hotai.newsAn archive of 20 AI news briefs collected on 2026-07-07. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
Anthropic's guide to the Claude Code effort level and model selection: when to raise or lower effort—low, medium, high, and max—and how to choose between Claude Fable, Opus, and Sonnet.
Source links can be verified on the original siteRead original ↗Checking the site connection security
Source links can be verified on the original siteRead original ↗The current environment is abnormal. After completing verification, you can continue to access. Go to verification
Source links can be verified on the original siteRead original ↗The current environment is abnormal. After completing verification, you can continue to access. Go to verification
Source links can be verified on the original siteRead original ↗In the latest iOS 27 developer beta, Siri AI can already read third-party app data, such as the remaining battery of electric vehicles, and will request access permission from users. This marks Siri evolving from a mere voice assistant to an agent capable of deeply integrating app data, and may unlock more scenarios in the future.
Source links can be verified on the original siteRead original ↗A half-yearly report from the Bank of England points out that the AI boom, while pushing up the stock prices and borrowing of related companies, has also significantly amplified the cyber risks of the financial system. The report warns that the market's overly optimistic bets on and leverage of AI may amplify future volatility, and the AI technology itself also makes banks more vulnerable to attacks. Regulators are paying attention to the new challenges brought by AI autonomous systems.
Source links can be verified on the original siteRead original ↗The company just raised $7 million in seed funding, and is launching its app for iPhone and Android on Tuesday.
Source links can be verified on the original siteRead original ↗Midjourney founder asks online, voicing the thoughts of many programmers: after using the latest AI coding models, efficiency is extremely high, but accompanied by a huge sense of physical and mental exhaustion. Former Meta engineer, Claude Code lead and others left comments one after another, analyzing the root causes and sharing ways to alleviate them, revealing that 'AI fatigue' is becoming a new trouble for the developer community.
Source links can be verified on the original siteRead original ↗The product focuses on children's education, supports call wake-up, touch and shake multi-feedback, connects to a large model with a massive knowledge base, and a companion App can record children's daily activities.
Source links can be verified on the original siteRead original ↗This modular framework, named Avalon, integrates various attack capabilities, allowing hackers to load different functions as needed, offering flexibility far beyond traditional single-purpose malware. Although its code quality is low, the design is complete, demonstrating that generative AI is helping hackers with limited technical skills rapidly develop complex tools, posing new challenges to enterprise defenses.
Source links can be verified on the original siteRead original ↗Checking the site connection security
Source links can be verified on the original siteRead original ↗Chinese AI models are gaining ground with US companies because they cost far less than systems from OpenAI and Anthropic, CNBC reports.
Source links can be verified on the original siteRead original ↗Chinese startup Deepseek is building its own AI chip, Reuters reports.
Source links can be verified on the original siteRead original ↗Skip to content Look for the robot helpers Top robotics researchers and founders explain how robot autonomy is evolving. Agility's Digit humanoid robots are working in warehouses and on factory floors. Credit: Agility Robotics
Source links can be verified on the original siteRead original ↗Checking the site connection security
Source links can be verified on the original siteRead original ↗Ant Lingbo Technology releases spatial perception model LingBot-Depth 2.0, with training data surging from 3 million to 150 million, taking 12 first places among 16 depth completion benchmark evaluations. The new model performs prominently in scenarios where traditional depth cameras easily fail, such as glass and mirrors, with depth error halved compared to the previous generation. The simultaneously launched vision foundation model LingBot-Vision achieves better-than-DINOv3 precision with a smaller amount of data.
Source links can be verified on the original siteRead original ↗OpenAI, Anthropic, and major cloud providers are racing to outbid each other with free compute credits to pull startups into their ecosystems. Some individual offers top $3 million. At Y Combinator alone, OpenAI and Anthropic could hand out up to $800 million in credits per year combined. The discount war comes at a time when both companies need to improve their margins ahead of upcoming IPOs.
Source links can be verified on the original siteRead original ↗At the first UN Global Dialogue on AI Governance, Guterres told participants, 'We may be the last generation able to set the terms for human-machine coexistence.'
Source links can be verified on the original siteRead original ↗As prices for top U.S. AI models continue to rise, cheaper Chinese models with rapidly improving performance are gaining favor among American companies. Data from OpenRouter shows that the peak share of tokens used by U.S. enterprises calling on Chinese models has reached 46%. Models from vendors such as DeepSeek and Zhipu AI have performed close to top-tier products in multiple evaluations, yet cost only one-fifth or even less.
Source links can be verified on the original siteRead original ↗Pruning agent context down to what the answer actually needs, while keeping 96% of recall
Source links can be verified on the original siteRead original ↗