Good morning, AI enthusiast.

Amazon is pouring $20 billion into new data centers in Pennsylvania to fuel its AI ambitions. A key part of the plan involves directly connecting a facility to a nuclear power plant for a massive, carbon-free energy supply.

This aggressive push into direct energy sourcing highlights how crucial massive power infrastructure has become in the AI race. But with regulators already putting the nuclear connection on hold, will Amazon's bold energy strategy become a blueprint for the industry or a cautionary tale of overreach?

In today's AI recap:

• Amazon's $20B nuclear-powered AI bet

• Apple's research on AI 'accuracy collapse'

• New AI training video (watch here)

• China's nationwide AI chatbot shutdown

Amazon's Nuclear-Powered AI

The Recap: Amazon plans to invest at least $20 billion in Pennsylvania to build new AI and cloud computing data centers. In a novel approach to meet AI's colossal energy demands, one of the first facilities will be directly powered by an adjacent nuclear power plant.

Unpacked:

• The move directly tackles AI's energy consumption problem by securing a massive amount of carbon-free nuclear power. Amazon is partnering with Talen Energy to directly connect a data center to its 2.5 GW Susquehanna Steam Electric Station, ensuring a stable, non-grid-reliant power source.

• This project represents the largest private sector investment in Pennsylvania's history, creating at least 1,250 direct high-tech jobs and thousands more in construction. The deal highlights how the demand for AI infrastructure is driving significant economic development and a re-industrialization of local economies.

• The strategy underscores Big Tech's commitment to find alternative energy solutions and bypass congested power grids. This follows a growing trend, with Microsoft also striking a deal to power its data centers with Pennsylvania's revived Three Mile Island nuclear plant.

Bottom line: This investment shows how the race for AI dominance is physically reshaping our energy and industrial landscapes. Co-locating data centers directly with power plants could become the new standard for sustainably scaling the technology.

Apple's AI Reality Check

The Recap: A new Apple research paper reveals that even advanced 'reasoning' AI models suffer a 'complete accuracy collapse' when faced with complex problems, suggesting current approaches may be hitting a fundamental wall.

Unpacked:

• Researchers tested models including OpenAI's o3 and Google's Gemini Thinking on puzzles, finding they all failed completely once problems reached a certain complexity threshold.

• Counterintuitively, the study found that as problems grew harder, models actually reduced their reasoning effort just before failing, indicating a core scaling limitation.

• The "pretty devastating" findings challenge the industry's race toward AGI, suggesting current methods rely more on pattern matching than true generalizable reasoning.

Bottom line: This research injects a dose of realism into the AI hype cycle, signaling that the path to human-level AI is not a straight line of simply scaling up current models. Future breakthroughs will likely require entirely new approaches beyond what the industry is currently pursuing.

AI Training

We Used AI to Clone Zapier & Canva’s SEO Strategy (free n8n template)

The Recap: In this video, I walk through how I used n8n to build a fully automated system for generating high ranking SEO listicle style articles that can be scaled programmatically for your website — just like how Canva and Zapier do things. Today, these blog posts are generating thousands of page views per month for my website.

P.S We also launched a free community for AI Builders looking to master the art and science of building AI Automations — Come join us!

China's National AI Shutdown

The Recap: In an unprecedented move to ensure academic integrity, major Chinese tech firms like Alibaba and Tencent temporarily suspended their AI chatbots for 13.3 million students during the nation's critical 'gaokao' college entrance exams.

Unpacked:

• The coordinated freeze saw companies like Alibaba, Tencent, and ByteDance disable or limit features, particularly photo recognition, to prevent students from using AI to solve exam questions during the high-stakes, multi-day tests.

• Chatbots didn't just go silent; they explicitly informed users that services were paused to ensure the fairness of the examinations, demonstrating a clear, top-down directive to prevent AI-assisted cheating.

• In an ironic twist, while student-facing AI was disabled, authorities deployed AI surveillance systems in testing centers to monitor for irregular behavior, using AI to enforce the rules that other AIs were blocked from breaking.

Bottom line: This event marks one of the largest real-world stress tests of AI governance and societal control to date. It offers a fascinating look at how a country might enforce digital guardrails when the integrity of a national institution is at stake.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

• Real peer support from people building actual AI businesses

• Complete access to download our automation library of battle-tested n8n templates

• Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

The Shortlist

Anthropic's CEO warned in an essay that the industry does not fully understand how its own AI creations work, calling the lack of "interpretability" an unprecedented and concerning risk in the history of technology.

Duolingo's CEO clarified the company's "AI-first" strategy after intense customer backlash, stating the goal is to have employees use AI to handle routine work so they can focus on more creative tasks, not to replace its workforce.

Australian researchers developed an AI model that identifies contaminated construction and demolition wood waste with 91.67% accuracy, enabling automated sorting to improve recycling efforts and promote a circular economy.

Signing off,

David, Lucas, Mitchell — The Recap editorial team