Weekly Dose of Optimism #144
CRISPR Baby, Flozins, AlphaEvolve, Meta FAIR, Alchemy, Overdose & Murder Drop
Hi friends 👋,
Happy Friday and welcome back to our 144th Weekly Dose of Optimism. We got baby-saving treatments, a new class of Ozempic-level-impressive drugs, two of the world’s largest companies open sourcing AI science tools, alchemy (sort of), and the first good news in overdoses I’ve read in…forever? These are the types of weeks we live for here at Not Boring.
Let’s get to it.
BONUS: Baby is Healed With World’s First Personalized Gene Editing Treatment
By Gina Kolata at The New York Times
Instead, KJ has made medical history. The baby, now 9 ½ months old, became the first patient of any age to have a custom gene-editing treatment, according to his doctors. He received an infusion made just for him and designed to fix his precise mutation.
A BONUS?!? At the beginning of the Weekly Dose?! Oh baby, it must be important.
Just after we wrapped writing yesterday, the New York Times dropped the mother of all whitepills: a baby became the first human to receive a personalized gene edit.
Baby KJ was born with CPS1 deficiency, a rare, often fatal genetic disorder. In a race against time, Dr. Kiran Musunuru and his team designed a custom CRISPR base-editing treatment tailored to KJ’s exact mutation. Delivered via lipid nanoparticles to his liver and administered in three escalating doses, the therapy slashed his need for medication and let him eat protein like any healthy baby, thereby saving his life.
Why this matters:
Gene-editing therapies can now be personalized and deployed fast, bypassing the traditional years-long drug development cycle.
The same method could be adapted for thousands of other rare, and even common, genetic diseases.
Most importantly, Baby KJ is hopefully going to live a long and happy life. Look at that smile!
Packy note: Nowhere near as cool as the treatment itself but still cool is where it happened: the Children’s Hospital of Philadelphia (CHOP), our hometown, home to the Super Bowl Champion Philadelphia Eagles, and most importantly, where a baby Dan McCormick got the open heart surgery that, almost thirty years later, means that he’s alive to bring you stories like the one about KJ. Go CHOP.
Today’s Weekly Dose is brought to you by… INBOUND
Dwarkesh Patel. Dario Amodei. Marques Brownlee. Sean Evans. Amy Poehler. And more. Under one San Franciscan roof, for INBOUND 2025.
HubSpot just unveiled the INBOUND 2025 speaker lineup, and it’s stacked with tech leaders and creatives. Art and science.
This isn’t a boring conference. It's a rare opportunity to learn directly from icons like Sean Evans (Hot Ones) and Dwarkesh Patel about content creation that builds thriving businesses, while gaining AI expertise from leaders like Synthesia co-founder Victor Riparbelli and Anthropic co-founder Dario Amodei.
It’s all happening in San Francisco, the home of AI, and the perfect place to learn about content and distribution, building authentic connections, and leveraging emerging technologies to do what you do, better.
Not Boring readers get 10% off their INBOUND General Admission ticket with code NOTBORING10 (Valid thru 7/31). You read about these people in Not Boring; join them live.
(1) Everything drugs
From The Works in Progress Newsletter
Sodium-glucose cotransporter-2 (SGLT2) inhibitors, also called flozins, began as diabetes drugs. Surprisingly, they turned out to also be very effective at improving heart health. Then they were discovered to slow the progression of chronic kidney disease, one of the leading causes of death and disability worldwide. Preliminary evidence indicates that they show promise in helping several other conditions, but no one knows exactly how they achieve this yet. Could SGLT2 inhibitors be a new medical Swiss Army knife?
First GLP-1s, now flozins. We’re living in an extremely transformative period of modern medicines.
Originally developed to lower blood sugar in type 2 diabetes, SGLT2 inhibitors, or flozins, turn out to have far broader effects. Does this story sound familiar to you…?
SGLT2 inhibitors work by blocking a kidney protein that reabsorbs glucose, leading to more sugar being excreted in urine. After proving effective in lowering blood sugar, they unexpectedly showed major benefits for heart health and kidney disease, significantly reducing hospitalizations, cardiovascular deaths, and progression to end-stage kidney failure, even in non-diabetics. We’re living the GLP-1 story all over again.
The exact mechanism behind the the rangy drugs is unclear, but theories include improved kidney cell function and heart metabolism. The early trial results are promising: flozins lowered blood sugar and restored insulin sensitivity, while cutting heart failure hospitalizations by 35%, cardiovascular deaths by 38%, and reducing progression to end-stage kidney disease by 36%. Those are some crazy results and if you apply those rates to the number of people with each condition, the potential impact of these drugs becomes very obvious. About 18 million people die each year from cardiovascular disease and nearly 1 million people have chronic kidney disease. These drugs could potentially help millions of people each year from otherwise very serious conditions.
I have a feeling we’re going to start hearing a lot more about flozins over the next couple of months and years.
(2) AlphaEvolve: A coding agent for scientific and algorithmic discovery
From Google’s DeepMind
In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code.
Google’s DeepMind introduced AlphaEvolve, an evolutionary coding agent that uses LLMs to improve algorithms by modifying code. The coding agent refines algorithms iteratively, similar to biological evolution, by repeatedly modifying and testing them. It uses feedback from evaluators, which score the algorithm's performance, to guide the modification process.
And AlphaEvolve works: it achieved state-of-the-art results in matrix multiplication, even surpassing a 56-year-old record, and discovered novel algorithms that outperform existing solutions in mathematics and computer science. Importantly, Google is already using AlphaEvolve to optimize real-world systems, such as data center scheduling.
