OpenAI

2d

Researchers trained an open source AI search agent, Harness-1, that outperforms GPT-5.4 on recalling relevant information

A joint research collaboration between researchers at the University of Illinois at Urbana-Champaign (UIUC), UC Berkeley, and the open source AI-native vector database platform Chroma unveiled Harness-1, a 20-billion parameter open-source search agent built atop OpenAI's gpt-oss-20B open source model that fundamentally redesigns how AI executes complex retrieval tasks. Harness-1 achieves a massive leap in performance, scoring 73% average on its ability to recall relevant information correctly fr...

Apple

2d

Apple announces visionOS 27 with Siri AI

Apple's latest visionOS 27 update introduces powerful new AI capabilities for the Vision Pro, particularly an upgraded Siri. This 'visual intelligence' feature lets users instantly query both physical and digital spaces, potentially streamlining operations for SMBs. The update also boosts Wi-Fi speeds and customization options, making the Vision Pro a more robust tool for business applications.

Anthropic

19h

Latest coverage: Anthropic Releases Claude Fable 5 and Claude Mythos 5: Same Underlying Model, Different Safeguards, New Mythos-Class Tier

Anthropic has just launched Claude Fable 5, their most advanced AI model to date. This 'Mythos-class' AI is particularly adept at software development and image recognition, showing strong performance in long, complex tasks. SMBs facing talent shortages in these areas might find this a powerful new tool.

OpenAI

2d

OpenAI, a major player in the AI space, has confidentially submitted its S-1 filing to the U.S. Securities and Exchange Commission. This marks the initial phase of its journey towards an Initial Public Offering (IPO). This move follows a similar confidential filing by its competitor, Anthropic, earlier in June, intensifying the race between these two AI powerhouses to go public.

Google

1d

Google has unveiled Gemini 3.5 Live Translate, a significant leap in real-time translation technology. This new AI model, part of the Gemini 3.5 family, can instantly translate spoken conversations between more than 70 languages, even replicating the speaker's tone and pace. For SMBs dealing with international clients or customers, this tool promises to dramatically improve communication and unlock new business potential.

PET imaging, crucial for medical diagnosis, often struggles with image noise, making AI processing difficult. A new deep learning model, U-TTT, uses a 'test-time training' approach to adapt to various conditions, consistently removing noise and improving diagnostic accuracy. This advancement could lead to more precise medical diagnoses.

Google

6/3/2026

While many AI open source model providers are pursuing larger and more powerful models, Google is still giving attention to the smaller, more local side of the market. Today, the tech giant released Gemma 4 12B, an 11.95-billion-parameter open-weights model with permissive Apache 2.0 license optimized to execute locally on a standard enterprise laptop using just 16GB of VRAM or unified memory. That means those enterprise users looking to keep working with AI while on a flight without WiFi, or tr...

Microsoft's research team introduced AsyncWebRL, a novel reinforcement learning framework designed to accelerate the training of visual language web agents. This innovation aims to resolve common inefficiencies in AI training, such as idle GPUs and unproductive learning paths. By improving speed and performance, AsyncWebRL could contribute to the development of smarter, more efficient AI systems for web interaction.

Large language model (LLM) agents are increasingly tackling diverse tasks, but current rigid systems struggle with varied execution patterns. HarnessForge introduces a novel framework that co-evolves an agent's external execution structure ('harness') and internal reasoning logic ('policy'), demonstrating significant performance gains in complex scenarios. This research highlights the critical importance of compatibility between these two components for agent adaptability.

Google

2d

Google's AI note-taking app, NotebookLM, has been significantly updated with the new Gemini 3.5 model. This upgrade promises more accurate and reliable information, transforming how SMBs conduct research. Users can now start research projects simply by asking NotebookLM questions, leveraging Google Search to find relevant sources efficiently.

NVIDIA

5d

Authors: Yi Yang, Zhihong Liu, Siqi Kou, Yiyang Chen, Yanzhe Hu arXiv Links arXiv | PDF AI summary Abstract World-language-action models combine textual instruction processing with robot state prediction through an autoregressive transformer backbone, enabling efficient long-horizon task execution and cross-embodiment learning. Generated by Qwen/Qwen2.5-Coder-32B-Instruct 摘要 世界语言动作模型通过自回归变压器主干将文本指令处理与机器人状态预测结合起来，从而实现高效的长视野任务执行和跨实体学习。 由 Qwen/Qwen2.5-Coder-32B-Instruct 生成 Abstract Generated by Qwe...

A new AI training method, OPRD (On-Policy Representation Distillation), dramatically improves efficiency by addressing key challenges in traditional approaches. It speeds up learning by focusing on internal AI processes, not just final outputs. This innovation could lead to more powerful and accessible AI services in the future.