The Big Story

Scientists at Anthropic just announced that their AI model Claude found a completely new enzyme system hiding in the DNA of bacteriophages — viruses that infect bacteria.

Think of bacteriophages like tiny robots that target only bacteria. The AI scanned huge amounts of genetic data and noticed a strange repeating pattern next to a gene that looked similar to CRISPR, the famous gene-editing tool.

This new system appears to have the ability to cut, copy, and paste DNA, just like a few other rare natural tools that scientists already know about. What makes it exciting is that discoveries like this have historically led to big medical breakthroughs.

CRISPR itself turned into the foundation for new genetic medicines, so researchers are wondering whether this unknown system could eventually do something similar. The team is being careful though — they still need to run many more experiments in the lab to figure out exactly what the system does and whether it can be safely used.

For you, this shows how AI is starting to act like a super-fast research assistant that can spot patterns in biology that humans might miss. It doesn't replace scientists, but it speeds up the early idea stage so real people in lab coats can test the most promising leads.

You can't try the biology lab yourself, but you can explore the same kind of AI reasoning by going to claude.ai and asking it to help you brainstorm ideas for a school science project or explain a biology concept in simple terms. Source: x.com

Explain Like I'm 14

You know how when you're looking for a specific song on your phone, you can type just a few words and the search instantly pulls up the right track even though there are millions of songs?

Now picture that same idea but instead of songs, the AI is searching through billions of letters of DNA code from viruses. It doesn't read every single letter one by one like a person would. Instead it has learned patterns from huge amounts of training data, so it can quickly flag when something looks unusual or important.

Next it shows the strange pattern to human biologists, who then decide whether it's worth testing in a real lab. The AI is basically doing the first round of "does this look interesting?" filtering at a speed no human team could match.

What makes this powerful is that the AI isn't guessing randomly — it's using the same kind of pattern recognition you use when you recognize a friend's face in a crowd. Only instead of faces, it's recognizing shapes and repeats inside genetic code.

The result is that scientists get a short list of promising ideas instead of having to check everything themselves. That's why the discovery happened so fast.

Cool Stuff & Try This

Turn your favorite things into animated Skyrim loading screens Ethan Mollick simply told an AI called Opus to make a sequence of fully animated movie-style Skyrim loading screens featuring whatever the AI liked best. The result is a fun, looping website full of moving scenes that feel like they belong in the game. Anyone who loves Skyrim or just wants to see what AI can create from a short prompt will enjoy this. Go to elder-favorites.netlify.app on your phone or laptop and scroll through the different animated screens — notice how the AI kept the classic Skyrim art style while swapping in its own favorite objects. Source: x.com

Bring an old childhood game back to life with modern graphics Ethan Mollick asked Claude to rebuild Rescue Raiders, a game he loved as a kid, but with updated graphics, new goals, and tech trees. The AI worked with helper agents to keep improving the game until it felt fun and playable. If you enjoy retro games or want to see how AI can recreate and upgrade old favorites, this is a great example. Visit rescue-raiders.netlify.app and try playing a few levels — pay attention to how the graphics and gameplay feel compared to the original 1980s version.

Quick Bits

Superforecasters got AI timelines very wrong Last year the best human forecasters gave AI only a 1.7% chance of solving a major unsolved math problem by now. The actual progress turned out to be much faster, which shows how quickly things are moving.

AI benchmark costs keep dropping fast A new graph shows that the price of reaching good scores on tough math and science tests is falling quickly even as the AI gets smarter. This matters because cheaper AI means more people and schools can use powerful tools without spending a lot.

💬 Reply to this email — Patrick reads every one.

Share: X · LinkedIn · WhatsApp

Forwarded this email? Subscribe here — it's free.