Welcome back friends.

Type 1 diabetes (the really bad one caused by your own immune system…) may have a surprisingly simple solution: someone else’s immune system.

Stanford researchers combined blood stem cells and pancreatic islet cells from mismatched mice donors (i.e. not a perfect genetic match), creating hybrid immune systems. Result: 19 out of 19 mice avoided diabetes entirely and 9 out of 9 with diabetes were cured. No immune suppressants and no insulin, and they stayed healthy for six months.

This builds on earlier work with kidney transplants, but cracks the harder autoimmune problem where the immune system targets insulin-producing cells like tiny cellular fascists. It turns out one commonly used autoimmune medication in the pre-transplant mix was enough to fix the issue.

Even better, every component already exists in clinical practice so human trials look feasible. The catch is supply: islets come only from deceased donors, and you need blood stem cells from the same person but Stanford is now exploring lab-grown islets to eliminate the queue.

🧐 What's in it for me? If this translates to humans, Type 1 diabetes could become curable rather than manageable. For autoimmune diseases broadly, this also hints at a new paradigm where we can reset the immune system instead of suppressing it forever. Trials could start within a few years, assuming the islet supply problem gets solved. 

💵 Out of the Lab: Autoimmune diseases represent a massive untapped and terrifyingly fast growing market, whilst Stanford's conditioning approach could unlock stem cell transplants far beyond diabetes.

Academia has a new existential crisis, and for once it's not about funding. An AI called Kosmos can apparently do six months of PhD-level research in a single day, and unlike a lot of AI hype, the numbers actually check out.

FutureHouse's latest creation reads 1,500 papers, runs 42,000 lines of analysis code, and churns out discoveries that beta testers reckon would have taken them half a year to reach themselves. The system has already made seven legitimate scientific discoveries, including reproducing findings from unpublished papers and spotting novel links in Alzheimer's research. It achieved 79% accuracy and is being commercialised through a spinout called Edison Scientific. 

The algorithm uses structured world models which lets it maintain coherence across tens of millions of tokens without losing the plot. Every conclusion is traceable back to specific papers or code making it inherently auditable. Sounds like a modest boast, but this part is key given the black box nature of most advanced models. 

🧐 What's in it for me? Whilst there’s still doubt about the full potential of the platform in the broader scientific community, if you're a researcher drowning in literature reviews, this could compress months of drudgery into an afternoon. For everyone else, the implications are stranger: scientific breakthroughs might start arriving faster than we can validate them. 

💵 Out of the Lab: If AI scientists can genuinely compress research timelines by orders of magnitude, pharma companies will be queuing up with chequebooks.

It turns out there’s some science behind Dracula’s eternal youth.

Researchers studying Paracoccus sanguinis (a blood-dwelling microbe discovered in 2015) found it produces 12 compounds, 6 previously unknown to science. When tested on stressed human skin cells, 3 produced interesting results, reducing inflammation, preventing collagen breakdown and cutting cellular damage. Even though the cells were subjected to conditions designed to age them rapidly, they recovered nicely.

The fact these molecules are already circulating in your blood is particularly useful. Humans have spent a very long time not rejecting them which puts them several steps ahead of most skincare ingredients. Whether they'll work as well when applied topically or taken therapeutically remains to be seen, but at least they've passed the important test of not killing you.

🧐 What's in it for me? This is firmly in early-stage research territory so don't go full vampire just yet. But if these metabolites make it through trials we could see genuinely novel skincare treatments that work from the inside rather than being painted on your face. The real payoff is broader: understanding how blood microbes influence aging could unlock therapies far beyond skin health.

💵 Out of the Lab: The skincare industry is worth over $180B annually and desperately seeking innovation beyond incremental peptide formulations.

When researchers engineered mice with a tweaked liver protein and fed them the same soybean oil-rich diet as normal mice, the engineered mice stayed lean whilst their untampered with compatriots expanded rapidly. Interestingly, the culprit wasn't necessarily the oil itself but what happens when your liver breaks it down into compounds called oxylipins. More oxylipins = more fat.

The tricky bit is that oxylipins only show up in liver tissue, not blood, so standard health checks won't catch them until the damage is done, and the enzymes that create them vary by genetics, which explains why some people gain weight whilst others don’t.

The extraordinary thing is that soybean oil now makes up 10% of US calories, up from 2% a century ago and some estimates suggest 60% of processed foods now contain it. The researchers are also testing whether other vegetable oils trigger the same response, suggesting this might be less about soybeans and more about modern diets. Regardless, it looks like we have our first major scapegoat.

🧐 What's in it for me? If you're metabolically unlucky, soybean oil could be sneakily sabotaging you in ways that don't show up in routine tests. And in case you weren’t already full of bad news, other high-linoleic oils like corn and sunflower oils may trigger the same response. If, on the other hand, you are Robert F. Kennedy Jr., well done, hears some vindication.

💵 Out of the Lab: The processed food industry won't welcome this research, but alternative oil producers certainly will.

GPUs used to run AI require enough electricity to power a small town and generate enough heat to warm several more. However, as avid Toast readers you'll already know the most exciting potential solution: photonics.

And there’s a new breakthrough here, specifically in tensor operations (the mathematical heavy lifting behind AI). Think of tensors as a multidimensional container of information, like a box full of maths problems. Computers traditionally look at each calculation sequentially. Light can solve all of them at once.

A Finnish team encoded data into light waves then let physics do the work as the light propagates. No chips heating up, no power-hungry circuits, just light interfering with itself in mathematically useful ways. And they've now demonstrated it works for the operations AI actually needs like convolutions, with different wavelengths acting as separate computational channels.

Professor Zhipei Sun reckons this can be integrated onto photonic chips within three to five years and for an industry strangled by power bills and thermal limits, this is the sort of breakthrough that makes venture capitalists scramble for their chequebooks.

🧐 What's in it for me? AI could become dramatically faster and cheaper which could mean less money gets burnt using it, but more likely means there’ll just be a lot more AI. This could also unlock AI applications that are currently impossible due to power constraints. Data centres may start piloting photonic accelerators within a few years if the commercialisation holds up.

💵 Out of the Lab: Data centres are haemorrhaging cash on power bills, and photonics offers an escape hatch.

Scientists have discovered an ingenious way to get seagulls to sod off: yell at them. 

Researchers in Cornwall [no surprises there] played recordings of men shouting "No! Stay away! That's my food!" and compared the response to neutral speaking voices and robin song. Gulls exposed to shouting were significantly more likely to flinch, stop pecking, and fly off in a hurry. Those spoken couldn’t have cared less. 

Until next time, stay curious.

Like what you're reading? Share toast with a friend.