AWS and OpenAI Announce Multi-Year Strategic Partnership ($38B Commitment)

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...

print("News & Trends")

AWS and OpenAI Announce Multi-Year Strategic Partnership (3 min. read)

OpenAI and AWS have inked a $38 billion, multi-year deal to supercharge AI workloads using AWS's cutting-edge infrastructure. This partnership grants OpenAI immediate access to hundreds of thousands of NVIDIA GPUs and the scalability to tens of millions of CPUs, all via Amazon EC2 UltraServers. The collaboration aims to rapidly expand OpenAI's compute capacity, enhancing the performance and reach of AI applications like ChatGPT.

Wharton AI Study Shows Surging Enterprise Adoption (8 min. read)

The 2025 Wharton AI Adoption Report reveals a significant leap in generative AI integration, with 82% of enterprise leaders now using it weekly, up from 72% in 2024. Notably, 72% are actively measuring AI's ROI, focusing on productivity and profit gains, and 75% report positive returns. However, 43% express concerns about skill degradation, emphasizing the need for robust training programs to sustain AI-driven advantages.

Former xAI Researcher Targets $1B for Human-First AI Lab (4 min. read)

Former xAI researcher Eric Zelikman and an early-employee of Google are reportedly in talks to raise $1 billion for their new startup called Humans&, aiming for a $5 billion valuation. The venture is planning to build AI models that better collaborate with humans, focusing on training systems to remember and respond to a person’s preferences and interests, rather than purely replacing human tasks. The move signals investor enthusiasm about the next stage of human-centric AI foundations.

Jupyter AI: AI Coding in Notebooks (Course)

DeepLearning.AI's latest course, led by Andrew Ng and Brian Granger, introduces Jupyter AI—a tool that seamlessly integrates AI coding assistance into Jupyter notebooks. You'll learn to generate, debug, and explain code directly within your notebook environment. The course guides you through building practical applications like a book research assistant and a stock data analysis workflow, emphasizing best practices for AI-assisted development. It's a concise, hands-on experience designed to enhance your coding efficiency with AI.

print("Applications & Insights")

Simple, Battle-Tested Algorithms Still Outperform AI (8 min. read)
In this insightful piece, Jose Crespo, PhD, challenges the prevailing AI hype by highlighting scenarios where traditional, well-established algorithms continue to outperform modern AI solutions. He emphasizes that while AI offers impressive capabilities, it often introduces unnecessary complexity and resource consumption for tasks where simpler algorithms suffice. Crespo advocates for a balanced approach, urging data scientists and ML engineers to critically assess when to deploy AI and when to rely on time-tested methods, ensuring efficiency and effectiveness in problem-solving.

What's Trending in Data Science and ML? Preparing for 2026 (7 min. read)
Data science is shifting from static dashboards to agentic analytics, where orchestrated multi-agent systems and small language models handle real work like EDA, forecasting, and proactive monitoring. The takeaway for 2026 is practical: learn to architect agent workflows, adopt sLMs you can run cheaply, and build a semantic layer so agents share one source of truth; the article calls out use cases such as conversational dashboards, EDA/cleaning agents, foundation models like TimeGPT, and coordinated pipelines with tools like Tableau Pulse, Power BI Copilot, and causaLens.

Ladder of Evidence in Understanding Effectiveness of New Products (7 min. read)
Meta's analytics team delves into the hierarchy of evidence for assessing new product effectiveness, emphasizing that while Randomized Control Trials (RCTs) are the gold standard, there are scenarios where they're impractical. In such cases, causal inference methods become essential. The article outlines when to use experiments, causal inference, or both, and stresses the importance of clear communication about the confidence levels of different analytical approaches. A practical framework is provided to guide data scientists in selecting appropriate methods based on project constraints and decision impact.

Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning
Large Language Models (LLMs) often falter on multi-step reasoning tasks. Traditional Supervised Fine-Tuning (SFT) rigidly imitates demonstrations, while Reinforcement Learning with Verifiable Rewards (RLVR) struggles when correct solutions are rare. Enter Supervised Reinforcement Learning (SRL): a novel framework that trains models to generate internal reasoning steps before each action, providing smoother, step-wise rewards by aligning model actions with expert trajectories. This approach enables smaller models to tackle complex problems previously out of reach and generalizes effectively to tasks like agentic software engineering.

print("Tools & Resources")

TRENDING MODELS

Text Generation
MiniMaxAI/MiniMax-M2
⇧ 810k Downloads
A 230B parameter model designed for advanced text generation tasks, offering high performance across various applications.

Image-Text-to-Text
deepseek-ai/DeepSeek-OCR
⇧ 2.25M Downloads
A 3B parameter model optimized for optical character recognition, enabling accurate text extraction from images.

Text Generation
moonshotai/Kimi-Linear-48B-A3B-Instruct
⇧ 19.4k Downloads
A 49B parameter model tailored for instruction-based text generation, facilitating nuanced and context-aware outputs.

Text-to-Image
briaai/FIBO
⇧ 3.03k Downloads
A model designed for generating images from textual descriptions, enabling creative visual content creation.

Image-to-Image
dx8152/Qwen-Edit-2509-Multiple-angles
⇧ 191 Likes
A model focused on image-to-image transformations, offering capabilities for editing and enhancing images from various perspectives.

TRENDING AI TOOLS

• 🧠 Guru: Knowledge management tool to capture and share information effortlessly.

• 🚀 CodeBanana: Google Docs–style collaboration, AI context, and live VMs to speed up software development.

• 🔍 Researcher: Streamline research with AI-powered insights in Microsoft 365 Copilot.

• Composer: AI-powered tool to enhance coding productivity and streamline software development.

print("Everything else")

• DBT Labs open sources MetricFlow to enhance data interoperability with an independent schema.

• Amazon's CEO clarified that AI is not responsible for the company's recent layoffs.

• GitHub Copilot CLI now allows users to use custom agents and delegate tasks to the Copilot coding agent.

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