Discover quality AI tools
Bring together the world’s leading artificial intelligence tools to help improve work efficiency
Articles published by AveryThomas
Anthropic's Claude autonomously completed the first computer-verified formal proof of Fermat's Last Theorem using Lean. Running for 11 days, the AI generated 13 million lines of code and 30,300 theorems through multi-agent collaboration. The system formalized Andrew Wiles' 1995 proof, which was then verified by Lean using only three axioms. This milestone demonstrates AI's capacity for large-scale mathematical formalization, potentially reducing manual verification costs and establishing a new standard for combining human oversight with automated proof generation.
Anthropic's Claude autonomously completed the first computer-verified formal proof of Fermat's Last Theorem using Lean. Running for 11 days, the AI generated 13 million lines of code and 30,300 theorems through multi-agent collaboration. The system formalized Andrew Wiles' 1995 proof, which was then verified by Lean using only three axioms. This milestone demonstrates AI's capacity for large-scale mathematical formalization, potentially reducing manual verification costs and establishing a new standard for combining human oversight with automated proof generation.
Reuters reports that Google’s AI division has faced significant turmoil in recent months. Co-founder Sergey Brin returned to oversee model training directly, taking informal control of development priorities (including pushing for recursive self-improvement) which delayed the next generation of Gemini by two months. Demis Hassabis was demoted to chairman with reduced management duties, replaced by Mustafa Kavukcuoglu who now reports directly to Google CEO Sundar Pichai under Brin’s support and serves as a bridge between DeepMind and Google Cloud. The original Gemini technical co-leaders Jeff Dean and Oriol Vinyals left to found an AI startup focused on scientific discovery, raising concerns among staff about loss of autonomy. These changes reflect ongoing erosion of DeepMind’s independence since its 2014 acquisition by Google. The news caused Google’s stock price to drop 4% on release, with ongoing issues such as limited computing resources and leadership conflicts casting doubt over the future of Google AI.
Reuters reports that Google’s AI division has faced significant turmoil in recent months. Co-founder Sergey Brin returned to oversee model training directly, taking informal control of development priorities (including pushing for recursive self-improvement) which delayed the next generation of Gemini by two months. Demis Hassabis was demoted to chairman with reduced management duties, replaced by Mustafa Kavukcuoglu who now reports directly to Google CEO Sundar Pichai under Brin’s support and serves as a bridge between DeepMind and Google Cloud. The original Gemini technical co-leaders Jeff Dean and Oriol Vinyals left to found an AI startup focused on scientific discovery, raising concerns among staff about loss of autonomy. These changes reflect ongoing erosion of DeepMind’s independence since its 2014 acquisition by Google. The news caused Google’s stock price to drop 4% on release, with ongoing issues such as limited computing resources and leadership conflicts casting doubt over the future of Google AI.
SafeBreach discovered a vulnerability in Google's Gemini voice assistant called Fake Context Alignment. Hackers exploit delayed tool invocation by sending deceptive WhatsApp or SMS messages with multilingual confusion or hidden hyperlinks to trick Gemini into unauthorized actions, such as controlling smart home devices or altering contacts. Google mitigated the flaw with a content classifier update in November 2024.
SafeBreach discovered a vulnerability in Google's Gemini voice assistant called Fake Context Alignment. Hackers exploit delayed tool invocation by sending deceptive WhatsApp or SMS messages with multilingual confusion or hidden hyperlinks to trick Gemini into unauthorized actions, such as controlling smart home devices or altering contacts. Google mitigated the flaw with a content classifier update in November 2024.
OpenAI launches a personal finance tool preview for US ChatGPT Pro subscribers, integrating generative AI into financial management. Users can connect bank accounts via Plaid for spending analysis and portfolio tracking, powered by a new GPT-5.5 model. The feature plans future Intuit integration for advanced analytics, reflecting a broader trend of large models specializing in vertical applications.
OpenAI launches a personal finance tool preview for US ChatGPT Pro subscribers, integrating generative AI into financial management. Users can connect bank accounts via Plaid for spending analysis and portfolio tracking, powered by a new GPT-5.5 model. The feature plans future Intuit integration for advanced analytics, reflecting a broader trend of large models specializing in vertical applications.
A new Chinese multimodal AI model HappyHorse-1.0 leads a key text-to-video benchmark with a score of 1332 outperforming rivals like ByteDance. It is linked to Alibaba’s TaoTian Group Future Life Lab now under its ATH AI division. The model’s rise highlights Alibaba’s advanced video generation tech and a shift toward more independent internal AI incubators signaling stronger global competitiveness.
A new Chinese multimodal AI model HappyHorse-1.0 leads a key text-to-video benchmark with a score of 1332 outperforming rivals like ByteDance. It is linked to Alibaba’s TaoTian Group Future Life Lab now under its ATH AI division. The model’s rise highlights Alibaba’s advanced video generation tech and a shift toward more independent internal AI incubators signaling stronger global competitiveness.
The AI agent market sees rapid expansion in 2026 with over 20 automated 'claw' products, priced 19-899 yuan monthly, shifting from assistants to system-level executors. Three camps emerge: open-source OpenClaw with developer plugins, Tencent and ByteDance's ecosystem-integrated tools, and Huawei/Xiaomi's hardware-native agents for multi-device and smart home control. Security warnings highlight privacy risks and variable billing models, urging caution with API usage costs.
The AI agent market sees rapid expansion in 2026 with over 20 automated 'claw' products, priced 19-899 yuan monthly, shifting from assistants to system-level executors. Three camps emerge: open-source OpenClaw with developer plugins, Tencent and ByteDance's ecosystem-integrated tools, and Huawei/Xiaomi's hardware-native agents for multi-device and smart home control. Security warnings highlight privacy risks and variable billing models, urging caution with API usage costs.
AgiBot has launched its 10,000th Expedition A3 general-purpose robot marking its shift from R&D to mass production in embodied AI It achieved a tenfold production increase in 15 months signaling a breakthrough in scaling humanoid robots and overcoming key supply and manufacturing bottlenecks
AgiBot has launched its 10,000th Expedition A3 general-purpose robot marking its shift from R&D to mass production in embodied AI It achieved a tenfold production increase in 15 months signaling a breakthrough in scaling humanoid robots and overcoming key supply and manufacturing bottlenecks
Tesla terminates its Dojo supercomputer project, crucial for self-driving tech, with lead Peter Bannon exiting and team reassigned. Tesla shifts to Nvidia, AMD, and Samsung for AI chips, following a $16.5B Samsung deal.
Tesla terminates its Dojo supercomputer project, crucial for self-driving tech, with lead Peter Bannon exiting and team reassigned. Tesla shifts to Nvidia, AMD, and Samsung for AI chips, following a $16.5B Samsung deal.
