OpenAI and Broadcom Team Up

Hey there 👋

We hope you're excited to discover what's new and trending in AI, ML, and data science this week.

Here is your 5-minute pulse...

But first, a quick message from this week’s partner 👇

The World’s Most Wearable AI

Limitless is your new superpower - an AI-powered pendant that captures and remembers every conversation, insight, and idea you encounter throughout your day.

Built for tech leaders who need clarity without the clutter, Limitless automatically transcribes and summarizes meetings, identifies speakers, and delivers actionable notes right to your fingertips. It’s securely encrypted, incredibly intuitive, and endlessly efficient.

Order now and reclaim your mental bandwidth today.

Discover Limitless

print("News & Trends")

Thinking Machines Lab Co-Founder Joins Meta (3 min. read)

In a strategic move, the co-founder of Thinking Machines Lab has joined Meta to bolster its AI research initiatives. This transition underscores Meta's commitment to advancing artificial intelligence by integrating top-tier talent from leading AI research organizations. The co-founder's expertise is expected to significantly enhance Meta's capabilities in machine learning and data science, potentially accelerating the development of innovative AI-driven solutions.

Salesforce Unveils Agentforce 360 Amid Intensifying Enterprise AI Competition (3 min. read)

Salesforce has introduced Agentforce 360, the latest iteration of its AI agent platform, aiming to enhance enterprise AI capabilities. This update features Agent Script, a tool for flexible AI agent programming, and Agentforce Builder, a unified environment for building, testing, and deploying AI agents. Additionally, deeper integration with Slack is planned, transforming it into a central hub for AI-driven enterprise applications. These advancements position Salesforce to compete more effectively in the rapidly evolving enterprise AI landscape.

OpenAI and Broadcom Announce Strategic Collaboration to Deploy 10 Gigawatts of OpenAI-Designed AI Accelerators (3 min. read)

OpenAI and Broadcom are teaming up to develop and deploy 10 gigawatts of custom AI accelerators, integrating OpenAI's chip designs with Broadcom's Ethernet solutions. This multi-year partnership aims to enhance AI infrastructure, with deployments starting in the latter half of 2026 and completing by 2029. By embedding insights from frontier model development directly into hardware, OpenAI seeks to unlock new levels of AI capability and intelligence.

Microsoft Introduces MAI-Image-1, Debuting in the Top 10 on LMArena (3 min. read)

Microsoft AI has unveiled MAI-Image-1, its inaugural in-house image generation model, now ranking among the top 10 on LMArena. Designed to deliver photorealistic images with remarkable speed and quality, MAI-Image-1 emphasizes visual diversity and practical value, steering clear of repetitive or generic outputs. This advancement promises to enhance creative workflows, enabling faster ideation and iteration. The model will soon be integrated into Copilot and Bing Image Creator, offering users a more immersive and dynamic experience.

print("Applications & Insights")

Not Another Workflow Builder (4 min. read)
LangChain's Harrison Chase critiques visual workflow builders, arguing they often fail to simplify agent creation for non-technical users and become unwieldy with complex tasks. He suggests that for simpler applications, no-code agents combining prompts and tools are more effective, while intricate workflows are best managed through code, leveraging tools like LangGraph. Chase emphasizes focusing on developing reliable, user-friendly agents and enhancing code generation models to better serve diverse complexity levels.

LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings (White paper)
This study introduces Semantic Similarity Rating (SSR), a method that leverages large language models (LLMs) to simulate consumer survey responses. By mapping LLM-generated textual outputs to Likert scales through embedding similarities, SSR achieves 90% of human test-retest reliability on a dataset of 9,300 responses. Notably, it maintains realistic response distributions and provides qualitative insights, offering a scalable and interpretable alternative for consumer research.

Is AI a Bubble? (5 min. read)
Azeem Azhar and Nathan Warren tackle the pressing question: Is AI experiencing a bubble? They introduce a practical framework with five key indicators to assess AI's current trajectory against historical bubbles. Drawing from past tech booms and busts, they provide a nuanced analysis to help investors and professionals navigate the AI landscape. This piece is essential reading for those seeking to understand the sustainability of AI's rapid growth.

Build Your Own RAG Application (GitHub repo)
This GitHub repository offers a comprehensive guide to constructing Retrieval-Augmented Generation (RAG) applications using LangChain. It provides step-by-step instructions, from setting up the environment with Python 3.11.11 to running detailed Jupyter notebooks that cover various aspects of RAG, including multi-query handling, routing, indexing, and advanced retrieval techniques. The project also includes guidance on configuring essential environment variables for integrating with services like OpenAI, LangSmith, Pinecone, and Cohere, making it a valuable resource for developers aiming to enhance their applications with RAG capabilities.

print("Tools & Resources")

TRENDING MODELS

Text Generation
MoonshotAI/Kimi-K2
⇧ 1.64k Downloads
Kimi K2 is a trillion-parameter model developed by Moonshot AI, utilizing a mixture-of-experts architecture with 32 billion active parameters per inference. It excels in coding and math benchmarks, demonstrating efficiency and high performance in complex tasks.

Text-to-Speech
facebook/mms-tts
⇧ 18.5k Downloads
The MMS-TTS model by Facebook is a 0.7 billion parameter text-to-speech system designed to generate natural-sounding speech across multiple languages. It offers high-quality audio synthesis, making it suitable for various applications requiring multilingual speech generation.

Text Generation
EleutherAI/gpt-neox-20b
⇧ 32.4k Downloads
GPT-NeoX-20B is a 20 billion parameter autoregressive language model developed by EleutherAI, capable of generating coherent and contextually relevant text. It serves as a versatile tool for tasks such as text completion, summarization, and creative writing.

Text Generation
bigscience/bloom
⇧ 1.58k Downloads
BLOOM is an open-access multilingual language model with 176 billion parameters, developed by the BigScience initiative. It supports text generation in 46 languages and 13 programming languages, promoting inclusivity and diversity in AI applications.

Text-to-Image
stabilityai/stable-diffusion-2
⇧ 226 Downloads
Stable Diffusion 2 by Stability AI is an advanced text-to-image model capable of generating high-resolution images from textual descriptions. It offers improved image quality and prompt adherence, making it a powerful tool for creative and design tasks.

TRENDING AI TOOLS

• 🌐 Gemini Enterprise: Accelerate AI with scalable, secure, and comprehensive machine learning solutions.

• 🔗 Jellyjoin: Efficiently join large datasets with minimal memory usage.

• 🛠️ Super Intern: Your AI teammate inside group chats.

• 📝 markitdown: A simple markdown editor for creating and editing markdown documents.

print("Everything else")

• Google plans to invest $10 billion in an AI data center project in India.

• Andrej Karpathy releases nanochat, a minimal full-stack ChatGPT implementation.

• Sora hit 1M downloads faster than ChatGPT.

• Cursor shares a video on 10 pro tips for working with Cursor Agents.

That’s it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve the pulse experience for you.

See you soon,

Andres