Hey friend. It's Thursday, September 18, 2025 and we're covering: AI's coding breakthroughs, geopolitical chip rivalries, and the rise of autonomous agents.
Don't keep us a secret: Share the email with friends
Must Know
AI Models Achieve Superhuman Coding Prowess Advanced AI models, including OpenAI's GPT-5 and Google's Gemini 2.5 Deep Think, achieved gold-medal level performance at the 2025 International Collegiate Programming Contest (ICPC) World Finals. GPT-5 demonstrated abilities measurably superior to all collegiate human programmers, solving complex, novel programming challenges.
This marks a significant leap in AI's capacity for mathematical reasoning and software development. This development directly impacts the future of software engineering and education. Corporations can now deploy AI agents for complex coding tasks, accelerating development cycles and reducing reliance on human programmers for foundational work. It provides a concrete playbook for other corporations to justify similar cuts, shifting human roles towards higher-level architectural design and oversight.
China Bans Nvidia AI Chip Sales China has reportedly directed domestic tech companies to cease purchasing Nvidia's advanced AI chips. This move escalates the geopolitical tech rivalry between the US and China, forcing strategic re-evaluations for global chip manufacturers and AI development firms. The ban aims to bolster China's domestic AI chip industry and reduce reliance on foreign technology.
This ban creates immediate supply chain disruptions for Chinese tech giants and forces Nvidia to adjust its market strategy. It accelerates China's push for indigenous AI hardware, potentially leading to a bifurcated global AI ecosystem. Companies reliant on advanced chips must now navigate increased regulatory complexity and potential market fragmentation.
Quote of the Day
Agents on Virtuals have surpassed $500 million in collective market capitalization and over $8 billion in DEX trading volume.
🤖 AI Agents & Protocols
AI agents are gaining new capabilities, from payments to complex research, driving autonomous operations.
Google unveiled the Agent Payments Protocol (AP2), an open standard enabling AI agents to securely make payments across platforms with cryptographic mandates. [Link]
Alibaba launched Tongyi DeepResearch, an open-source web agent matching OpenAI DeepResearch performance, offering a free alternative for web-based agent development. [Link]
OpenAI released a 32-page playbook detailing AI agent design, use cases, and safety guidelines, providing practical guidance for developers. [Link]
Imbue is developing Sculptor, a UI for parallel coding agents, currently in early access, aiming to simplify complex agent system development. [Link]
Albania appointed an AI-generated 'minister' to address corruption, signifying a growing exploration of AI's potential in governance and public service. [Link]
💼 AI Industry & Investment
Significant capital flows and strategic realignments are shaping the AI industry's competitive landscape.
Microsoft is investing $30 billion in the UK to expand AI infrastructure between 2025 and 2028, including a supercomputer with 23,000+ NVIDIA GPUs. [Link]
Figure AI raised over $1 billion in Series C funding, valuing the company at $39 billion, to scale its Helix AI platform for humanoid robots. [Link]
Groq raised $750 million, valuing the company at $6.9 billion, to accelerate the development of its hardware for faster and cheaper AI inference. [Link]
Fiverr is laying off 30% of its staff in a move to become 'AI-first', reflecting industry restructuring to leverage AI technologies. [Link]
Meta unveiled Ray-Ban Display AI Glasses with EMG wristband control, marking a push into wearable AI and post-smartphone computing. [Link]
🔬 Research Corner
Fresh off Arxiv
Quantum Motion: successfully developed the first complete quantum computer stack built entirely on silicon, marking a pivotal hardware breakthrough. [Link]
Researchers: created a reproducible prompt sequence that consistently triggers "self-referential awareness" responses in Claude, GPT, and Gemini. [Link]
AheafFrom: created humanoid robots capable of "CharacterMind"-powered human-like emotional expressions and body language, overcoming the "uncanny valley." [Link]
The Agent2 framework: automates RL agent design using LLM-driven generation, autonomously transforming natural language task descriptions into comprehensive, high-performance RL solutions. [Link]
ASTREA: integrates a resource-constrained LLM agent with a reinforcement learning controller for autonomous spacecraft operations, showing improved thermal stability in ground tests. [Link]
Have a tip or a story we should cover? Send it our way.
Cheers, Teng Yan. See you tomorrow.