Meta Enables Multi-Object Tracking in One Pass with SAM 3.1

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 our partner 👇

Most data scientists know how to prompt AI tools, but lack the actual AI skills needed to amplify their impact and stay relevant in the job market.

That is why we created the AI Workflows Bootcamp. Every week, you build a real project, get support from mentors, and gain exposure to industry leaders and practitioners via our live sessions.

📌 Enrollment for our April 2026 cohort is now open: Secure your spot (Seats are limited)

print("News & Trends")

Introducing Codex Plugin for Claude Code (2 min. read)

The Codex plugin for Claude Code provides a straightforward way to integrate Codex into your Claude Code workflow. It’s useful for standard reviews, more adversarial passes, and handing off work when a second perspective from another agent is needed. The plugin routes requests through the local Codex CLI and app server, leveraging your existing authentication, configuration, environment, and MCP setup.

Meta Introduces Segment Anything Model 3 (3 min. read)

Meta's Segment Anything Model 3 (SAM 3) is revolutionizing computer vision by enabling precise object segmentation across diverse images and videos with minimal user input. This advancement streamlines tasks like image editing, medical imaging analysis, and autonomous driving, making complex segmentation tasks more accessible and efficient for data scientists and ML engineers.

Details Leak on Anthropic's 'Step-Change' Mythos Model (3 min. read)

Anthropic's upcoming AI model, Claude Mythos, has been inadvertently revealed through a data leak, showcasing a significant leap in capabilities over the current Opus line. Dubbed "Capybara" in internal documents, Mythos excels in coding, academic reasoning, and cybersecurity, prompting Anthropic to restrict early access to cyber defense organizations due to potential security risks. This development underscores the rapid evolution of AI and the critical need for robust safeguards.

print("Applications & Insights")

Things I Learned at OpenAI (5 min. read)
Karina Nguyen reflects on her tenure at OpenAI, emphasizing the art of crafting effective evaluations and benchmarks that drive the field forward. She highlights the nuanced process of post-training, where human creativity shapes models' subjective capabilities like emotional intelligence and humor. Nguyen also underscores the importance of internal tooling for rapid iteration and the value of designing experiments that maximize insights per computational effort. Her insights offer a compelling glimpse into the evolving landscape of AI research.

Architectural Governance at AI Speed (15 min. read)
In the fast-paced world of AI, traditional architectural governance can feel like a bottleneck. This article explores how to adapt governance frameworks to keep up with AI's rapid evolution, emphasizing the need for agility without sacrificing oversight. It delves into strategies for balancing speed and control, ensuring that governance structures are robust yet flexible enough to support continuous innovation in AI-driven environments.

6 Research Papers Every Data Scientist Should Read In 2026 (6 min. read)
This article delves into six pivotal research papers that are shaping the future of data science and AI in 2026. It highlights advancements in autonomous AI agents, such as DeepAnalyze-8B, designed to automate the entire data science pipeline. The piece also explores frameworks like AutoML-Zero, which evolves machine learning algorithms from scratch, and discusses the implications of these technologies on the role of data scientists. A must-read for professionals aiming to stay ahead in the rapidly evolving AI landscape.

Scaling Global Storytelling: Modernizing Localization Analytics at Netflix (6 min. read)
Netflix has revamped its localization analytics to better serve its global audience. By modernizing their data pipelines and analytics tools, they've enhanced the efficiency and accuracy of translating and adapting content across diverse languages and cultures. This overhaul not only streamlines operations but also ensures a more personalized and engaging experience for viewers worldwide.

print("Tools & Resources")

TRENDING MODELS

Image-Text-to-Text
Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
⇧ 337k Downloads
A 28-billion parameter model designed for advanced reasoning tasks, distilled from larger models to enhance efficiency. Updated 7 days ago, it has gained significant traction for its performance in complex image-to-text applications.

Automatic Speech Recognition
CohereLabs/cohere-transcribe-03-2026
⇧ 50.5k Downloads
A state-of-the-art speech recognition model released by CohereLabs, optimized for transcribing audio with high accuracy. Updated about 23 hours ago, it has quickly become popular for its robust performance in various transcription tasks.

Text-to-Speech
mistralai/Voxtral-4B-TTS-2603
⇧ 3.72k Downloads
A 4-billion parameter text-to-speech model developed by Mistral AI, offering natural and expressive voice synthesis. Updated 4 days ago, it has been well-received for its ability to generate human-like speech across different applications.

Image-Text-to-Text
baidu/Qianfan-OCR
⇧ 17.6k Downloads
An optical character recognition model by Baidu, designed to extract text from images with high precision. Updated 5 days ago, it has been widely adopted for its efficiency in converting image-based text into editable formats.

Text Generation
chromadb/context-1
⇧ 2.39k Downloads
A 21-billion parameter text generation model by ChromaDB, tailored for generating contextually relevant and coherent text. Updated 1 day ago, it has gained attention for its ability to produce high-quality text outputs in various applications.

TRENDING AI TOOLS

• 🌏 Brave Search API: Provides private, ad-free search results for seamless integration into AI applications.

• 📝 lat.md: Markdown-based knowledge graph system for structuring and linking codebase context, optimized for AI agents and developers.

• 🙋 Unwrap Customer Intelligence: AI-driven insights to understand customer behavior and improve engagement.

• 🔍 Critique: Enhance research with multi-model intelligence for deeper insights.

print("Everything else")

• A recent arXiv paper advocates for domain-specific superintelligence models to enhance AI sustainability and reasoning capabilities.

• Claude Code introduces a computer use feature, enabling it to control macOS applications via the CLI for tasks like UI testing and debugging.

• The Meta-Harness project introduces an end-to-end optimization framework for model harnesses, enhancing performance across tasks like text classification and mathematical reasoning.

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