AI News — Sunday, May 17, 2026

9
Warp-as-History: Generalizable Camera-Controlled Video Generation from One Training Video

Researchers introduce Warp-as-History, a novel method enabling generalizable camera-controlled video generation from a single training video, significantly advancing video synthesis capabilities.

Hugging Faceresearch
9
OpenAI Co-founder Greg Brockman Takes Charge of Product Strategy

OpenAI co-founder Greg Brockman is reportedly taking over product strategy, signaling a significant shift in the company's direction and focus.

TechCrunchindustry
8
ArXiv to Ban Authors for One Year for AI-Generated Submissions

The research repository ArXiv announces a new policy to ban authors for a year if they are found to have used AI to generate their entire research papers, aiming to maintain academic integrity.

TechCrunchindustry
8
OpenAI and Malta Partner to Bring ChatGPT Plus to All Citizens

OpenAI has partnered with the government of Malta to provide free access to ChatGPT Plus for all Maltese citizens, marking a significant national-level AI adoption initiative.

OpenAI Blogproduct
7
RouteProfile: Elucidating the Design Space of LLM Profiles for Routing

This research explores RouteProfile, a framework that clarifies the design space for LLM profiles used in routing, optimizing how language models handle diverse tasks.

Hugging Faceresearch
7
PREPING: Building Agent Memory without Tasks

A new paper introduces PREPING, a method for building robust agent memory architectures without requiring explicit task-specific training, enhancing agent adaptability.

Hugging Faceresearch
7
EvolveMem: Self-Evolving Memory Architecture via AutoResearch for LLM Agents

EvolveMem proposes a self-evolving memory architecture for LLM agents that uses 'AutoResearch' to continuously improve its memory capabilities, leading to more adaptive AI agents.

Hugging Faceresearch
7
Realiz3D: 3D Generation Made Photorealistic via Domain-Aware Learning

Realiz3D presents a new approach to 3D generation that achieves photorealistic results by incorporating domain-aware learning, significantly improving visual fidelity.

Hugging Faceresearch
6
I Ran Hermes Agent on the Same Task for 7 Days. The Skill File on Day 7 Looked Nothing Like Day 1.

A developer shares insights from running an AI agent for a week, observing how its 'skill file' dynamically evolved and changed significantly over time, highlighting adaptive learning in agents.

Dev.toresearch
6
Long Context Pre-Training with Lighthouse Attention

This paper introduces Lighthouse Attention, a novel mechanism designed to enable more efficient and effective long-context pre-training for large language models.

Hugging Faceresearch
6
The Haves and Have Nots of the AI Gold Rush

TechCrunch explores the growing disparity between major tech players and smaller entities in the ongoing AI gold rush, examining who benefits most from the current boom.

TechCrunchindustry
6
Google Partners with XPRIZE and Range Media Partners for $3.5M Future Vision Film Competition

Google announces a collaboration with XPRIZE and Range Media Partners to launch the $3.5 million Future Vision film competition, encouraging creative exploration of AI's impact on society.

Google AI Blogindustry
5
My AI Agent Kept Blind-Reading My Files, So I Built a Local CLI to Give It 'Eyes'

A developer created a local command-line interface to provide AI agents with controlled access and 'vision' into local files, addressing privacy and efficiency issues with current agent designs.

Dev.toopen-source
5
Orchard: An Open-Source Agentic Modeling Framework

Orchard is introduced as a new open-source framework designed for agentic modeling, providing tools and structures for developing and deploying AI agents.

Hugging Faceopen-source
5
I Built an Agent That Actually Reviews Your Pull Requests

A developer showcases an AI agent designed to autonomously review pull requests, aiming to streamline code review processes and improve development workflows.

Dev.toopen-source