Good morning, AI enthusiast.

A significant forecast regarding AI's future impact on employment has come from Anthropic's CEO, Dario Amodei. He predicts a potential 50% reduction in entry-level white-collar positions within the next five years, signaling a rapid transformation in the job market due to artificial intelligence.

This projection arrives amidst data already showing a downturn in tech hiring for new graduates. With such a substantial shift anticipated, how will industries and individuals adapt to these evolving employment realities?

In today’s AI recap:

• Anthropic CEO’s warning on AI and entry-level jobs

• Our first AI Training video! (watch here)

• The New York Times and Amazon's new AI content partnership

• Meta AI achieves one billion monthly active users

• DeepSeek’s latest R1 reasoning model update

AI Job Disruption: Anthropic CEO Warns of 50% Entry-Level Cuts

The Recap: Anthropic CEO Dario Amodei has issued a stark warning that AI could eliminate 50% of entry-level white-collar jobs within five years, as he detailed in an Axios interview. This projection highlights the accelerating pace of AI's impact on the workforce.

Unpacked:

• Recent data underscores this trend, with a SignalFire report revealing Big Tech's hiring of new graduates has already fallen by roughly 50% from pre-pandemic levels, partly due to AI adoption.

• Experts outside of Anthropic echo these concerns; University of Utah AI professor Tim Kapp agrees with Amodei's assessment, noting a 27.5% drop in entry-level jobs in the developing world, and that approximately 40% of CEOs are considering workforce reductions.

• Amodei states that AI companies and governments are "sugarcoating" the risks, emphasizing an obligation for transparency, a principle Anthropic supports by sharing insights from its AI safety research.

Bottom line: This significant forecast from a leading AI figure emphasizes the urgent need for professionals to understand AI's capabilities and potential job market shifts. Proactive adaptation and skill development will be crucial for navigating this evolving employment landscape.

AI Training

How to Automate Your Web Scraping in LLM-Ready Format (n8n + Firecrawl + rss.app)

The Recap: We just posted our very first (of many!) AI automation training videos on YouTube. In this video, we walk through how to use n8n, firecrawl, and rss.app to scrape virtually any piece of web content and transform it into LLM-ready output. Make sure to watch to the end to see the exact data pipeline we've used to generate SEO-friendly blog posts, automated newsletters, and all sorts of content here at The Recap.

Have a business process you'd like us to automate? Let us know in the comments what you'd like to automate next!

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!

NYT & Amazon: A New Chapter for AI and News

The Recap: The New York Times and Amazon announced a landmark agreement, allowing Amazon to license NYT's extensive content to train its AI models and enhance services like Alexa. This multi-year collaboration highlights a growing trend of media and tech joining forces over AI development.

Unpacked:

• Amazon gains access to decades of archival content from The New York Times, including materials from NYT Cooking and The Athletic, to refine its AI models.

• The partnership enables Amazon's AI, like Alexa, to deliver real-time summaries and excerpts, ensuring Times content is properly attributed with direct links back to the source.

• NYT's leadership views this as affirming that quality journalism is valuable, securing fair compensation while its intellectual property powers AI advancements.

Bottom line: This alliance showcases a key strategy for legacy media to engage with the AI boom, turning trusted archives into new revenue streams. For AI developers and users, it points towards AI tools enriched with more authoritative and diverse information sources.

Meta AI Hits 1 Billion User Milestone

The Recap: Meta AI now boasts one billion monthly active users across its apps, a milestone Mark Zuckerberg announced at the company's annual shareholder meeting. This rapid user adoption signals Meta's accelerating push into the personal AI space.

Unpacked:

• Meta AI's user base impressively doubled from nearly 500 million monthly active users reported back in September 2024.

• Zuckerberg outlined Meta's current focus: deepening the AI experience with an emphasis on personalization, voice conversations, and entertainment.

• This growth follows the recent launch of a stand-alone Meta AI app, as Meta explores future monetization through paid recommendations or subscriptions.

Bottom line: This surge in users highlights the growing integration of AI assistants into everyday digital interactions. Meta's expanding AI footprint positions it as a formidable player in the evolving personal AI landscape, paving the way for new user experiences and business models.

DeepSeek Boosts AI Reasoning with R1 Update

The Recap: Chinese AI firm DeepSeek has rolled out its R1-0528 reasoning model, detailed in an official update, promising enhanced thinking and inference capabilities. This move intensifies competition in the global AI landscape.

Unpacked:

• The updated R1 model ranks competitively on code generation benchmarks like LiveCodeBench, placing it near OpenAI’s o4 mini and o3 models and ahead of some xAI and Alibaba offerings.

• Alongside the main R1, DeepSeek introduced DeepSeek-R1-0528-Qwen3-8B, a distilled version built on Alibaba's Qwen3-8B that can run on a single GPU, showcasing impressive efficiency with its base model's known hardware needs.

• This smaller model is available under a permissive MIT license, meaning developers can use it commercially without restriction, with hosts like LM Studio already providing access.

Bottom line: DeepSeek's continuous advancements underscore China's rapid progress in cutting-edge AI development. These new models, especially the efficient distilled version, offer powerful and more accessible tools for professionals seeking to automate tasks and boost productivity.

The Shortlist

Anthropic open-sourced its circuit tracing tools, developed with Decode Research, allowing researchers to generate attribution graphs and better understand the internal workings of models like Gemma-2-2b and Llama-3.2-1b.

Meta partnered with Palmer Luckey's Anduril to develop next-generation extended reality (XR) technology for the U.S. military, leveraging Anduril's Lattice AI platform and Meta's Llama AI models.

Grammarly secured $1 billion in new funding from General Catalyst, aiming to scale sales, marketing, and pursue strategic acquisitions to expand its AI-powered productivity platform.

Testing revealed DeepSeek's updated R1-0528 AI model is substantially more censored regarding topics sensitive to the Chinese government, making it their most restricted model yet for such criticisms.

Signing off,

David, Lucas, Mitchell — The Recap editorial team