- Evolving AI Insights
- Posts
- 🧬 Google's Self-Evolving AI that Writes its own Code
🧬 Google's Self-Evolving AI that Writes its own Code
Also: GPT-4.1 Now Available in ChatGPT

Welcome, AI enthusiasts
Google Deepmind has unveiled AlphaEvolve, a system that uses large language models to automate the discovery and refinement of algorithms. The aim is to develop new algorithms, systematically improve them, and adapt them for practical use. Let’s dive in!
In today’s insights:
Google's Self-Evolving AI that Writes its own Code
GPT-4.1 Now Available in ChatGPT
AI Campus Deal Reshapes US-UAE Tech Ties
Read time: 4 minutes
LATEST DEVELOPMENTS
Evolving AI: A new coding agent evolves algorithms using large language models and evaluators.
Key Points:
AlphaEvolve uses Gemini models and automated evaluators to improve algorithms through an evolutionary approach.
It's been deployed across Google to boost efficiency in data centers, chip design and AI training.
It has discovered new solutions to open problems in math, including improving matrix multiplication.
Details:
AlphaEvolve is DeepMind’s new system that uses Gemini Flash and Pro to write and improve code automatically. The agent works like evolution: it generates algorithm candidates, tests them with automated evaluators, and evolves the best ones through many iterations. This self-improving loop has already made an impact inside Google — recovering compute resources, speeding up large model training, and designing better chips. It’s not just fine-tuning; it’s creating better algorithms from scratch.
Why It Matters:
This isn’t just a coding shortcut. AlphaEvolve can now outperform human experts in finding better algorithms, solving long-standing math problems, and optimizing how computers work. It means parts of computer science once reserved for top researchers are now accessible to AI. The tools that power today’s digital world — how we search, stream, train AI, even simulate proteins — can now be rebuilt and improved by machines, without needing human guidance. That changes who controls progress, how fast it moves, and what becomes possible next.
Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.
OPENAI
🚀 GPT-4.1 Now Available in ChatGPT
By popular request, GPT-4.1 will be available directly in ChatGPT starting today.
GPT-4.1 is a specialized model that excels at coding tasks & instruction following. Because it’s faster, it’s a great alternative to OpenAI o3 & o4-mini for everyday coding needs.
— OpenAI (@OpenAI)
5:36 PM • May 14, 2025
Evolving AI: OpenAI rolls out GPT-4.1 and GPT-4.1 mini to ChatGPT users, boosting speed and context.
Key Points:
GPT-4.1 is now available to all paid ChatGPT users, with Enterprise and Edu access coming soon.
SGPT-4.1 mini becomes the default model for all accounts, including free users.
Both models support up to one million tokens and improve coding and instruction tasks.
Details:
OpenAI has launched GPT-4.1 and GPT-4.1 mini in ChatGPT, replacing previous models across tiers. Paid users can choose GPT-4.1 in the dropdown menu, while GPT-4.1 mini now powers all default chats, including those on free plans. Both versions significantly outperform earlier models, especially for coding and task-following, and introduce a million-token context window—nearly ten times GPT-4o’s limit.
Why It Matters:
GPT-4.1's enhancements in coding tasks mean it can better assist developers by generating more accurate and efficient code snippets. Improved instruction following ensures that the AI adheres more closely to user prompts, reducing the need for repeated clarifications.
UNITED ARAB EMIRATES
🇦🇪 AI Campus Deal Reshapes US-UAE Tech Ties
Evolving AI: UAE strikes deal to build world's biggest AI campus outside US, easing prior chip restrictions.
Key Points:
UAE will build a 10-square-mile AI campus with 5GW capacity, managed by US companies.
The deal enables UAE to import 500,000 Nvidia AI chips annually starting 2025.
Trump administration eased prior export controls, signaling trust in UAE's tech handling.
Details:
The UAE and US have finalized an agreement allowing the Gulf nation to build the largest AI campus outside America, a move previously blocked due to China-related concerns. The campus, based in Abu Dhabi, will support up to 2.5 million Nvidia B200 chips. Operated by American firms, it will include data centers and cloud infrastructure. Companies like Qualcomm and AWS are also involved in cybersecurity and engineering efforts. Under the deal, the UAE must match its US investments in AI infrastructure and align security standards.
Why It Matters:
The UAE gains access to half a million top-tier AI chips a year. It marks a shift in AI power — more global, but still under U.S. oversight and control.
QUICK HITS
🚀 New AI and accessibility updates across Android, Chrome and more.
🖥️ Microsoft starts testing ‘Hey, Copilot!’ in Windows.
👏 OpenAI pledges to publish AI safety test results more often.
🤖 Google's Gemini AI helps YouTube find the perfect advertising spaces.
☢️ UK needs more nuclear to power AI, says Amazon boss.
🚀 Anthropic is reportedly testing Claude models that can fix their own mistakes.
📈 Trending AI Tools
What'd you think of today's edition? |
Reply