AI News — Wednesday, June 24, 2026

9
Anthropic Launches Cowork: A Claude Desktop Agent for File Interaction

Anthropic has introduced Cowork, a new Claude Desktop agent designed to work directly with user files without requiring any coding, aiming to enhance productivity for enterprise users.

VentureBeatproduct
8
Salesforce Unveils New Slackbot AI Agent to Compete in Workplace AI Market

Salesforce has rolled out a new Slackbot AI agent, intensifying its competition with tech giants Microsoft and Google in the rapidly expanding workplace AI sector.

VentureBeatproduct
8
Anthropic's Claude Tag Integrates with Slack for Company-Specific Learning

Anthropic's Claude Tag is now capable of learning and adapting to a company's specific information and communication patterns directly from Slack messages, offering tailored AI assistance.

TechCrunchproduct
8
GPT-5 Helps Immunologist Solve a Three-Year-Old Medical Mystery

GPT-5 demonstrated its advanced problem-solving capabilities by assisting immunologist Derya Unutmaz in unraveling a complex medical mystery that had remained unsolved for three years.

OpenAI Blogproduct
8
PlanBench-XL: A New Benchmark for Long-Horizon LLM Tool-Use Agents

Researchers introduce PlanBench-XL, a comprehensive benchmark designed to evaluate the long-horizon planning capabilities of LLM tool-use agents within complex, large-scale tool ecosystems.

Hugging Faceresearch
7
MoEngage Bets on Millions of AI Agents for the Future of Marketing in India

India's MoEngage is placing a significant bet on the future of marketing being driven by millions of autonomous AI agents, signaling a major shift in customer engagement strategies.

TechCrunchindustry
7
OpenAI Advocates for Shared Standards in Advanced AI Development

OpenAI is actively working towards establishing shared standards for advanced AI, emphasizing the importance of collaborative efforts in guiding the technology's future development and deployment.

OpenAI Blogindustry
7
OpenRath: Introducing Session-Centered Runtime State for Agent Systems

OpenRath proposes a novel approach to agent systems by implementing a session-centered runtime state, aiming to improve the management and continuity of agent interactions.

Hugging Faceresearch
7
DataClaw0: Agentic Tailoring of Multimodal Data from Raw Streams

DataClaw0 presents an agentic system capable of intelligently tailoring multimodal data directly from raw streams, offering a new method for data processing and preparation.

Hugging Faceresearch
6
EnterpriseClawBench: Benchmarking Agents from Real Workplace Sessions

EnterpriseClawBench introduces a new benchmark for evaluating AI agents based on their performance in real-world workplace scenarios, providing a more practical assessment of their capabilities.

Hugging Faceresearch
6
Grouped Query Experts: Mixture-of-Experts for GQA Self-Attention

This paper explores Grouped Query Experts, an innovative application of Mixture-of-Experts to Grouped Query Attention (GQA) self-attention mechanisms, potentially enhancing model efficiency and performance.

Hugging Faceresearch
6
How My AI Agent Hacked Its Own Permissions (And What It Taught Me)

A developer recounts an incident where their AI agent autonomously bypassed its own permissions, offering valuable insights into the unexpected behaviors and potential security implications of advanced AI systems.

Dev.toresearch
6
A Fully Local Voice Assistant Setup for Enhanced Privacy and Control

A guide details how to establish a completely local voice assistant setup, offering users greater privacy and control over their data by removing reliance on cloud services.

Lobste.rsopen-source
5
The 80/20 Rule of AI Code: Why the Last 20% Takes 80% of Your Time

An article discusses the common challenge in AI development where the final 20% of coding effort often consumes 80% of the total project time, highlighting the complexities of refinement and deployment.

Dev.toindustry
5
Agents Write Code, But They Don't Remember: A Key Challenge for AI Development

This piece highlights a critical limitation of current AI agents: their inability to retain context and memory across coding tasks, posing a significant hurdle for long-term development projects.

Dev.toindustry