All AI Intelligence

A new open-source WhatsApp MCP server enables AI clients like Claude to search and read chat history via the Model Context Protocol. Deployed on Cloudflare Workers, it uses a 'phone-as-source-of-truth' architecture with periodic syncs to a local SQLite database, offering read-only access to enhance privacy while solving the challenge of integrating IM data into agents.

• •Enables AI assistants to directly retrieve WhatsApp message history without manual context copying
• •Adopts a one-way read-only design to prevent accidental sends, storing data within the user's own Cloudflare account
• •Leverages Cloudflare Durable Objects to run SQLite, providing a low-cost private deployment solution

Netflix releases the v1 set of models for its Emmy-winning perceptual video quality metric, VMAF. The update features speed optimizations, a new API design, and a codec-evaluation-friendly NEG mode to enhance practical utility.

• •Netflix releases the v1 model set for VMAF, continuing its Emmy-winning multi-method fusion architecture.
• •The update focuses on computational speed optimization and a more user-friendly API design.
• •A new NEG mode (No Enhancement Gain) is introduced specifically to facilitate codec evaluation scenarios.

Anthropic announced the creation of a 'public benefit fund' by committing 5% of its shares to an internal entity, with options to contribute more as company value increases. The move aims to provide flexibility for societal good projects, though specific governance and allocation criteria remain undisclosed.

• •Anthropic commits 5% of shares to a wholly-owned public benefit entity
• •Option to add 2% more shares for every 50% increase in company valuation
• •Funds intended for large nonprofit donations or direct distributions

The article argues that developers should abandon building complex 'personal software factories' and instead use AI chat interfaces directly for rapid prototyping. The author contends that over-engineered custom toolchains are a waste of time in the current era, as non-developers are achieving more creative results with simple prompts.

• •Building personal software factories often serves as an escape into comfort zones rather than a genuine productivity boost.
• •At this stage, directly prompting AI and observing results is more efficient than maintaining complex automated workflows.
• •Non-technical users are achieving highly creative applications through simple AI interactions, proving the value of low-barrier tools.

Anthropic has established the Anthropic Public Benefit Fund, committing an initial 5% of its shares with a mechanism to add 2% for every 50% increase in company valuation. This structure aims to provide flexible capital for public benefit initiatives like AI safety research and labor market mitigation, distinct from internal R&D.

• •Anthropic establishes an independent Public Benefit Fund with an initial 5% equity stake.
• •Implements a dynamic equity mechanism: adding 2% stake for every 50% increase in company valuation.
• •Focuses on externalities such as AI safety, biomedical research, and mitigating labor market disruptions.

Linear founder Karim Vardanyan argues that while AI agents accelerate development, great products still stem from human intuition and taste. He rejects the 'software factory' model, advocating for deep problem understanding and subjective judgment in an era of increasing automation.

• •While AI agents automate requirements and building, they cannot replace a team's unique understanding and style.
• •Linear values 'intuition over data,' defining intuition as rapid pattern recognition born from experience, not arbitrary preference.
• •As coding barriers drop, competitive advantage shifts from execution speed to deep insight into problem essence and aesthetic consistency.

Following Character.AI's bankruptcy filing, Disney has agreed to hire at least 60 employees, the majority of its workforce, as part of a broader $20 million asset acquisition deal. This move allows Disney to rapidly integrate top-tier AI talent under its new CTO while providing a soft landing for the startup team amidst ongoing safety controversies.

• •Disney commits to hiring at least 60 Character.AI employees (the majority) as part of a $20M asset acquisition.
• •Former Character.AI CEO Karandeep Anand now serves as Disney's first-ever CTO, leading its AI strategy.
• •The staffing deal follows Character.AI's bankruptcy filing amid child safety investigations, representing a 'soft landing' for the team.

Meta released Muse Spark Open, its strongest US open-weight model featuring a sparse MoE architecture with million-token context and multimodal input, optimized for long-horizon agent tasks. While excelling in tool invocation and error recovery, it requires dual M5 Ultra Mac Studios for local deployment and trails competitors like Zhipu and Kimi in coding benchmarks.

• •Muse Spark Open features 816B total parameters with ~42B active per token and supports a 1M token context window.
• •Core strengths lie in end-to-end tool invocation and failure recovery for long-horizon agents, but it lacks native image/audio/video generation capabilities.
• •Local deployment requires two M5 Ultra Mac Studios; it scores lower than Zhipu and Kimi on coding benchmarks like Terminal-Bench.

