- Synthetic
- Posts
- Google AlphaEvolve Transforms Coding 🧬
Google AlphaEvolve Transforms Coding 🧬
Plus, Why Moderna Merged Its HR and Technology Departments 🧠

Subscribe to Synthetic
This week’s most interesting and relevant AI news and analysis
Remember, you can read past editions of Synthetic here.
This Week in AI 🧠
In a letter signed by more than 400 British musicians, writers, and artists Dua Lipa, Sir Elton John, Sir Ian McKellen, Sir Paul McCartney, Kate Bush, Robbie Williams, Richard Curtis (the director of Notting Hill and other great British comedies), Coldplay, Florence Welch, and others have called on the British Prime Minister to update copyright laws to protect them from AI, saying that failing to do so will force them to give away their work to big tech firms.
HUMAIN is a newly-launched arm of the Saudi Public Investment Fund. It announced this week that it has struck a major deal with Nvidia and will make a substantial investment to build AI factories in the Kingdom of Saudi Arabia, with a projected capacity of several hundred thousand GPUs over the next five years, eventually consuming 500 megawatts of energy. Their initial focus will be to build an NVIDIA Omniverse system to simulate physical AI and accelerate the development of advanced robotics systems. The effort is part of the Kingdom’s Vision 2030 initiative to reduce its dependence on oil. Analysts increased their price targets for both Nvidia and AMD on the news. 📈
OpenAI CEO Sam Altman, AMD CEO Lisa Su, CoreWeave CEO Michael Intrator, and Microsoft President Brad Smith urged policymakers to streamline regulations and policies on AI and support AI-related projects in the United States. “It’s very difficult to imagine us figuring out how to comply with 50 different sets of regulations,” said Altman, advocating for a single federal framework of easy-to-understand, “light-touch” regulations that would help the industry move faster and compete with China. China has stated it wants to take a leadership position in AI by 2030. You can watch the tech leaders’ opening statements here (20 mins) and the full hearing here (over 3 hours). 🇨🇳
In related news, Republicans propose prohibiting U.S. states from regulating AI for the next ten years. The Washington Post published an opinion piece calling this approach a “dangerous plan.”
Quick Hits
TikTok Launches AI Alive Feature - The new feature transforms static photos into moving videos that are added to stories. Content will be embedded with C2PA metadata, tagging it as AI-generated. 🎬
Introducing LegoGPT! - A team at Carnegie Mellon University developed a new AI anyone can use for free to turn text prompts into their own physically stable LEGO designs. 🧱
Google Might Replace “I’m Feeling Lucky” Button - After a couple of decades of rare use, the legendary button might be switched out for “AI Mode.” ⏺️
Elon Musk Says Everyone Will Want a C-3PO-like Robot - He also warned AI could lead to a Terminator-like scenario. 🤖
Waymo Recalls 1,200 Robotaxis After Low-Speed Collisions - The cars have trouble seeing gates and chains, causing minor collisions, an issue that is addressed with a software update. 🚙
AI Safety Advocate Believes 90% Probability Superintelligence Will Pose Existential Risk - Max Tegmark urged big-tech AI companies to calculate the so-called Compton constant, which defines the probability that AI will escape human control. Yikes. 😬
Video: Nvidia Director of AI Talks About Roadmap for Embodied AI
Jim Fan is a distinguished scientist and Director of AI at Nvidia. He has a PhD from Stanford, is a great speaker, and co-leads the Project GR00T humanoid robotics project at Nvidia. In this amusing and interesting short talk, Fan proposes a “physical Turing test” to determine when we have built a robot with truly human-like capabilities. Fan goes on to explain why robotics is such a difficult challenge and how Nvidia is using simulation to build physical AI. He describes the key concepts of digital twins, digital cousins, and what he refers to as “digital nomads” (17m 33s).
AI Tech and Innovation 💡

AlphaEvolve is a significant breakthrough from Google DeepMind, demonstrating the powerful combination of RL with LLMs to “evolve” code and deliver novel innovation.
The whizzes in Google DeepMind’s research labs have done it again. The company reserves its “Alpha” prefix to signify that they are releasing a technology that makes a significant leap in capability. For example, AlphaGo, AlphaFold, and AlphaGeometry. Its latest release, AlphaEvolve, is an AI coding agent that evolves algorithms to solve math and other practical problems in computing. This release has profound significance, marking the convergence of two major AI technology strands. Google used AlphaEvolve to enhance its operations, enabling it to reduce energy consumption in its data centers, refine the design of Google’s TPU chips, and improve AI training algorithms. AlphaEvolve has solved mathematics problems that have stumped humans for decades. It’s a big deal.
