Briefings

The morning briefing.

While you slept: AI agents shift strategies at ByteDance and Microsoft, chip shortages drive global infrastructure investment, and new research reveals AI's hidden learning costs.

RIGHT NOW, IN ONE BREATH

AI Agent Evolution. The landscape for AI agents is rapidly shifting, with major players like ByteDance and Alibaba sunsetting existing agent features on platforms like Doubao and Qwen. This move comes as Microsoft signals its intent to consolidate its consumer and enterprise Copilot apps, introducing new "AutoPilot" agents for background tasks. These developments suggest a strategic pivot towards more integrated and perhaps less visible agent functionalities. OpenAI cofounder Greg Brockman also envisions a future with "almost no interface" where software learning becomes obsolete.

Compute and Infrastructure. The global demand for AI is driving a significant surge in semiconductor prices, with over 20 chip manufacturers announcing second price hikes this year and lead times extending to mid-2027. This scarcity is prompting governments, including those in France and India, to actively court AI giants for data center investments and cloud infrastructure. Huawei is also advancing home broadband with its AI-FAN architecture, aiming for 10-gigabit connectivity, highlighting the foundational infrastructure required for widespread AI adoption.

Societal Impact and Policy. New research reveals a concerning long-term impact of AI on learning, with a 26,000-student study showing AI users performing up to 24% worse on exams after two years, despite initial homework benefits. This highlights the hidden costs of AI in education. Simultaneously, UK parents are receiving stark warnings from the National Crime Agency regarding the dangers of posting children's images online due to the rise of AI-generated sexual abuse material. These issues underscore the urgent need for robust policy and ethical considerations as AI integrates further into daily life.

Frontier Research and Applications. In a significant move, Anthropic is launching its own drug discovery programs focused on neglected diseases, aiming to leverage AI where Big Pharma sees no profit. This initiative showcases AI's potential for high-impact applications beyond commercial viability. Meanwhile, Alibaba DAMO Academy's ElementsClaw AI agent has discovered four new superconductors in just 28 GPU hours, demonstrating rapid material science breakthroughs. ByteDance researchers have also identified a new scaling law for AI agents, suggesting a path to sustained improvement in real-world task learning.

68
Sources scanned
5,660
Headlines processed
#231
Edition
28.5k
Discussing now

Anthropic launches drug discovery programs for neglected diseases

Anthropic is initiating its own drug development program to address neglected diseases that the pharmaceutical industry deems unprofitable. This effort aims to leverage AI to potentially cut development time and increase success rates for new treatments.

ByteDance's Doubao and Alibaba's Qwen to sunset AI agent features

ByteDance's Doubao and Alibaba's Tongyi Qianwen (Qwen) have announced the simultaneous removal of their AI agent features. This change will take effect on July 15, impacting users of both platforms.

Microsoft to merge Copilot apps and introduce new AutoPilot agents

Microsoft reportedly plans to merge its consumer and enterprise Copilot apps into a single platform by August. The company will also introduce new "AutoPilot" AI agents designed to handle background tasks for an additional fee.

Semiconductor prices skyrocket as AI demand fuels chip shortage

Over 20 chip manufacturers have announced their second price hikes for 2026, with AI-specific MLCCs seeing a 3-10x surge. Delivery lead times for these components are now extending to mid-2027 due to high AI demand.

Governments court AI giants for data center investment

French President Macron and Indian Prime Minister Modi are actively courting tech CEOs to attract AI data center investment and cloud infrastructure. Both nations are seeking to bolster their domestic AI capabilities.

Study shows AI's hidden learning cost takes two years to surface

A study of over 26,000 Chinese students revealed that while AI users completed homework faster and scored higher initially, they performed up to 24% worse on exams after two years. This suggests short-term studies may underestimate AI's long-term educational impact.

UK parents warned over posting images of children amid AI sexual abuse fears

The UK's National Crime Agency and Internet Watch Foundation have issued guidance warning parents against posting children's photos online. This advisory addresses the rising threat of AI-generated sexual abuse material.

Kuaishou spins off Kling AI with $3B funding round

Kuaishou has spun off its Kling AI video generation unit, securing $3 billion in funding at an $18 billion valuation. Tencent, Alibaba Cloud, and Baidu participated as investors in this significant round.

Embodied AI robots begin entering Chinese homes for household chores

Embodied AI robots are now available for hire in Chinese homes for household chores, with services starting at 74 yuan for three hours. This development sparks debate on the practical value of substantial investments in embodied AI.

Mistral's open-source Leanstral 1.5 aces formal math benchmarks and catches real bugs

Mistral AI has released Leanstral 1.5, an open-source model for formal verification in Lean 4, which excels in math benchmarks. The model also identified five previously unknown bugs across 57 open-source repositories.

Alibaba DAMO Academy's ElementsClaw AI agent discovers 4 new superconductors

Alibaba DAMO Academy's AI agent, ElementsClaw, has discovered four previously unknown superconductors. The agent achieved this by screening 2.4 million crystal structures in only 28 GPU hours.

ByteDance discovers new scaling law that could sustain AI boom

Researchers at ByteDance's Seed AI team have discovered a new scaling law that governs how AI agents improve by performing real-world tasks. This finding suggests AI agents can double their learning speed every three months through interaction.

Apple turns Safari into something AI agents can control

Apple's WebKit team has shipped Safari Technology Preview 247 with a built-in Model Context Protocol server. This integration allows AI agents to control aspects of the Safari browser.

Confidential computing's core trust mechanism is broken

The core trust mechanism in confidential computing, specifically attested TLS, is reportedly broken, making it difficult to verify the identity of the other end of a connection. Experts suggest a definitive fix for this vulnerability may not yet exist.