Anthropic, Google offer free AI curriculum

The thing going viral is not one giant shared program. It’s a stack of official learning hubs from the companies building the tools. Anthropic now has a fuller Academy presence, Google still runs its long-standing machine-learning courses, NVIDIA offers free self-paced Deep Learning Institute classes, and OpenAI has expanded Academy and developer training. Put together, that does start to look like a serious free curriculum. (anthropic.skilljar.com) ### What actually changed? The freshest piece is Anthropic. Its Academy now points people to structured courses on Claude, prompt engineering, MCP, agents, Claude Code, and API work, instead of just scattered docs and blog posts. That matters because a curriculum feels different from documentation — it tells learners what order to do things in. (anthropic.skilljar.com)ly, yes — but with a catch. The learning materials themselves are free across these platforms, and some providers offer certificates on at least part of the catalog. NVIDIA says many popular self-paced courses are free and notes that select courses include a certificate of competency. Anthropic’s Academy pages are open to sign-up-based learners, a(anthropic.skilljar.com) and recorded sessions. (nvidia.com) ### What do you learn from each company? Google is still strongest on classic ML foundations — things like supervised learning, problem framing, and core concepts through Machine Learning Crash Course and related pathways. NVIDIA is strongest on the compute layer — GPUs, accelerated computing, and applied generative-AI workflows. Anthropic leans into Claude-specif(nvidia.com)AI splits between user fluency and developer practice — prompting, workflows, API guides, and build hours. (developers.google.com) ### So is the “$200k engineer” line real? Not literally. No one is handing out the equivalent of a four-year computer science degree for free. But the phrase is getting at something real: a lot of the job-ready AI stack has moved into public vendor education. If you want to become useful fast, the bottleneck is less “Where do I find material?” and more “Can I turn tutorials into projects?” That’s why these lists spread so quickly. (developers.google.com) ### What skills does this bundle cover? Basically, the baseline toolkit for 2026 entry-level AI work. You can learn model basics, prompting, evaluation, retrieval, API integration, agent workflows, and some deployment patterns. That is enough to build demos, internal automations, and small production features. It is not enough by itself to make someone a research scientist or senior ML engineer. (anthropic.com) ### What’s missing from the viral version? Math depth, data engineering, and long-form software practice. Vendor courses are great at teaching their tools and patterns. They are weaker as a substitute for statistics, distributed systems, debugging ugly production failures, or learning how to work with messy proprietary data. Think of this as a very good on-ramp, not the whole highway. (developers.google.com) ### Why are companies giving this away? Because trained users buy platforms. The easier Anthropic, Google, NVIDIA, and OpenAI make it to get productive, the more likely developers are to build on Claude, Gemini-era Google tooling, NVIDIA hardware ecosystems, or OpenAI APIs. Free education here is marketing, ecosystem building, and support reduction all at once. That’s not cynical — it’s just the business model. (anthropic.com) ### Bottom line The viral post is overselling one part and underselling another. It’s overselling the idea that there is a single magic curriculum. But it’s underselling how much official, high-quality AI training is now free and organized in public. For anyone trying to break in, that’s the real shift. (anthropic.skilljar.com)