Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about how AI is engineering phages that outsmart resistant bacteria, and more!
AI in Action: AI is writing viruses to beat nature's own
Bacteriophage ΦX174 has a genome under 6,000 base pairs, tiny by biology's standards but lethal enough to kill E. coli. In a new Stanford study, chemical engineer Brian Hie and graduate student Samuel King fed a short stretch of its DNA into Evo 2, a generative AI model Hie built to write entire genomes in one pass rather than edit existing ones. Evo 2 generated thousands of candidate phage genomes end to end, with no manual tweaking. The team synthesized nearly 300 of them, tested each against E. coli in the lab, and kept 16 that worked. A handful of Evo 2's designs turned out to have higher fitness than the natural phage they were modeled on.
The real payoff shows up against resistance. Bacteria eventually evolve immunity to any single phage, the same way they do to antibiotics — but a cocktail of genetically distinct phages is much harder to outmaneuver. Hie's team mixed their 16 AI-designed phages together and showed the blend rapidly overcame E. coli that had already become immune to the natural version, a result published in Science this week. Hie says the same approach could extend to phages targeting tuberculosis, MRSA, and Pseudomonas aeruginosa — bacteria behind some of the toughest hospital-acquired infections.
Hie has released Evo 2 as open-source software, free for anyone to use. His reasoning: naturally occurring pathogens are already easier to access and produce than anything an AI model could design, and unlike evolution, an AI tool can have safety checks built into it from the start. He's now pushing Evo 2 toward longer, more complex DNA — including whole bacterial genomes engineered to produce useful chemicals and medicines.
🔥 Rapid Fire
Commentary: The AI Demand Bubble
OpenAI accounts for 70% of Microsoft’s AI revenue
Without two unsustainable AI labs, Big Tech’s era of hypergrowth is over
Analysis: AI Is Getting Way Too Expensive
Big Tech’s Anthropic and OpenAI stakes distort corporate earnings
Over $1 trillion in data center leases have yet to hit Big Tech’s balance sheets
“The commitments show that a substantial part of Big Tech’s AI spending spree has already been locked in, without yet appearing as debt-like lease liabilities on company balance sheets.”
Korean stock market crash threatens to upset tech boom
OpenAI’s recent math “breakthroughs” commit research misconduct
LinkedIn wants users to lean less on AI — a lot less
The New Rules of Online Visibility
Your customers are searching in places your strategy doesn’t reach.
So before your business is buried and left behind, you need to understand the new rules of SEO.
BELAY's SEO in the Age of AI report explains how search is changing, what AI means for your visibility, and the practical steps small businesses like yours can take to stay visible.
BELAY’s U.S.-based Marketing Assistants turn strategy into execution, helping your business stay visible, credible, and competitive in every search.
📖 What We’re Reading
A model that watches a video, reads its captions, and hears its narration has to agree with itself about what just happened. That agreement does not come from the architecture. It comes from the labels underneath it, and from the people who decided what each frame, sentence, and sound clip actually means.
Multimodal foundation models raise the stakes on that decision, because a single training example now carries several modalities that all have to point at the same interpretation. This is why data labeling services have moved from a back-office chore to a determinant of model quality.
