AI News — Friday, April 17, 2026

9
Seedance 2.0: Advancing Video Generation for World Complexity

Researchers introduce Seedance 2.0, a significant advancement in video generation capabilities, enabling the creation of more complex and realistic virtual worlds.

Hugging Faceresearch
9
Factory hits $1.5B valuation to build AI coding for enterprises

Factory, an AI coding platform, has achieved a $1.5 billion valuation, signaling strong investor confidence in AI-powered development tools for large organizations.

TechCrunchindustry
9
A new way to explore the web with AI Mode in Chrome

Google introduces an 'AI Mode' in Chrome, offering users novel ways to interact with and explore web content through artificial intelligence.

Google AI Blogproduct
8
Luma launches AI-powered production studio with faith-focused Wonder Project

Luma has unveiled a new AI-powered production studio, launching with the 'Wonder Project' specifically tailored for faith-based content creation.

TechCrunchproduct
8
Upscale AI in talks to raise at $2B valuation, says report

Upscale AI is reportedly in discussions to secure new funding that would value the company at $2 billion, highlighting rapid growth in the AI sector.

TechCrunchindustry
8
Codex for (almost) everything

OpenAI announces an expansion of its Codex model's capabilities, aiming to apply its code generation and understanding across a broader range of tasks and domains.

OpenAI Blogproduct
8
Introducing GPT-Rosalind for life sciences research

OpenAI unveils GPT-Rosalind, a specialized AI model designed to accelerate and enhance research within the life sciences sector.

OpenAI Blogproduct
7
GameWorld: Towards Standardized and Verifiable Evaluation of Multimodal Game Agents

Researchers propose GameWorld, a new framework for standardized and verifiable evaluation of multimodal AI agents operating within game environments.

Hugging Faceresearch
7
RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time

A new research paper introduces RationalRewards, a method that uses reasoning-based rewards to improve the scalability of visual generation models during both training and testing.

Hugging Faceresearch
7
Claude Code costs up to $200 a month. Goose does the same thing for free.

A VentureBeat article highlights a free alternative called Goose that offers similar AI coding capabilities to the paid Claude Code, potentially disrupting the market for developer tools.

VentureBeatindustry
7
Exploration and Exploitation Errors Are Measurable for Language Model Agents

New research demonstrates that exploration and exploitation errors, critical aspects of decision-making, can be quantitatively measured in language model agents.

arXivresearch
7
SciFi: A Safe, Lightweight, User-Friendly, and Fully Autonomous Agentic AI Workflow for Scientific Applications

SciFi introduces a new autonomous, agentic AI workflow designed to be safe, lightweight, and user-friendly for various scientific applications.

arXivresearch
7
Numerical Instability and Chaos: Quantifying the Unpredictability of Large Language Models

This paper explores the inherent numerical instability and chaotic behavior in large language models, providing a framework to quantify their unpredictability.

arXivresearch
6
The creator of Claude Code just revealed his workflow, and developers are losing their minds

The workflow of Claude Code's creator has been revealed, generating significant excitement and discussion among developers keen to optimize their AI-assisted coding practices.

VentureBeatindustry
6
Join the OpenClaw Challenge: $1,200 Prize Pool!

Developers are invited to participate in the OpenClaw Challenge, an initiative with a $1,200 prize pool aimed at fostering innovation in AI agent development.

Dev.toopen-source