AI Stops Searching, Starts Building

Awakend.ai tracks where AI value is moving: off trophy models and into the machines and balance sheets that run them.

This week, Anthropic entered final talks on a $6.3 billion data center in Japan, Micron posted the largest revenue year in the company’s history, and Microsoft Research asked whether robot brains belong in the cloud.

The longer, last, is antibiotic resistance. When a patient’s infection beat every drug and every known alternative, an AI designed something that had never existed. Whether it saved them is still unknown.

When an antibiotic stops working, a doctor reaches for a stronger one, then another, then another. Eventually there is nothing stronger. Doctors call this an antibiotic-resistant infection, and it kills more than 35,000 Americans a year. Hospital cases rose 20% during the pandemic, and most have not come back down. The bacteria keep adapting, and medicine keeps answering with the same kind of drug.

But Medical AI is now starting to build a different kind.

The Other Weapon
There is another way to kill bacteria, and medicine found it before it found antibiotics. In 1917, French scientist Félix d’Hérelle identified viruses that attack bacteria and named them bacteriophages. They hunt bacteria and leave human cells alone.

Two years later, a boy with severe dysentery at a Paris children’s hospital got better after a single dose. It was the first time anyone had treated a patient this way, and it worked.

Then penicillin arrived. Antibiotics treated almost any infection without anyone having to know which bacterium caused it, and phage therapy fell out of use across the West for the next seventy years.

Why Phages Lost
The reason phages lost is aim. Each one attacks a narrow set of bacterial strains and ignores everything else, so treating a patient means hunting through collections of known phages until one matches the exact strain making them sick.

This approach is also known as precision medicine, a treatment matched to one patient’s infection rather than aimed at every infection at once. Dr. Marisa Azad, an infectious disease clinician at the Ottawa Hospital who collaborates with Radical Numerics, says finding a suitable phage is very time consuming. An antibiotic only had to be found once.

When Nothing Matches
This year a patient arrived with an infection no antibiotic could touch, caused by E. coli, a common gut bacterium most people carry safely. That is when a doctor turns to phages. A lab tested a sample of the bacteria against the collection, one phage at a time, but nothing killed it.

A phage collection is assembled from sewage, soil, hospital samples, and the standing water at zoos and nature reserves. One database alone holds hundreds of thousands. But none of that matters without a match, and this patient did not have one.

Nobody can request a phage that has not been picked up yet, so the only move left is to go back out and search again. But the infection does not wait for that.

A Phage From Scratch
Radical Numerics, a company that launched in June 2026, trained a model called Omnii-1 on 28 kinds of biological data at once. It reads DNA the way a language model reads text, which means it can also write.

So instead of searching the collection for a phage that matched this strain, the company had Omnii-1 design new ones. In laboratory testing, those phages killed the bacteria taken from the patient. The FDA authorized the treatment, and on October 20 the company reported that doctors had given it to the patient. 

Radical Numerics says it is the first time an AI-designed phage has been used as medicine in a person. Precision medicine has always depended on finding the right match first, and a model that writes one skips that step entirely.

Why It Matters: The real story here isn’t a new phage. It’s AI making treatment built for one patient faster. Every phage in every collection was found by someone out gathering samples from sewage and soil, one at a time, across a hundred years. This one was written by an AI model for a strain nothing anyone had gathered could kill.

That difference is what matters to patients. AI is now shortening the same step everywhere, pulling rare disease diagnosis from years down to weeks. Nobody knows yet whether these phages worked, but it could point to a new way to use AI in medicine, providing individualized treatment that targets even the most resistant bacterial infections.

AI Drug Design Is on Track to 3x by 2031

Takeaway:
The money is moving from finding drugs to designing them. Building a molecule from scratch is already the biggest use of this software in a market growing 28.1% a year, but pharma is paying mostly in promises, with Isomorphic Labs alone holding nearly $3 billion in deals. Not one of those engines has an approved drug. The spending is real, the pipeline is early, and the first AI-designed phage to reach a patient came from the smallest company in the field.

The Timeline Guide:

Each timeline and graph represents the realistic stage of the covered technology, plotted from concept to scale, capturing where it stands today and when broad deployment is likely.

Hit reply with anything else on your mind; we actually read every note.
See you in the next upload

Note: Images in this newsletter are often AI generated for illustrative purposes only. Market forecasts are hypothetical, inherently uncertain, represent our best-guess estimates only, and should not be considered investment advice.