PLUS: The Blackwell era begins, cracking hashes with hardware, and what Ford's CEO says about AI jobs

Note to self: don't make changes to the automation without testing it. Especially before a wedding...

Happy 4th of July AI Rockstars! Wishing every one of you an amazing day to celebrate our nation's 250th birthday!

A new open-source project provides a clear blueprint for building your own custom AI research agent. It shows how to combine Vercel's AI SDK and specialized search APIs to automate data gathering from platforms like YouTube and Reddit.

This showcases the growing trend of assembling specialized AI agents from powerful, accessible building blocks. As these components become easier to use, are we about to see an explosion of niche agents designed for every possible workflow?

In today's Lean AI Report:

• A blueprint for building your own AI research agent
• CoreWeave becomes the first to deploy NVIDIA's Blackwell hardware
• Using specialized hardware to accelerate password cracking
• Ford CEO's warning on the future of white-collar jobs

Build Your Own AI Researcher

The Report: A new open-source project called "yurei-app" provides a blueprint for building a custom social media research agent. It demonstrates how to combine Vercel's AI SDK with specialized search APIs to create powerful, automated tools.

Broaden your horizons:

• The agent works by searching across different platforms, including YouTube and Reddit, using APIs to gather information on a given topic.
• It leverages Exa's search API, which is designed to find high-quality content by understanding a user's intent rather than just matching keywords.
• The entire framework is built using Vercel's AI SDK, making it easier for developers to create conversational AI experiences and stream responses back to the user.

If you remember one thing: This project showcases the growing trend of assembling specialized AI agents from powerful, accessible components. It offers a clear, practical template for anyone looking to automate their own research workflows.

The Blackwell Era Begins

The Report: AI cloud provider CoreWeave has become the first to deploy NVIDIA's next-generation GB300 NVL72 systems. The move unlocks a massive leap in performance for AI training and inference workloads.

Broaden your horizons:

• The new systems deliver a 50x increase in output for reasoning models and a 5x improvement in throughput per watt compared to the previous Hopper architecture.
• This early access to top-tier hardware helps CoreWeave differentiate itself from larger cloud competitors like AWS, Google, and Microsoft.
• The deployment was built through key hardware partnerships with companies like Dell, which assembled and tested the liquid-cooled systems in the U.S.

If you remember one thing: Early access to the most powerful hardware is a critical advantage for specialized AI cloud providers. This deployment solidifies CoreWeave’s position as a key enabler for companies racing to build the next generation of AI.

Cracking Hashes with Hardware

The Report: A developer showcases a new project that uses specialized hardware to build a high-performance password cracker. The system parallelizes SHA-256 hash calculations to rapidly find the original string for a given hash.

Broaden your horizons:

• The setup uses a Litefury FPGA board connected to a Raspberry Pi 5, demonstrating how powerful custom hardware can be paired with accessible single-board computers.
• To maximize performance, the design integrates 12 parallel SHA-256 cores running on the FPGA, enabling it to test numerous password candidates simultaneously.
• The developer created a Python driver to manage the hardware and has shared all project files on GitHub for others to explore and build upon.

If you remember one thing: This project is a powerful demonstration of how FPGAs can vastly accelerate specific computational tasks like cryptography. It highlights the growing importance of specialized hardware for tackling complex problems that are inefficient for general-purpose processors.

CEOs Warn of AI Job Cuts

The Report: Ford CEO Jim Farley has joined a growing chorus of corporate leaders openly predicting that artificial intelligence will significantly reduce the number of white-collar jobs.

Broaden your horizons:

• At the Aspen Ideas Festival, Farley made the stark prediction that AI will eventually replace “half of all” white-collar workers in the U.S., urging a greater focus on skilled trades.
• This sentiment is not isolated; executives at companies like Anthropic and Amazon have issued similar warnings, signaling a broader shift in how corporate leadership views AI's role in the workforce.
• These predictions are supported by hiring data, as a recent SignalFire report noted that Big Tech's hiring of new graduates has fell around 50% from pre-pandemic levels, partly due to AI's influence.

If you remember one thing: The conversation around AI and jobs is moving from abstract theory to concrete business strategy. This signals a critical moment for professionals to focus on developing skills that complement and manage AI systems, rather than compete with them.

The Shortlist

Meta seeks to raise a staggering $29B from private capital firms to fund its massive AI data center build-out, signaling the enormous infrastructure costs of competing in the AI race.

Reddit races to protect its forums from AI-generated content, aiming to preserve the value of its human-generated conversations, which it licenses to AI companies.

Germany asked Apple and Google to remove the Chinese AI app DeepSeek from its app stores, citing concerns that the company illegally transfers user data to China.

Lyft integrated Anthropic's Claude into its customer care platform, reducing resolution times by 87% and showcasing how frontier models are being deployed to enhance core business operations.