Hello, friends.

For decades, engineers have been building artificial neurons: tiny electronic circuits designed to fire electrical pulses the way real brain cells do. They've gotten rather good at it, producing convincing signals on the lab bench while almost never having to prove themselves against an actual living neuron. A bit like spending your life claiming to speak French without ever having to meet a French person.

Now though, a team at Northwestern has run the proper test, setting up the rendezvous with the help of graphene, molybdenum disulfide and electronic ink. The result is a bendy, low-cost artificial neuron whose signals arrive at the right speed and shape to talk to actual brain cells. When applied to slices of mouse brain, the printed devices reliably triggered living neurons to fire. Not imitation, but actual conversation.

Brain implants are the obvious application, but the ambition runs deeper. The brain is roughly five orders of magnitude more energy-efficient than a digital computer which becomes a fact worth taking note of whilst your AI data centre is ordering its own nuclear reactor.

🧐 What's in it for me? Two directions worth watching. In the near term, neuroprosthetics that can genuinely communicate with the nervous system, which matters for restoring hearing, vision and movement. Longer term, brain-inspired hardware that doesn't require a Greenland’s worth of power. Given where AI energy demand is heading, the second is less a nice-to-have than physics slowly becoming policy.

💵 Out of the Lab: Neuromorphic computing has drifted in and out of fashion for thirty years, but the AI power crunch is finally giving it a commercial reason to exist. A printable, biocompatible neuron is also conveniently well positioned for medical-device spinouts. Worth noting: Northwestern's Mark Hersam, who led the work, has previously founded two companies (NanoIntegris and Volexion), so commercial translation is not unfamiliar territory.

The body has several ways of killing off its own cells when they go wrong. One of the grimmer options is necroptosis, a form of programmed self-destruction triggered by a protein called MLKL. Researchers therefore tended to treat MLKL as a villain: its day job was executing cells, so the only question worth asking was whether any had recently been dispatched. If not, great. 

However, researchers then decided to ask something different. What if MLKL is doing something nasty short of actual murder? Their answer is, unfortunately, a big yes. 

Under stress, MLKL briefly visits the mitochondria (the cell's power stations) and causes low-grade damage. No death, no obvious distress, just weaker energy production, impaired self-renewal, and the slow drift toward the tired immune system you tend to find in older bodies. When researchers knocked MLKL out in mice, the stem cells aged dramatically more gracefully.

🧐 What's in it for me? The immune decline that comes with age is one of the reasons older people fare worse with infections, cancers, etc. If MLKL is genuinely driving part of that process, it becomes a drug target and the most immediate beneficiaries would be patients recovering from things like chemo, radiation, or bone-marrow transplants, where stem cell function is the bottleneck.

💵 Out of the Lab: The broader thesis, that "cell-death" proteins do quieter, slower harm long before they kill anything, may reshape how ageing drugs get designed. Necroptosis modulators were already a biotech category. They may now be a longevity one too.

Quantum computing has spent the last decade doing a passable impression of nuclear fusion: undeniably exciting, perpetually twenty years away, and curiously unhelpful with whatever it was supposed to be solving. The industry's problem has always been the same. The hardware is noisy, the algorithms are fragile, and running a full quantum simulation is a bit like insisting on a Formula 1 car for the school run.

Enter a team from UCL who whilst toying around with a 20-qubit IQM machine, found what may be a fantastically pragmatic workaround. Rather than run an entire simulation on quantum hardware, they use the quantum computer once to extract the stable statistical patterns lurking in whatever they’re modelling, and then feed those into the training of an ordinary AI model running on a classical supercomputer.

The result, published in Science Advances: 20% more accurate than conventional AI, stable over much longer windows, and using hundreds of times less memory. Less quantum supremacy, more the next potential leap in intelligence that we’re all now somehow expecting. 

🧐 What's in it for me? This is the first convincing glimpse of what commercially useful quantum computing actually looks like. If that pattern holds, it changes quantum's economics entirely: instead of needing a million-qubit machine to do anything useful, you may just need modest quantum access stitched into the AI stack people already have.

💵 Out of the Lab: The convergence of quantum and AI is probably the most underpriced frontier-tech story going. The winners will be the providers who make quantum-as-a-co-processor trivial to plug into existing classical infrastructure.

It’s nice to think that in the grand question of how long we’ll live, nurture will have an outsized influence over nature. And historically the science has been nice, suggesting that so long as we eat our greens and only do the things that kill us in moderation, our destiny is ours to dictate. 

Well, bad news. A new study from the Weizmann Institute suggests nature has been quietly underestimated, from a widely agreed on 20% influence to a much more disconcerting 50%.

The authors reanalysed large twin datasets from Sweden and Denmark, including, crucially, twins raised apart. Earlier estimates had been skewed by "extrinsic mortality" (deaths from accidents, infections, and other generic misfortunes) which obscured the underlying genetic signal. Filter those out with some careful maths and the genetic contribution roughly doubles.

For those pursuing immortality via green smoothies, this is moderately inconvenient. For longevity researchers, it's excellent news: if genes matter this much, there must be variants to hunt down. Dementia risk, the study notes, turns out to be roughly 70% heritable up to age 80. Time to put the kale down and invest in Biotech. 

Until next time.

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