AI News — Friday, August 14, 2026

10
OpenAI Introduces 'Ultrafast' Mode for GPT-5.6 Sol, Boosting Speed by 14x

OpenAI has launched 'Ultrafast' mode for its GPT-5.6 Sol model, enabling it to operate at up to 14 times its previous speed, significantly enhancing performance for users and developers.

TechCrunchproduct
9
Databricks Secures $5 Billion in Funding at a Staggering $190 Billion Valuation

Databricks has closed a $5 billion funding round, valuing the AI and data analytics company at an impressive $190 billion, reflecting strong investor confidence in its market position.

TechCrunchindustry
9
Google Redesigns Search Box for the First Time in 25 Years, Signaling AI Integration

Google has unveiled the first redesign of its iconic search box in 25 years, a move that is expected to deeply integrate AI capabilities and fundamentally change user interaction with search.

VentureBeatproduct
9
OpenART: Scaling AI Agent Red Teaming Through Open-Ended Environment Evolution

Researchers introduce OpenART, a novel approach that scales AI agent red teaming by evolving open-ended environments to rigorously test and improve agent safety and robustness.

Hugging Faceresearch
9
Spark-to-Paper: End-to-End Research Paper Generation as a Composable AI Skill

A new research paper, 'Spark-to-Paper,' demonstrates an AI system capable of end-to-end generation of research papers by composing various AI skills, hinting at future automation in scientific writing.

Hugging Faceresearch
8
Writer Unveils New AI Model and Upgraded Harness to Optimize Token Costs

Writer has launched a new AI model alongside an enhanced 'harness' system designed to significantly reduce and contain token costs, offering more efficient AI solutions for enterprises.

TechCrunchproduct
8
Google Enhances Gemini API Managed Agents with 3.6 Flash and New Hooks

Google has updated its Gemini API Managed Agents, introducing the 3.6 Flash model and new hooks, providing developers with more powerful and flexible tools for building AI applications.

Google AI Blogproduct
8
AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

New research explores 'AI4AI at Test-Time,' a method for transferring capabilities from strong AI models to weaker ones using harnesses, improving efficiency and adaptability.

Hugging Faceresearch
8
Mechanist: AI as a Scientific Instrument for Discovering Mechanisms of Intelligence

A new paper introduces 'Mechanist,' proposing AI as a scientific instrument to uncover the underlying mechanisms of intelligence, pushing the boundaries of AI's role in fundamental research.

Hugging Faceresearch
7
OpenAI Appoints Dali Rajic as Chief Revenue Officer to Drive Monetization

OpenAI has announced the appointment of Dali Rajic as its new Chief Revenue Officer, signaling a strategic focus on accelerating monetization and scaling the company's commercial operations.

OpenAI Blogindustry
7
Google Introduces New AI Tools to Evolve Marketing Strategies

Google has released a suite of new AI-powered tools designed to help businesses evolve their marketing strategies, offering advanced analytics and automation for more effective campaigns.

Google AI Blogproduct
7
Google Sheets Canvas Brings Spreadsheet Data to Life with Enhanced Visuals

Google has launched Sheets canvas, a new feature for Google Sheets that aims to bring spreadsheet data to life with improved visualization and interactive capabilities, likely powered by AI.

Google AI Blogproduct
7
I Stopped Trusting AI Agents With Tools. So I Built a Gatekeeper.

A developer shares their experience of building a 'gatekeeper' system to manage and control AI agents' access to tools, addressing concerns about reliability and safety in autonomous AI operations.

Dev.toopen-source
7
SkillZip: Contract-Preserving Graph Compression for Scalable Agent Skill Libraries

Researchers present SkillZip, a method for compressing graph-based agent skill libraries while preserving critical 'contracts,' enabling more scalable and efficient AI agent systems.

Hugging Faceresearch
6
The Most Dangerous AI-Generated Code Is the Code That Passes All Tests

An article argues that the most insidious threat from AI-generated code comes from code that appears flawless by passing all tests, potentially masking deeper, subtle vulnerabilities or inefficiencies.

Dev.toindustry