07 Aug 2025

AI News Digest

🤖 AI-curated 6 stories

Today's Summary

So, OpenAI’s about to drop ChatGPT 5.0, which is set to change the game with some pretty cool multimodal abilities—think text, images, and maybe even voice working together. Meanwhile, Wix is stepping up for small businesses by rolling out financial tools like Wix Checking and Wix Capital, making it easier for merchants to manage their cash flow without juggling multiple platforms. And for those keeping an eye on AI research, Hugging Face is making waves with their ‘Reflect, Retry, Reward’ framework, which could mean smarter AI that learns from its mistakes—no synthetic data required.

Stories

OpenAI's ChatGPT 5.0 Launches August 2025 with Multimodal AI

OpenAI is set to release ChatGPT 5.0 in August 2025, introducing advanced multimodal capabilities that integrate text, images, and potentially voice inputs. This upgrade aims to enhance AI interactions, offering more intuitive and human-like experiences. The new model emphasizes ethical safeguards, natural prompting, and user well-being features, potentially revolutionizing various industries by providing a safer and more productive AI tool.

Wix Launches Financial Services for Merchants

Wix has introduced Wix Checking and Wix Capital to assist merchants in managing cash flow and funding growth. Wix Checking offers an integrated business checking account that syncs automatically with Wix Payments, eliminating the need for external banking tools. Wix Capital allows clients to request cash advances in exchange for a fixed fee and a percentage of future sales, providing flexible financial solutions for businesses.

InternVL3: Advancing Multimodal Pre-training Paradigms

InternVL3 introduces a native multimodal pre-training approach, jointly acquiring multimodal and linguistic capabilities from the outset. This model achieves state-of-the-art performance on benchmarks like MMMU, rivaling proprietary models such as ChatGPT-4o. The research signifies a shift towards integrated training methodologies for enhanced multimodal understanding.

Reflect, Retry, Reward: Enhancing LLMs via Reinforcement Learning

The 'Reflect, Retry, Reward' framework proposes a method for large language models to self-improve by generating self-reflections upon failure, retrying tasks, and receiving rewards for successful attempts. This approach enhances performance on complex tasks using binary feedback, eliminating the need for synthetic data, and marks a significant advancement in reinforcement learning applications for AI.

Tavily Secures $25 Million Series A to Enhance AI Agents' Web Connectivity

Tavily, a startup providing compliant internet access for AI agents, has raised $25 million in a Series A funding round led by Insight Partners. Founded in 2024, Tavily's platform enables enterprise AI agents to securely search and crawl the web while adhering to company-specific policies. The funding will accelerate product development and expansion, aiming to onboard 'the next billion agents to the web.'

CoreWeave to Acquire Core Scientific in $9 Billion All-Stock Deal

CoreWeave, a U.S.-based cloud infrastructure company specializing in AI workloads, has entered into an all-stock agreement to acquire Core Scientific for approximately $9 billion. This strategic move aims to bring data center operations in-house, enhancing CoreWeave's infrastructure control and scalability amid growing demand for AI and high-performance computing solutions. The transaction is expected to close in the fourth quarter of 2025, pending regulatory approvals.