Chen Guangyu, a 17-year-old researcher who contributed to Kimi's core architectures like Attention Residuals and MoMoE, has left Moonshot AI for the San Francisco Bay Area. He explicitly denied rumors of joining Anthropic, leaving his next employer unconfirmed.

• •Chen Guangyu officially left Moonshot AI for the SF Bay Area but explicitly denied joining Anthropic.
• •He was a co-first author on key Kimi infrastructure projects including Attention Residuals and MoMoE.
• •The departure of this prodigy, who joined system-level LLM R&D at age 16, sparks discussion on young AI talent mobility.

UT Austin mathematician Francesco Maggi analyzes the potential and limits of LLMs in generating mathematical proofs. The post contrasts a 'golden age' scenario—integrating AI to empower scientists while preserving independence—with a 'dark age' where short-term output spikes may erode human scientific capability. It highlights structural risks in AI-assisted research.

• •LLMs show promise in proof generation but lack clear quality boundaries
• •'Golden Age' requires balancing empowerment with scientific openness and independence
• •'Dark Age' risk involves short-term efficiency gains eroding human research capacity

OpenAI, Meta, and Google want AI agents to do everything for you. But popular services on the web aren’t letting that happen. STK414AICVIRGINIAC Autonomous agents are the hottest thing in AI. OpenAI, Meta, Google, Instinct, and others promise their agents can scour the internet on your behalf, doing anything from booking a dinner reservation and cleaning out your inbox to buying new shoes. But what happens when the rest of the internet won’t play nice? In recent weeks, users of services like Met...

Nvidia mandates Samsung, SK Hynix, and Micron to use TSMC's 3nm process for next-gen NVHBM base dies to ensure performance consistency across suppliers. Samsung breaks its traditional one-stop production model to comply, while retaining in-house fabrication as a capacity backup plan.

• •Nvidia standardizes NVHBM base dies on TSMC 3nm and core dies on 1c DRAM to eliminate supplier variability and streamline validation.
• •Samsung abandons its one-stop production model for Nvidia's custom HBM for the first time, signaling a shift in high-end memory manufacturing division.
• •Samsung retains in-house production for standard HBM and AMD clients, while validating internal capabilities for NVHBM as a hedge against TSMC capacity limits.

Anthropic has proposed an AI safety bill that only triggers regulatory action upon incidents causing at least 50 deaths or $1 billion in damages. The proposal aims to define severe AI accidents through high-threshold accountability standards but remains a policy suggestion rather than enacted law.

• •Anthropic's proposed AI safety bill sets an extremely high trigger threshold: 50 deaths or $1 billion in damages.
• •The proposal attempts to focus regulation on 'catastrophic' outcomes rather than general model errors or minor violations.
• •As an industry leader, Anthropic's stance may influence the direction of US and global AI regulatory frameworks.

Distributor Oderço reported that a shipment of Nvidia GeForce RTX 5090 GPUs worth approximately 2.5 million reais vanished after arriving at São Paulo's Guarulhos International Airport. While the pallets were confirmed to have entered the terminal, the devices disappeared during subsequent handling with no current tracking information.

• •A shipment of RTX 5090 GPUs valued at approximately 2.5 million reais (about 3.3 million CNY) went missing at São Paulo airport.
• •The cargo pallets were confirmed to have arrived at the terminal but disappeared during internal handling processes.
• •The distributor stated this is the first time such a large-scale loss has occurred in their decades-long import history.

NVIDIA achieved a 1.68x inference speedup for the 2.8T-parameter Kimi K3 model on GB300 NVL72 hardware via upstream vLLM optimizations, reaching 423 tok/s/user. The breakthrough relies on a massive 12,000-line Megakernel written in NVIDIA's new Cute DSL, demonstrating the future of AI-assisted generation of advanced compute kernels.

• •Kimi K3 (2.8T params) achieves 423 tok/s/user on GB300 NVL72, a 1.68x improvement over baseline.
• •The optimization centers on a 'Megakernel' decode-specific kernel with 12,000 lines of code, written in NVIDIA's new Cute DSL.
• •This case signals that AI-assisted generation of advanced compute kernels is becoming a key direction for solving LLM inference efficiency bottlenecks.