AlphaEvolve is a Gemini-powered AI coding agent that is able to make new discoveries in computing and mathematics
Synthetic’s Take: This is a significant breakthrough. It shows that Google DeepMind is combining reinforcement learning (RL) with large language models (LLMs) to evolve code over time, ultimately achieving an optimal result. It combines the technology behind platforms like AlphaZero and MuZero with the incredible coding abilities of an LLM to yield novel innovations that go beyond what humans have achieved. It’s now clear that researchers will not build AGI simply by scaling up LLMs. It will require a combination of approaches, and probably some additional breakthroughs. AlphaEvolve demonstrates that Google is already exploring how to combine RL with LLMs and move in that direction.
Meta has been building a vast trove of scientific data in hopes of turbo-charging scientific research. This week, it released the Open Molecule 2025 data set, a collection of more than 100 million molecular snapshots whose properties have been calculated using density functional theory. Meta also released a powerful new AI model named the Universal Frontier Model (UMA) for Atoms. The model took 6 billion compute hours to create (!) and is designed to help scientists accelerate research into new materials and drugs. It simulates the quantum mechanics of atoms and molecules to accurately represent the interactions between molecules.
AI Insights 🔮
Speaking at Queen’s College, Cambridge, Sir Demis Hassabis offered candid advice to students navigating the next decade of extreme change. His advice: “It’s important to use the time you have as an undergraduate to understand yourself better and learn how to learn,” so you can acquire new skills rapidly. He added, “Any time there’s disruption and change, there’s also a huge opportunity. We entered a period like that in the 1990s with the internet, mobile, and gaming. I think we’re in another of those eras, which is exciting, but you’ve got to be very nimble and embrace the new technologies that are coming down the line.”
Vaccine maker Moderna is rethinking its workforce planning to take full advantage of AI capabilities. They are one of the first companies to merge their technology and human resources teams into a single function, overseen by a new role, their Chief People and Digital Technology Officer (CPDTO). Moderna is redesigning teams across the company. Roles are being created, eliminated, and reimagined. Moderna partnered with OpenAI in 2023 and has since made significant strides in integrating AI into its daily workflows. 🙋♀️ 🤖
Synthetic’s Take: Moderna will be among the first of many companies that move in this direction. The workplace of the near future will host a blend of human, digital, and robotic workers, as AI agents and robots become increasingly essential components of the AI-enabled workforce.
Klarna was one of the first companies to rush out of the gate and start experimenting with AI to replace human labor and cut costs through automation. Their approach has failed, and Klarna is reportedly planning a major recruitment drive to attract gig workers, aiming to address the problems that AI has created. Klarna tore up its marketing contracts in 2023 and attempted to replace its entire customer service team with AI in 2024. However, they over-automated, which has negatively impacted customer satisfaction and the Klarna brand.
Synthetic’s Take: Klarna has become a case study of how NOT to integrate AI into business operations. Studies show that the most effective and sustainable approach to creating business value with AI is to pair it with humans to enhance overall performance. Augmentation works better than pure automation, especially when AI technology is still nascent and prone to errors.
Toolkit for the Future 🛠️
Building a helpful AI assistant for your website is so easy now. Just pick a hyper-realistic voice, give your new assistant a simple prompt (e.g., “You are a helpful customer service agent for a keynote speaker”), upload relevant FAQ documents, and integrate it with your website. Done! We used ElevenLabs for stevebrown.ai, and love it. Super easy to set up and configure. ElevenLabs is also great for creating audiobooks or multilingual voice tracks in up to 29 languages to reach new markets. Try it for free. It’s one of the best AI solutions out there.
Create and sell online courses from your website with this top-rated platform. It’s easy to use, with hundreds of beautiful templates and a powerful generative AI assistant that makes content generation a breeze. This is the platform where Synthetic will launch a new AI training course this summer. The AI helps you create lessons, exams, assessments, and more. Free to try. Use promo code 10XR to get 30% off your first two months.
With a support automation platform, you can relieve pressure on your support team and serve customers better and faster. AI answers 90% of queries and escalates the rest to your team. It automatically updates its knowledge database so it’ll never escalate the same issue again. The platform supports SMS, social, email, and phone communication.
Focus on the email that matters. Deep Clean automatically identifies old, unimportant emails so you can delete them. Snooze emails until a better time and easily track emails that need a response. Automatically filter unimportant emails into specific folders. It works across every device where you check your email. Get to inbox zero with award-winning SaneBox.
Try this easy-to-use, AI-powered alternative to DocuSign to write proposals, create quotes, get eSignatures on contracts, build online forms, collect payments, and much more. Great for sales, HR, legal, finance, and IT. Free 14-day trial.
ChatGPT at Work: Free Resource Bundle
Power up your productivity with Mindstream's exclusive ChatGPT toolkit, designed for professionals who want to work smarter, not harder.
Your free bundle includes:
ChatGPT Decision Flowchart
Advanced Prompt Templates
2025 AI Productivity Guide
Task Automation Framework
Industry-Specific Use Cases
Join thousands of AI-powered professionals by subscribing to our daily newsletter. Get the complete bundle instantly after signup - no extra steps required.