DeepLearning.AI

A new short course, Claude Code: A Highly Agentic Coding Assistant, is live! Claude Code is currently one of the most capable coding assistants. It can explore your codebase, plan features, write tests, refactor code, and even collaborate across multiple sessions—with surprisingly minimal input. In this course, you’ll learn how to guide Claude Code effectively: from setting up context and memory to integrating with GitHub and MCP servers. You’ll use it to extend a RAG chatbot, refactor a Jupyter notebook for e-commerce data analysis, build a web app from a Figma design, and more. Taught by Elie Schoppik and built in collaboration with Anthropic, this course is a must for AI builders. 👉 Enroll now: https://hubs.la/Q03Bxpg20

Researchers at Stanford and Carnegie Mellon analyzed over 1,000 Character AI users and 400,000 messages to gauge how AI companionship affects mental health. The study showed heavier reliance on bots for friendship or romance correlated with lower satisfaction and higher loneliness. Read our summary of the paper in The Batch: https://hubs.la/Q03BT7yv0

This week, in The Batch, Andrew Ng discusses why Meta is paying AI engineers top dollar. Plus: 🔓 OpenAI's re-opening 🏭 Reasoning's impact on carbon emissions 🤖 GLM-4.5, a new open contender 🩺 Autonomous surgical robot removes gallbladders Read The Batch: https://hubs.la/Q03BTcdj0

We’re hiring a Product Marketing Manager! This role is for marketers who like to build and builders who know how to bring products to market. At DLAI we create world-class AI education that empowers millions to build real-world applications and shape the future of the field. You’ll help bring that vision to life. You’ll work with our marketing, product, and content teams (as well as directly with Andrew Ng) to craft product narratives, drive go-to-market strategies, and speak to our learners. 📍San Francisco Bay Area | Full-time 🔗 Apply now: https://hubs.la/Q03Bz9PG0

The White House reversed its April ban and will let Nvidia sell H20 processors and AMD sell MI308 chips to Chinese customers. Nvidia CEO Jensen Huang lobbied President Trump by noting the parts lag U.S. flagship GPUs, pledging 500 billion dollars for on-shore fabrication, and highlighting that China made 13 percent of Nvidia’s revenue and 24 percent of AMD’s last year. Learn more in The Batch: https://hubs.la/Q03Bt0YY0

Production-ready RAG systems need observability. From tracking latency and throughput to evaluating response quality with human feedback or LLM-as-a-judge, robust observability gives you visibility into both system performance and output quality, on both a component and system-wide level. This lesson from our Retrieval Augmented Generation course breaks down the core components of an effective eval system and how to balance cost, automation, and accuracy when choosing your metrics. 📚 Learn more in the full course: https://hubs.la/Q03Bs-3b0

Alibaba released Qwen3-235B-A22B-Instruct-2507, a reasoning-enabled Thinking-2507 version, and a 480-billion-parameter Qwen3-Coder, all under the Apache 2.0 license. The new Instruct model tops other non-reasoning peers on 14 of 25 benchmarks, Thinking ranks mid-pack among proprietary reasoning systems, and Coder handles million-token contexts and edges Claude Sonnet 4 in multi-turn tool-use tests while costing as little as seventy cents per million input tokens. Learn more in The Batch: https://hubs.la/Q03BdzvR0

The future of AI-assisted coding is happening now. At Buildathon, hosted with AI Fund, we’re inviting developers to show just how fast they can build real products using today’s tools. 🖥️ If you’re ready to prototype fast and think even faster, apply now: https://hubs.la/Q03BdsC20 See you on August 16 in Menlo Park!

Researchers at Google built AlphaEvolve, an agent that lets Gemini 2.0 Flash and Pro repeatedly run, assess, and edit code until unit tests improve. Starting from placeholder functions, the loop produced a new routine for complex 4x4 matrix multiplication, matched or topped solutions on most of 50 math tasks, and cut Gemini training time 1 percent. The work shows that coupling large language models with automated tests can uncover new algorithms that produce real infrastructure gains better than one‑shot prompting or manual tuning. Read our summary of the paper in The Batch: https://hubs.la/Q03zNNfr0

The European Union published a “General Purpose AI Code of Practice,” explaining voluntary steps developers can take to meet the AI Act’s requirements for general‑use models. The code directs builders of models deemed to pose “systemic risks” to document data sources, log compute and energy use, and report security or safety incidents within two to ten days. Microsoft, Mistral, and all OpenAI opted in, while Meta declined. Read our full breakdown in The Batch to see what the guidelines mean for developers’ costs and timelines: https://hubs.la/Q03zNqSp0