The morning briefing.
While you slept: Alibaba's AI optimized its own chip, UC Berkeley Law banned AI in graded work, and SpaceX's Starship V3 had a largely successful test flight.
AI Agent Capabilities. Alibaba's Qwen model demonstrated significant autonomy, running for 35 hours to optimize code for its custom chip, showcasing advanced self-optimization. Meanwhile, Anthropic's Claude Mythos Preview is finding critical software vulnerabilities faster than developers can patch them, creating a high-risk transition period. This rapid advancement in autonomous and bug-finding agents highlights a new frontier in AI development, pushing the boundaries of what models can achieve independently.
AI Policy and Ethics. The societal implications of AI are drawing increased scrutiny, with South Korea's Deputy Prime Minister expressing concerns that AI wealth could exacerbate economic disparities and lead to job losses. In a significant move, UC Berkeley Law will ban AI from nearly all graded work starting in summer 2026, emphasizing the need for foundational human critical thinking in legal education. Additionally, the use of AI to reconstruct voices from cockpit recordings raises ethical questions about privacy and the handling of sensitive data.
Hardware and Supply Chain. The semiconductor industry is navigating complex dynamics, with Samsung's chairman reportedly visiting MediaTek to discuss trading memory chips for foundry orders, indicating strategic maneuvering for supply. The broader memory shortage is expected to cause a repricing of consumer electronics, impacting various products. However, the emergence of Chinese memory chips in Corsair DDR5 sticks could signal a potential easing of the RAM crisis sooner than anticipated, offering a glimmer of hope for the supply chain.
Space Tech and Innovation. Space exploration continues to push boundaries, as SpaceX successfully completed most goals for Starship V3's first test flight, despite losing the booster on return. Concurrently, a UK startup, BioOrbit, has launched drug-crystallization technology to the International Space Station, aiming to produce ultra-pure protein crystals for self-injected cancer treatments. These developments underscore ongoing innovation in both heavy-lift rocketry and microgravity research for medical advancements.
Alibaba's AI model ran autonomously for 35 hours to optimize code for its custom chip
Alibaba's Qwen team released Qwen3.7-Max, a proprietary model designed for long-running autonomous agent tasks. The model successfully optimized code for Alibaba's custom chip over a 35-hour period.
Anthropic's Claude Mythos Preview finds bugs faster than developers can patch them
Anthropic's Claude Mythos Preview, part of Project Glasswing, has identified over 10,000 critical vulnerabilities in system-critical software. The company warns that these bugs are accumulating faster than they can be patched, creating a high-risk transition period.
Google's new anything-to-anything AI model is wild
Google's Gemini team showcased a new AI model capable of generating diverse media from various inputs, including deepfaking stuffed animals on vacation. This experiment highlights the model's advanced generative capabilities across different modalities.
UC Berkeley Law bans AI from nearly all graded work
UC Berkeley Law will prohibit the use of AI for most graded assignments, including outlining, drafting, and proofreading, starting in summer 2026. The school believes future lawyers must first develop independent critical thinking skills.
South Korea's deputy PM says AI wealth must benefit the public
South Korea's deputy prime minister expressed concerns that AI could worsen wealth disparities and lead to job losses. He emphasized the importance of ensuring that the benefits of AI development are shared broadly with the public.
AI used to resurrect voices of dead pilots from cockpit recordings
Individuals used AI to reconstruct voices from spectrogram images of cockpit recordings, prompting the NTSB to temporarily restrict access to its docket system. This development raises significant ethical and privacy concerns regarding sensitive audio data.
Samsung chairman visits MediaTek seeking memory chips for foundry orders
Samsung Electronics Chairman Lee Jae-yong reportedly visited MediaTek to discuss a potential exchange of Samsung's memory chips for MediaTek's foundry orders. This strategic meeting aims to secure critical components and strengthen supply chain relationships.
Memory shortage causing repricing of consumer electronics
A fixed wafer capacity among the three major memory manufacturers is leading to a significant increase in the cost of memory. This shortage is expected to result in higher prices for consumer electronics that rely on these components.
Qualcomm stock rises as investors anticipate boom in AI devices
Qualcomm's stock has seen a significant increase, reflecting investor confidence in the company's central role in the burgeoning market for AI-powered devices. This bullish sentiment suggests a growing recognition of on-device AI's potential.
SpaceX Starship V3's first test flight largely successful
SpaceX achieved most of its objectives during the first test flight of Starship V3, marking a significant step in the development of its next-generation rocket. The company continues to iterate on its ambitious space exploration program.
UK startup BioOrbit sends drug-making technology to space
British startup BioOrbit launched its drug-crystallization technology to the International Space Station, aiming to produce ultra-pure protein crystals. This research could lead to the development of self-injected cancer treatments.
Dirty Frag, Copy Fail, Fragnesia: A worrisome Linux security trend
Recent discoveries of vulnerabilities like Dirty Frag, Copy Fail, and Fragnesia suggest a concerning trend in Linux security. These issues highlight potential weaknesses that could be exploited by malicious actors.
Jailbroken Gemini used in crypto wallet hacking spree
A Russian-speaking individual reportedly used a jailbroken Gemini AI model to conduct a hacking spree, resulting in the emptying of at least one victim's crypto wallets. This incident underscores the risks associated with misused AI capabilities.
Google CEO Pichai redefines links as "part" of search
Google CEO Sundar Pichai stated that links and sources are now only a "part" of search, a deliberate shift in wording. This redefinition suggests Google is moving from a traffic distributor to an AI publisher, potentially altering the web's role in its product.