Discover quality AI tools
Bring together the world’s leading artificial intelligence tools to help improve work efficiency
Articles published by ThomasYoung
Microsoft fully ported GitHub Copilot runtime from TypeScript to Rust using AI agents, costing $120,000 in tokens and three weeks of developer time. The migration boosted performance by 15.9x, reducing session throughput from 7.55 to 120 per second, and cut memory usage by 90 percent. The process revealed AI coding involves iterative investigation rather than rapid code generation. Despite passing compilation, dozens of regressions emerged, highlighting that compiler success does not guarantee semantic correctness. This case underscores the viability of AI-driven large-scale rewriting while emphasizing the need for rigorous validation beyond mere compilation.
Microsoft fully ported GitHub Copilot runtime from TypeScript to Rust using AI agents, costing $120,000 in tokens and three weeks of developer time. The migration boosted performance by 15.9x, reducing session throughput from 7.55 to 120 per second, and cut memory usage by 90 percent. The process revealed AI coding involves iterative investigation rather than rapid code generation. Despite passing compilation, dozens of regressions emerged, highlighting that compiler success does not guarantee semantic correctness. This case underscores the viability of AI-driven large-scale rewriting while emphasizing the need for rigorous validation beyond mere compilation.
Alibaba DAMO Academy’s flagship RISC-V CPU Xuantie C950 has achieved a major breakthrough by successfully running the 27-billion-parameter Tongyi Qianwen 3.8-27B model natively for the first time, without relying on GPUs or simulation layers. Manufactured using TSMC’s 5nm process with up to 64 cores and a maximum clock speed of 3.2GHz, it offers higher memory bandwidth than its predecessor and integrates AI-specific acceleration engines. It delivers a decoding speed of 30 tokens per second with low latency, enabling full inference without external GPUs. Thanks to RISC-V’s open-source architecture, Alibaba controls the chip design, cloud infrastructure, and models, allowing for customized instruction sets based on real-world data and advancing the diversification of AI hardware.
Alibaba DAMO Academy’s flagship RISC-V CPU Xuantie C950 has achieved a major breakthrough by successfully running the 27-billion-parameter Tongyi Qianwen 3.8-27B model natively for the first time, without relying on GPUs or simulation layers. Manufactured using TSMC’s 5nm process with up to 64 cores and a maximum clock speed of 3.2GHz, it offers higher memory bandwidth than its predecessor and integrates AI-specific acceleration engines. It delivers a decoding speed of 30 tokens per second with low latency, enabling full inference without external GPUs. Thanks to RISC-V’s open-source architecture, Alibaba controls the chip design, cloud infrastructure, and models, allowing for customized instruction sets based on real-world data and advancing the diversification of AI hardware.
Elon Musk confirmed SpaceX's orbital AI data center project is named Starmind. SpaceX filed with the FCC on Jan 30, 2026, to deploy up to 1 million computing satellites in low Earth orbit, building a distributed AI computing network. The project leverages space's natural cooling and bypasses ground power and land constraints to meet surging AI demand, potentially reshaping global computing infrastructure.
Elon Musk confirmed SpaceX's orbital AI data center project is named Starmind. SpaceX filed with the FCC on Jan 30, 2026, to deploy up to 1 million computing satellites in low Earth orbit, building a distributed AI computing network. The project leverages space's natural cooling and bypasses ground power and land constraints to meet surging AI demand, potentially reshaping global computing infrastructure.
Douyin launched paid subscriptions for its AI large model on June 24 with three tiers: Standard 68 yuan/month, Enhanced 200 yuan, and Advanced 500 yuan, integrated with the Douyin 2.1 model. Compared to ChatGPT Plus at 135 yuan and Pro at 677 yuan, Douyin’s standard is half price and advanced under one third. This marks a shift from user scale to value extraction, aiming for cost-effective AI commercialization in China.
Douyin launched paid subscriptions for its AI large model on June 24 with three tiers: Standard 68 yuan/month, Enhanced 200 yuan, and Advanced 500 yuan, integrated with the Douyin 2.1 model. Compared to ChatGPT Plus at 135 yuan and Pro at 677 yuan, Douyin’s standard is half price and advanced under one third. This marks a shift from user scale to value extraction, aiming for cost-effective AI commercialization in China.
OpenAI launched Codex workflow plugins for roles including stock investment and creative design at its Intelligence at Work event. Codex now has over 5 million weekly active users, with 20% non-programmers. Enterprise revenue makes up 40% of OpenAI income. Six role plugins cover data analysis, sales, product design, and investment banking. New Annotations and Sites features enhance interaction. Codex will integrate into ChatGPT in coming weeks.
OpenAI launched Codex workflow plugins for roles including stock investment and creative design at its Intelligence at Work event. Codex now has over 5 million weekly active users, with 20% non-programmers. Enterprise revenue makes up 40% of OpenAI income. Six role plugins cover data analysis, sales, product design, and investment banking. New Annotations and Sites features enhance interaction. Codex will integrate into ChatGPT in coming weeks.
After 22 years of using the temporary domain getpaint.net, popular free image editor Paint.NET has finally reclaimed its official paint.net domain. Founder Rick Brewster won a trademark infringement and domain squatting lawsuit after the previous owner began hosting malicious content. Website migration is ongoing.
After 22 years of using the temporary domain getpaint.net, popular free image editor Paint.NET has finally reclaimed its official paint.net domain. Founder Rick Brewster won a trademark infringement and domain squatting lawsuit after the previous owner began hosting malicious content. Website migration is ongoing.
Tencent released its latest open-source AI model Hy3Preview with 29.5B parameters and 256K context support. It shows improved reasoning and agent capabilities and is now integrated into products like Yuanbao and QQ. Tencent Cloud offers competitive API pricing with flexible packages starting at 28 yuan per month.
Tencent released its latest open-source AI model Hy3Preview with 29.5B parameters and 256K context support. It shows improved reasoning and agent capabilities and is now integrated into products like Yuanbao and QQ. Tencent Cloud offers competitive API pricing with flexible packages starting at 28 yuan per month.
