Good morning, AI enthusiast.

The push for more capable AI agents just received a major boost from Chinese startup DeepSeek. The company released a new open-source model with a unique hybrid inference system designed to tackle complex, multi-step problems.

This new system allows the model to switch between quick responses and deeper reasoning, but its compatibility with upcoming home-grown chips points to a larger strategic play. Is this the beginning of a powerful AI ecosystem that's less dependent on today's leading chipmakers?

In today’s AI recap:

• DeepSeek’s new agent-focused AI model

• Google’s AI energy consumption report

• China blocks key Nvidia chip purchases

• Meta pauses its AI talent hiring

• 8 trending AI Tools

DeepSeek's New Agentic AI

The Recap: Chinese AI startup DeepSeek just dropped DeepSeek-V3.1, a powerful open-source model built for the agent era. Its new hybrid inference system lets it switch between quick responses and deep "thinking" for complex tasks.

Unpacked:

• The model’s hybrid system operates in two modes: a “Non-Think” mode for standard, quick answers and a “Think” mode designed for complex, multi-step reasoning.

• Post-training optimization boosts the model's performance on agent benchmarks, showing major gains in tool use and multi-step coding tasks.

• DeepSeek also revealed the model was specifically designed for home-grown AI chips expected to be released soon, using an efficient data format to reduce hardware needs.

Bottom line: This release signals a clear focus on developing more autonomous AI agents capable of tackling complex, multi-step problems. Its compatibility with upcoming domestic hardware also points to a future where powerful models are less dependent on today's leading chipmakers.

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

AI Tools of the Day

1. 🤖 Tasker AI – Unleash intelligent AI agents to automate your daily workflows, from scheduling meetings to conducting complex research across your connected apps.

2. 🧩 Corpus – Integrate all your work apps like Salesforce, Notion, and Slack into a single intelligent command center that delivers real-time, actionable briefings.

3. 🎬 IdeaVido – Go beyond simple generation and modify any video using just text prompts to seamlessly replace backgrounds, swap objects, or change entire scenes.

4. 🧑‍💻 Marblism – Hire a team of specialized digital AI employees to handle your emails, manage social media, and provide customer service 24/7.

5. 🪄 Same – Instantly create a pixel-perfect, editable copy of any website's user interface just by entering its URL.

6. 🎯 Scop.ai – Craft expert-level system prompts for models like GPT and Claude using smart templates to get dramatically better and more consistent AI results.

7. 🔧 PixieBrix – Customize any website or web app with your own low-code modifications and AI-powered automations directly within your browser.

8. 📞 BabelPhone – Translate your phone conversations in real-time, allowing you to speak naturally with international contacts without any language barriers.

Explore the Best AI Tools Directory to find tools that will 10x your output 📈

Google Reveals AI's Energy Bill

The Recap: In a first for the industry, Google released a detailed paper quantifying its AI's environmental footprint, revealing that a median Gemini text prompt uses just 0.24 watt-hours of energy and five drops of water.

Unpacked:

• Google's calculation is comprehensive, factoring in not just the AI chips but also idle machines, CPU/memory usage, and data center overhead like cooling.

• The company reports that efficiency gains have been dramatic, with the energy used per prompt dropping by a factor of 33x over the last 12 months.

• For context, a median Gemini prompt's 0.24 Wh usage is lower than the 0.34 Wh figure previously mentioned for an average ChatGPT query, though methodologies differ.

Bottom line: Google is setting a new benchmark for transparency in AI's environmental impact, pressuring other major players to disclose their own data. This move could shift the industry's focus toward provable efficiency as a key competitive advantage, not just model performance.

AI Training

How to Build a WhatsApp Chatbot + AI Agent in n8n for Small Businesses (free template)

The Recap: In this video, I'm going to show you how to build a WhatsApp chatbot and AI agent for businesses that interact with their customers over WhatsApp. This system can scrape all details from a business or company's website, bundle that together into an LLM-ready knowledge base, and use that knowledge base to answer real customer inquiries. I'm also going to show you how to set up the WhatsApp n8n connection from scratch.

P.S. We also launched a free AI Automation Community for those looking to build and sell AI Automations — Come join us!

China's Nvidia Chip Blockade

The Recap: Beijing is pressuring its top tech companies to halt purchases of Nvidia’s H20 AI chips, according to a Financial Times report. The move comes after Chinese officials reportedly found comments from U.S. Commerce Secretary Howard Lutnick to be "insulting."

Unpacked:

• The backlash stems from Lutnick's remarks that the U.S. doesn't sell China its "best stuff" and that the goal is to get its developers "addicted to the American technology stack."

• The H20 chip was specifically designed by Nvidia to comply with existing U.S. export controls, offering a less powerful alternative to its top-tier AI hardware.

• China’s pressure campaign is a coordinated effort involving multiple agencies, including the Cyberspace Administration of China and the Ministry of Industry and Information Technology.

Bottom line: This standoff shows how sensitive the U.S.-China tech relationship is, where public comments can have major commercial fallout. The pressure could accelerate China’s drive for chip self-sufficiency, threatening long-term U.S. market share.

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).

Meta Hits Pause on AI Talent Spree

The Recap: After a massive talent hunt with offers reportedly worth up to $100M, Meta has abruptly frozen hiring for its AI division. The move comes amid growing fears of an AI bubble and a major internal restructuring of its superintelligence efforts.

Unpacked:

• The freeze marks a sharp reversal from an aggressive recruitment drive that recently brought on over 50 researchers and engineers from rivals like Google and OpenAI.

• This decision aligns with recent market turbulence, including a tech stock sell-off fueled by concerns that massive AI investments are not yet delivering clear returns.

• The pause coincides with a significant internal restructuring that divides Meta's AI unit into four new groups and follows the underwhelming release of its latest models.

Bottom line: This move could signal a broader shift in the AI talent wars, moving from a "growth-at-all-costs" mindset to a more measured focus on results. Companies are now facing increased pressure to demonstrate the tangible value of their multi-billion dollar AI investments.

The Shortlist

Microsoft AI CEO Mustafa Suleyman warned of a rise in "AI psychosis," a phenomenon where users are developing unhealthy attachments and delusions about chatbots they perceive as conscious.

xAI exposed hundreds of thousands of private Grok user conversations to public search engines after its "share" feature made the chats indexable by default without explicit warning.

Google expanded its AI Mode in Search to over 180 countries and introduced new agentic capabilities for subscribers, starting with finding restaurant reservations.

Anthropic launched a higher education advisory board and a series of free AI Fluency courses to help guide universities on the responsible integration of AI in the classroom.

Fastly reported that AI crawlers are putting a heavy load on websites, with Meta alone accounting for over half of all crawler traffic and some fetcher bots hitting sites with over 39,000 requests per minute.

Signing off,

David, Lucas, Mitchell — The Recap editorial team