into a robot controller
- LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B
- Google rolls out Gemini 4 Hydrogen, finally replacing 3.8 Flash
- Google’s latest Search upgrade promises better answers to your most complicated questions
- ‘It’s like the wild west’: how AI has rocked the literary world
- Vercel Releases UI SDK: 50+ UI Behaviors for AI Coding Agents
- [AINews] The No AI Slop Manifesto
- Jensen Huang’s exit puts Nvidia’s support for open source in question
- Building a self-service LLM gateway with Amazon Bedrock and LiteLLM
- Optimizing cost per task and proving agent ROI on Amazon Bedrock
- Introducing Kiro for Android