AI is going to automate the AI that drives new AI research and discovery. The feedback loops will be tight and will drive accelerated returns in science and R&D. AlphaEvolve is a relatively early example of what that could look like.
Google DeepMind is quietly cooking. Beyond pure LLMs, they’ve released some of the most interesting models, from AlphaGo to AlphaFold to AlphaEvolve. In LLMs, its Gemini is largely believed to be the best model.
What’s wild is that it seems easier for the company to develop world-class AI models than the products that make people want to actually use them, despite its vast resources, distribution advantage, and decades of experience making consumer products.
We here at Not Boring love that. Nothing makes us more optimistic than incumbents fumbling the bag and giving new startups the chance to win.
From Meta
Meta FAIR is sharing new research artifacts that highlight our commitment to achieving advanced machine intelligence (AMI) through focused scientific and academic progress. The work we’re sharing includes Open Molecules 2025, a dataset for advancing molecular discovery, and Meta’s Universal Model for Atoms.
OK class, everyone say it with me: “Thank you Mr. Zuckerbeeeerrrgg”
Meta FAIR (Fundamental AI Research) released a suite of groundbreaking AI research tools and findings aimed at advancing machine intelligence through open science. Key among them are Open Molecules 2025 (OMol25), the largest quantum chemistry dataset to date, and the Universal Model for Atoms (UMA), a powerful ML model trained on 30B+ atoms, both of which drastically accelerate atomic-scale discovery in healthcare, energy, and materials science. The lab also introduced Adjoint Sampling, which can train generative models without any data, using only reward signals which is a breakthrough for low-data domains like chemistry and physics. The best part: all of this will be open sourced.
Additionally, in a neuroscience first, Meta and the Rothschild Foundation Hospital mapped how language develops in children’s brains, revealing parallels with LLMs. They conducted the first large-scale study using brain implants in children with epilepsy to record how language understanding evolves in the brain. By analyzing neural activity as the children listened to stories, researchers found that the brain’s language representations develop over time in patterns that mirror how large language models learn.
Packy note: Dan’s kind of obsessed with Zuck, but this is cool.
(4) CERN scientists turn lead into gold (briefly)
From Swiss Info
Physicists at the CERN nuclear research centre in Geneva have turned lead into gold. During the collision of lead nuclei at almost the speed of light, the conversion of lead into gold was measured using a new mechanism, the organisation said.
Alchemy! Or at least brief alchemy! Or least atomic alchemy!
CERN physicists have (kinda) successfully turned lead into gold by smashing lead nuclei together at near light speed in the Large Hadron Collider (LHC). This transformation occurs when three protons are knocked off a lead nucleus (which has 82 protons), creating gold (79 protons). Unfortunately, the quantities are vanishingly small and fleeting, far from forming any usable material. Womp womp. But perhaps more importantly, the collisions also create quark-gluon plasma, the primordial soup believed to have existed just after the Big Bang, offering insights into the fundamental makeup of matter.
CERN has been running this experiment for over a decade. Although the nuclear research lab was quick to emphasize that they had not achieved alchemy, it was a nice symbolic moment where science achieved what ancient dreamers only imagined.
Maybe Sir Isaac Newton wasn’t as dumb as he looked…
Some more good news elsewhere in the nuclear research lab world: the National Ignition Facility in Livermore, California, has demonstrated a record fusion gain (also known as Q) of 4—a 300% improvement over the initial breakthrough gain achieved just two years ago. This achievement has not been widely reported, but the former head of the NIF posted about it on Linkedin (h/t Owen Lewis for the find), which is at once kind of awesome and weirdly unexpected. 7 Things Achieving Q=4 Taught Me About B2B Sales.
(5) Overdose Deaths Plummeted in 2024
Jan Hoffman for The New York Times
Overdose deaths in the United States fell by nearly 30,000 last year, the government reported on Wednesday, the strongest sign yet that the country is making progress against one of its deadliest, most intractable public health crises.
This is one of those “good news from a terrible situation” stories. Drug overdoses in the U.S. fell last year from nearly 110,000 in 2023 to 83,000 in 2024. Obviously that’s 83,000 deaths too many, but the good news is that overdose rates started to plummet in 2023 after experiencing steady growth for the past decade.
2024 was the first year of meaningful decline in over a decade and it’s due to a combination of factors: wider access to treatment programs and services, Naloxone (Narcan) is now sold over-the-counter, and a suspected drop of fentanyl in the drug supply. There are obviously very foundational issues we need to fix in this country to curb the use of dangerous drugs: treating mental health properly, addressing economic despair, reducing isolation and loneliness, and of course reducing the availability of dangerous drugs on the market. But our hope is that while we, as a nation, work to improve the underlying causes, we can also put in place the programs, increase availability to potentially life saving treatments, and otherwise support the communities and individuals that are so prone to the life crushing effects of dangerous drug use. It looks like we’re finally starting to make some progress.
Plus, it looks like murder is falling across the country. I can’t imagine the two stats (overdoses and murder) are completely uncorrelated. Shout out to Philly, for seeing the largest drop in homicides of any major city. Anecdotally, you can actually feel this when visiting Philadelphia: it feels safer and less chaotic.
Great work across the board. Fewer overdoses. Fewer murders. Let’s keep driving these to 0. The future’s going to be too good for all these people to miss.
Have a great weekend y’all.
Thanks to INBOUND 2025 for sponsoring. We’ll be back in your inbox next week.
Thanks for reading,
Packy + Dan