We missed last week (our bad!). To make up for it, below is a double-sized update stuffed full of the latest in AI.
what to know for now
🛠️ OpenAI expands AI agent toolkit. OpenAI’s new Responses API enables developers to build AI agents with web search, file retrieval, and system automation, replacing the Assistants API by 2026. The API includes GPT-4o search models, a file search tool, and the Computer-Using Agent (CUA) for automating workflows. OpenAI also released the open-source Agents SDK for integrating and monitoring AI agents. Read more
🚀 Google pushes AI across models, search, and image generation. Google releases Gemma 3, its latest open model, claiming top performance in LMArena benchmarks against Llama-405B and DeepSeek-V3. It scales from 1B to 27B parameters, supports function calling, and includes a 128k-token context window. AI-powered search expands with Gemini 2.0, improving coding, math, and multimodal queries while AI Mode enhances real-time reasoning in Search. A new experimental Gemini 2.0 Flash introduces native image generation, eliminating separate diffusion models and enabling conversational editing of images in real time.
🕵️ China's AI agent steps up. Manus, a new AI agent out of China, is gaining attention for outperforming OpenAI's Operator and Anthropic's Computer Use in personal assistant tasks. While still invite-only, it highlights China's continued AI progress, though concerns remain about security, reliability, and data privacy. Read more
🍔 McDonald's integrates AI across operations. The fast-food giant is rolling out edge computing, AI-powered drive-throughs, and predictive maintenance to improve efficiency and customer experience. Kitchen sensors will anticipate equipment failures, while computer vision ensures order accuracy. AI-driven virtual managers will assist with scheduling, and customer data will refine promotions. Read more
🤖 Beijing mandates AI education. Starting this fall, Beijing schools must provide at least eight hours of AI instruction per year, covering elementary through high school. The initiative aims to integrate AI ethics and a "teacher-student-machine" learning model. China continues its AI push, with homegrown startups and tech giants gaining ground against U.S. competitors. Read more
💽 Intel's CEO shakes up strategy. Lip-Bu Tan plans to restructure Intel’s manufacturing and AI operations, cutting middle management and refocusing Intel Foundry to attract major customers. He aims to accelerate AI chip production while fixing inefficiencies in Intel’s contract manufacturing efforts. Read more
Jake’s Take: Stanford Medicine researchers identified a naturally occurring peptide (called BRP), which suppresses appetite and reduces body weight similarly to semaglutide (Ozempic) but without side effects like nausea and muscle loss. Using AI, the team was able to quickly screen specific peptides and found that BRP significantly reduced food intake and fat accumulation in mice and minipigs. Clinical trials are planned to assess its effectiveness in humans.
If BRP delivers on its promise, AI-driven peptide discovery could easily accelerate the replacement of broad-spectrum metabolic drugs with more precise, side-effect-free interventions.
what to know for later
🚀 OpenAI sets high price for AI agents. OpenAI is reportedly planning to charge up to $20,000 per month for specialized AI "agents" designed for tasks like sales lead ranking, software engineering, and PhD-level research. Pricing tiers range from $2,000 for knowledge worker tools to $10,000 for developer-focused agents. SoftBank has committed $3 billion to these products, as OpenAI looks to offset its $5 billion loss from last year. Read more
💰 Anthropic secures $3.5B at $61.5B valuation. The funding round, led by Lightspeed Venture Partners, will fuel AI research, compute expansion, and international growth. Anthropic continues advancing Claude models, with 3.7 Sonnet excelling in coding and integrations driving enterprise adoption. Read more
🧠 Ilya Sutskever pursues new AI path. Safe Superintelligence (SSI), co-founded by former OpenAI chief scientist Ilya Sutskever, is researching an alternative method for developing AI beyond current scaling approaches. Sutskever claims to be climbing a "different mountain," focusing solely on artificial superintelligence (ASI) rather than incremental AI advancements. The secretive startup operates with a small team across Silicon Valley and Tel Aviv, avoiding public disclosures about its personnel. Read more
🇺🇸 TSMC expands U.S. chip production. Taiwan Semiconductor Manufacturing Company will invest $100 billion over four years to expand its Arizona operations, increasing its total U.S. commitment to $165 billion. The company will build six factories, create 25,000 jobs, and manufacture AI and smartphone chips domestically. This move follows tariff threats and pressure from the CHIPS Act, which aims to strengthen U.S. semiconductor manufacturing. Read more
🖥️ Meta tests in-house AI chip. Meta has begun testing its first proprietary AI training chip, aiming to reduce dependence on Nvidia and lower infrastructure costs. The chip, manufactured by TSMC, is part of Meta's long-term push to power recommendation systems and generative AI. If successful, it could shift Meta's AI strategy by 2026. Read more
🤖 Amazon builds a reasoning AI model. Amazon is developing a new AI model under the Nova brand, focusing on "hybrid reasoning" to balance quick responses with complex problem-solving. The company aims to surpass competitors like OpenAI and Anthropic by optimizing cost efficiency and securing a top ranking in external benchmarks. Read more
