AlphaFold 3 opens to the world

Welcome, AI enthusiasts.

The crown jewel of scientific AI research just went open-source, with DeepMind releasing its Nobel Prize-winning AlphaFold 3 to researchers worldwide.

With brilliant new minds now gaining access to this powerhouse tool, could the open-source turn trigger biology's biggest breakthrough yet? Let’s get into it…

In today’s AI rundown:

• DeepMind opens AlphaFold 3 to researchers worldwide

• Qwen unveils powerful new open-source coding AI

• Create multilingual subtitles with Gemini

• AI detects blood pressure and diabetes from short videos

• 5 new AI tools & 5 new AI jobs

• More AI & tech news

Read time: 4 minutes

Important: The Rundown is slowly being sent from a new email address. To ensure you get our newsletter, please add [email protected] to your contact list.

LATEST DEVELOPMENTS

GOOGLE DEEPMIND

🧬 DeepMind opens AlphaFold 3 to researchers worldwide

The Rundown: Google DeepMind just open-sourced its groundbreaking AlphaFold 3 protein prediction model, enabling academic researchers to access both code and training weights for the first time since its limited release in May.

The details:

• The Nobel Prize-winning technology can predict interactions between proteins and other molecules like DNA, RNA, and potential drug compounds.

• Academic researchers can access the model's full capabilities for non-commercial use, though commercial applications remain restricted.

• The system has already mapped over 200M protein structures, demonstrating unprecedented scale in structural biology.

• Several companies, including Baidu and ByteDance, have already created their own versions based on the original paper's specifications.

• DeepMind's spinoff, Isomorphic Labs, maintains exclusive commercial rights, having recently secured $3 billion in pharmaceutical partnerships.

Why it matters: Scientific research is one of the most exciting areas for AI, and the wider availability of AlphaFold via open-source should massively accelerate breakthroughs across biology and medicine – while also leveling the playing field beyond well-funded institutions or pharmaceutical companies.

TOGETHER WITH WEIGHTS & BIASES

📊 Multimodal GenAI tracking and evaluations made easy

The Rundown: Weave is a lightweight developer toolkit that simplifies tracking and evaluation of your multimodal LLM applications — bringing comprehensive observability to text, images, and audio.

With Weave, you can:

• Implement full LLM tracking with just 3 lines of code

• Build custom evaluation pipelines suited to your needs

• Integrate seamlessly with OpenAI, Gemini, Anthropic, and many more

Get started now for free and streamline your AI tracking and evaluation.

QWEN

🚀 Qwen unveils powerful new open-source coding AI

The Rundown: Alibaba Cloud’s Qwen just released a suite of new AI coding models, with its flagship 32B version matching GPT-4o and Claude 3.5 Sonnet's performances on key benchmarks while remaining completely open-source.

The details:

• The Qwen2.5-Coder series spans six different sizes (0.5B to 32B parameters), making it accessible for various computing environments and tasks.

• The 32B version achieves state-of-the-art performance among open-source models in code generation, repair, and reasoning tasks.

• The models integrate with popular development tools like Cursor and are proficient across over 40 programming languages.

• Each size has two variants: a base model for custom fine-tuning and an instruction-tuned version ready for direct use.

Why it matters: AI’s coding abilities continue to level up, and open-source models like Qwen are now matching and exceeding the top players in the industry. Advanced programming capabilities are quickly becoming available to a much wider audience — no coding background is necessary.

AI TRAINING

🎬 Create multilingual subtitles with Gemini

The Rundown: Google Gemini’s video input feature lets you transform any video into multiple languages with accurate transcripts and professional subtitles.

Step-by-step:

1. Visit Google AI Studio and select Gemini 1.5 Pro 002.

2. Upload your video and generate a timestamped transcript.

3. Use translation prompts to create subtitles in target languages.

4. Enhance subtitles with proper formatting and timing.

Pro tip: Mention the source language in your initial prompt for better accuracy, and use Subtitle Edit Online for final touches. We have also written detailed prompts that you can copy and paste from here.

PRESENTED BY SAMBANOVA

🏆 Build lightning-fast AI agents

The Rundown: SambaNova is hosting the Lightning Fast AI Hackathon, offering developers the opportunity to build AI agents faster than ever before. Utilize their custom hardware stack and compete to win up to $5,000!

By participating, you can:

• Experience the fastest inference on the planet (10x compared to traditional GPUs)

• Use open-source models you love, like Llama

• Deploy agents quickly with simple OpenAI-compatible APIs

Join the hackathon today — the competition ends November 22.

AI RESEARCH

🏥 AI detects blood pressure and diabetes from short videos

The Rundown: Japanese researchers just developed an AI system that can screen for conditions like high blood pressure and diabetes using a brief video of someone's face and hands—with accuracy at levels comparable to or exceeding those of cuffs and wearable devices.

The details:

• The system combines high-speed video capture with AI to analyze subtle changes in blood flow patterns, analyzing 30 regions of the face and palm.

• Initial tests show 94% accuracy in detecting high blood pressure and 75% accuracy for diabetes compared to traditional diagnostic methods.

• A 30-second video achieved 86% accuracy in blood pressure detection, while even a 5-second clip maintained 81% accuracy.

• Researchers envision future integration into smartphones or smart mirrors for more convenient at-home health monitoring.

Why it matters: It may be time to ditch the bulky blood pressure cuffs—a simple selfie will soon do the trick. Integrating this type of AI breakthrough into accessible forms like an app or website would dramatically increase access to vital screenings while making personal health monitoring much easier and more effective.

NEW TOOLS & JOBS

Trending AI Tools

• 📊 Spiky - AI-powered real-time insights for faster, smarter sales decisions

• 🗣️ Univerbal - Boost your speaking confidence in 20+ languages with personalized AI tutors

• 🧠 Theo - Give your AI assistant a complete picture of your unique business model and strategy for more nuanced, aligned outputs

• ⚙️ Lamatic AI - Build AI agents in low-code and deploy on edge

• 🍄 Truffle - Stay ahead of the conversation with AI-powered X tracking

New AI Job Opportunities

• 📧 The Rundown - Social Media Manager

• 🦺 Waymo - Field Safety Lead

• 👥 Meta - Recruiter

• 🖥️ Cohere - IT Operations Specialist

• 🤖 Metropolis - Senior Machine Learning Engineer

QUICK HITS

Free event: AI Safety and the Risks of AGI. Find out what people aren’t talking about, but should be (for the sake of humanity), when it comes to AGI. RSVP here.*

Lex Fridman released a new interview with Anthropic CEO Dario Amodei, who discussed the firm’s approach to AI safety and predicted AGI may arrive by 2026-2027, as well as conversations with researcher Amanda Askell and co-founder Chris Olah.

AI sales automation startup 11x secured $50M in new funding, valuing the company at $320M as it expands its AI bots that can handle sales tasks in 30 languages.

Anthropic hired Kyle Fish as its first dedicated "AI welfare" researcher, who will explore whether future AI models might experience consciousness and require moral consideration.

The Vatican and Microsoft unveiled a digital AI-powered twin of St. Peter’s Basilica created from 400,000 images, enabling virtual visits and help identifying structural damage ahead of the 2025 Jubilee.

Jerry Garcia's estate announced a partnership with ElevenLabs, bringing the late Grateful Dead icon's AI-recreated voice to audiobooks and written content in 32 languages.

Leading AI companies are reportedly rushing to develop new benchmarks and testing methods, with current standards falling behind the ability to measure increasingly sophisticated AI models.

*Sponsored listing

THAT’S A WRAP

See you soon,

Rowan, Joey, Zach, and Alvaro—aka The Rundown Team