Anthropic Unveils Claude 3.7 Sonnet & Claude Code

Hey there 👋

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

Here is your 5-minute pulse...

print("News & Trends")

Anthropic Unveils Claude 3.7 Sonnet and Claude Code (5 min. read)

Anthropic introduces Claude 3.7 Sonnet, a pioneering hybrid reasoning AI model offering both rapid responses and in-depth, step-by-step analysis. This release includes Claude Code, a command-line tool in limited research preview, enabling developers to delegate complex coding tasks directly from their terminals. Extended thinking mode enhances performance in areas like math, physics, and coding, with API users gaining fine-grained control over the model's reasoning duration.

Qwen’s Unveils New Open-Source Thinking Model: QwQ-Max-Preview (5 min. read)

Qwen has unveiled QwQ-Max-Preview, a model built upon Qwen2.5-Max, excelling in deep reasoning, mathematics, coding, and agent-related tasks. This preview offers a glimpse into the forthcoming QwQ-Max release, with plans for open-source availability under the Apache 2.0 License.

OpenAI Rolls Out Its AI Agent, Operator, in Several Countries (1 min. read)

OpenAI is expanding Operator, its AI agent that handles tasks like booking tickets and shopping, to ChatGPT Pro users in several countries outside the U.S. The $200/month tool runs in a separate browser window and faces competition from Google, Anthropic, and Rabbit, though their offerings remain limited in availability.

Google’s New AI video Model Veo 2 Will Cost 50 Cents Per Second (1 min. read)

Google has unveiled pricing for Veo 2, its AI video generator, at 50 cents per second—$1,800 per hour. While far cheaper than Hollywood budgets, it’s still a hefty cost. Meanwhile, OpenAI’s Sora model is bundled with a $200/month ChatGPT Pro subscription, offering a different value proposition for creators.

DeepSeek to Open-Source Parts of Online Services Code (1 min. read)

DeepSeek is doubling down on open source, pledging to release five production-tested code repositories starting this week. The move reinforces its commitment to community-driven AI and pressures OpenAI, whose CEO admits DeepSeek has narrowed the tech gap. With daily unlocks promised, the AI race is heating up—no ivory towers, just innovation.

print("Applications & Insights")

CASE STUDIES

How Orakl Oncology is using DINOv2 to accelerate cancer treatment discovery (4 min. read)
Orakl Oncology leverages Meta’s DINOv2, a self-supervised vision model, to accelerate cancer treatment discovery. By applying biomedical image analysis, DINOv2 enhances tumor characterization and drug screening, reducing manual effort and improving efficiency in identifying potential therapies. This open-source AI integration is pushing precision oncology forward.

How Rocket Companies modernized their data science solution on AWS (14 min. read)
Rocket Companies modernized its data science infrastructure on AWS, cutting model deployment time from 8 weeks to under an hour. Using Amazon SageMaker, Step Functions, and CI/CD pipelines, they streamlined ML workflows, reduced incident tickets by 99%, and now power 10 million AI-driven decisions daily.

TUTORIALS

Build and run a deep research agent with LangGraph Studio (Video)
“Deep research” AI assistants are gaining traction, enabling autonomous, in-depth exploration of user-supplied topics. This video overview covers common architectures across open and closed-source tools, detailing an implementation that runs locally with LangGraph Studio. Users can customize models, search APIs, report structures, and research depth for tailored insights.

Understanding RAG Part VI: Effective Retrieval Optimization (6 min. read)
Optimizing retrieval is key to boosting RAG performance. This guide explores advanced techniques like hybrid search, query expansion, and reranking to enhance accuracy and relevance. By fine-tuning embeddings, leveraging metadata, and balancing precision with recall, you can refine retrieval for more effective AI-driven responses.

Build and deploy a Brick Breaker game in under 3 minutes (Video)
Learn how to create and deploy a Brick Breaker game using Grok 3 and Replit—without writing code. Using Grok for HTML generation and Replit’s AI-powered setup, you can configure, run, and deploy your app in under five minutes.

print("Tools & Resources")

TRENDING MODELS

Text Generation
perplexity-ai/r1-1776
⇧ 8.06k Downloads
A language model developed by Perplexity AI, designed for generating human-like text responses.

Image-Text-to-Text
microsoft/OmniParser-v2.0
⇧ 4.89k Downloads
OmniParser V2 by Microsoft converts UI screenshots into structured, interpretable elements, enhancing GUI automation capabilities.

Audio-Text-to-Text
stepfun-ai/Step-Audio-Chat
⇧ 751 Downloads
Step-Audio-Chat is an AI model designed to facilitate audio-based conversations, transcribing and understanding spoken language.

Text-to-Video
stepfun-ai/stepvideo-t2v
⇧ 1.17k Downloads
StepVideo-T2V enables the generation of video content from textual descriptions, bridging the gap between text and visual media.

Image-to-Video
Skywork/SkyReels-V1-Hunyuan-I2V
⇧ 24.1k Downloads
SkyReels-V1-Hunyuan-I2V is a model that transforms images into videos, allowing for dynamic visual content creation from static images.

TRENDING AI TOOLS

• 🧑‍💻 The Windsurf Editor: The first agentic IDE

• 🐼 PandasAI: Makes data analysis conversational using LLMs and RAG.

• 🛠️ Flags SDK: Free open-source library for feature flags and A/B tests in apps

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