In this week’s newsletter

✅ AI speeds up bioengineering - and science. ✅ A missed opportunity in the Senate hearing on longevity. ✅ Longevity fast grant program from the Foresight Institute. ✅ ”My passion is to solve aging.”

🤙🏼 Want to connect? Add me on LinkedIn. 🙏🏼 Not subscribed to the LEVITY podcast on Youtube yet? Do it here. 🎧 More of a listener? The LEVITY podcast is also available on Spotify, Apple Podcasts and other places.

The Co-Scientist and the Genome Designer

As my last couple of newsletters has been focused on artificial intelligence I had decided to start with something else this week. One idea was to revisit [redacted, as I’ll save it for another day]. I even had other topics lined up, ready to take the top spot above the ad.

But, as has happened before (and will undoubtedly happen again), my newsfeed had other plans. Several genuinely exciting AI developments dropped, and I couldn’t resist. So, let’s take a closer look at each of them.

Google DeepMind’s Co-Scientist

You might recall how I a few weeks ago made a big deal out of OpenAI’s tool Deep Research and underscored that it could provide novel insights. Well, these days two weeks is a long time in AI development.

Not only can Google DeepMind’s Co-Scientist generate novel research hypotheses, but these ideas have already been experimentally verified in the real world. For example it generated drug repurposing candidates to inhibit tumor viability in Acute Myeloid Leukemia. And it proposed and ranked hypotheses for identifying epigenetic targets with promising anti-fibrotic activity in liver fibrosis.

Co-Scientist is a multi-agent system built on the multimodal model Gemini 2.0, capable of taking a plain-language research goal and breaking it down into generating, refining, and ranking novel hypotheses - essentially following the scientific method. It employs a mix of specialized agents (for generation, reflection, evaluation, and so on) to create and iterate on ideas through rapid, self-improving cycles.

One particularly eye-catching experiment involved Co-Scientist re-discovering critical gene transfer mechanisms behind antimicrobial resistance.

Re-discovering?

Yes, in previous work, human researchers had experimentally validated a novel mechanism contributing to the spread of antimicrobial resistance. DeepMind’s Co-Scientist started with the same literature these researchers had access to when they began forming their hypothesis.

Not only did Co-Scientist arrive at the same conclusion - it did so in just two days.

Is that fast?

Considering it took human researchers twelve years to reach the same point, I'd say that's not bad for a few day's work. And I suspect Professor José R Penadés at Imperial College London might agree.

After all, Penadés and his team were behind that decade-long effort. He told the BBC ”of his shock when he found what it had done, given his research was not published so could not have been found by the AI system in the public domain.”

And if that wasn't enough, Co-Scientist didn’t just replicate existing research - it went further.

”It's not just that the top hypothesis they provide was the right one”, Penadés told the BBC. ”It's that they provide another four, and all of them made sense. And for one of them, we never thought about it, and we're now working on that.”

Evo 2 from Arc Institute

As impressive as Co-Scientist is - and let's not forget, these are still early days, so it's only going to get better - if I had to pick the AI Development of the Week, my vote would go to the foundation model Evo 2.

Subscribe to LEVITY Premium to read the rest.