🍼 Born in the USA? Your baby probably hasn't eaten real food

Remember when baby food was just, you know, mushed-up food? Apparently, the US decided that was too simple. This week, we learned that nearly three-quarters of American baby foods are ultra-processed, scientists reconstructed what 3,500-year-old Egyptian embalming balm smelled like (and put it in museums), and researchers at Cambridge created an RNA molecule so small it almost solved the origin of life. Also: Illumina is building a five-billion-cell atlas with pharma giants, and a Helsinki startup just raised $5.7M to fix how cells ship proteins. Let's get into it.

If you've ever fed a baby in the United States, there's a 71% chance you were serving them something that qualifies as ultra-processed food. That's not hyperbole. That's the finding from a new study published in Nutrients that analyzed 651 infant and toddler food products from America's top grocery chains.

Researchers at The George Institute for Global Health used the NOVA classification system (the gold standard for categorizing food processing) and found that the same 71% of products also contained additives. Over 105 unique additives showed up across the products, with flavor enhancers in 36% and thickeners in 29%. The ultra-processed options pack nearly double the sugar of their less processed counterparts (14.0g per 100g versus 7.3g).

Snack-size packages were the worst offenders at 94% ultra-processed. Those convenient squeeze pouches everyone loves? 73% ultra-processed, and their sales have grown roughly 900% since 2010.

The implications are still being studied, but emerging research links certain additives to gut microbiome disruption and behavioral effects in children. Lead author Elizabeth Dunford noted that infants and toddlers are particularly vulnerable because their systems are still developing.

The Consumer Brands Association pushed back, arguing there's no agreed scientific definition of ultra-processed foods. Which is technically true in the same way there's no agreed definition of "a lot of sugar," but you know it when your toddler is bouncing off the walls.

Here's a sentence you probably didn't expect to read today: scientists have recreated the smell of 3,500-year-old Egyptian embalming balm, and you can sniff it at a museum.

A team of researchers from institutions including the Max Planck Institute of Geoanthropology and Museum August Kestner just published a framework in Frontiers in Environmental Archaeology for how museums can translate biomolecular archaeological data into actual visitor experiences. Their proof of concept? The canopic jars of Lady Senetnay, wet nurse to Pharaoh Amenhotep II, were excavated by Howard Carter in 1900 from Tomb KV42 in the Valley of the Kings.

The original 2023 analysis identified the balm's ingredients: beeswax, plant oils, Pinaceae resins (probably larch), bitumen, and dammar or Pistacia tree resin. Researchers described it as the richest, most complex balm identified from this period. Perfumer Carole Calvez then recreated the scent using modern equivalents, naming it "The Scent of the Afterlife."

Museum August Kestner in Hannover now uses scent-printed cards during guided tours, while Denmark's Moesgaard Museum installed a permanent scent station. Curators report the experience transforms visitor understanding, moving conversations away from horror movie clichés about mummies toward appreciation of ancient funerary practices.

Turns out the afterlife smelled surprisingly pleasant. Resinous, earthy, a little sweet. Who knew.

For decades, origin-of-life researchers have chased a holy grail: an RNA molecule that could copy itself, proving that life could have bootstrapped from chemistry alone. A team at Cambridge just got tantalizingly close.

Published in Science, the study introduces QT45 (short for "Quite Tiny 45"), a 45-nucleotide RNA polymerase ribozyme. That's an RNA molecule that can catalyze the synthesis of other RNA molecules. Previous ribozymes capable of this were over 150 nucleotides long, making their spontaneous emergence from prebiotic soup implausible. QT45 is one-third the size.

Here's what it can do: in eutectic ice at -7°C, QT45 can build copies of its complementary strand using trinucleotide building blocks. It can also use that complementary strand to rebuild copies of itself. Both reactions work with about 94% fidelity per nucleotide.

Here's what it can't do (yet): perform both reactions in sequence without human intervention. Lead author Edoardo Gianni told Chemistry World that what they've achieved is "self-synthesis," not true self-replication. The two reactions require different conditions and don't yet happen in one continuous cycle.

Still, this is a major milestone. As Jack Szostak of the University of Chicago noted, if repeated replication cycles are achieved, "RNA could leap from being a possibility to a probable start to life." The fact that QT45 represents an entirely new RNA lineage (not derived from the class I ligase ribozyme that's dominated the field for 30 years) makes it even more significant.

We're not quite at life-from-scratch territory. But we're closer than we've ever been.

Illumina just announced it's building the world's largest genome-wide genetic perturbation dataset. The name is appropriately ambitious: the Billion Cell Atlas.

Unveiled at the J.P. Morgan Healthcare Conference, the project aims to map five billion cells over three years, systematically documenting how cells respond to CRISPR-mediated activation and inactivation of roughly 20,000 genes across more than 200 disease-relevant cell lines. About 150 million cells have been sequenced so far, with a billion expected by the end of 2026.

The scale is staggering. Illumina says it will generate data at a rate of 20 petabytes annually, processed through its DRAGEN pipeline and Single Cell 3' RNA prep platform (essentially a microfluidics-free workflow that can handle millions of cells per experiment).

Three pharma giants have already signed on as founding participants: AstraZeneca, Eli Lilly, and Merck. Merck's Iya Khalil, VP of Data, AI & Genome Sciences, said they're using the atlas to train AI foundation models and build virtual cell simulations. CEO Jacob Thaysen told STAT News that "several others have expressed interest," including technology companies.

The project lives within BioInsight, Illumina's new business unit launched in October 2025, specifically for data-driven discovery. Whether this becomes the Google Maps of cellular biology or an expensive experiment remains to be seen. But if you're wondering where pharma's AI drug discovery ambitions are heading, this is a pretty clear signal.

Here's a problem you probably didn't know existed: when cells manufacture therapeutic proteins, those proteins need molecular "address labels" called signal peptides to reach the right destination. The current industry approach to optimizing these signal peptides is basically trial and error, one peptide at a time.

Avenue Biosciences, with its split presence in Helsinki and Palo Alto, a little since startup spun out of the University of Helsinki in 2023, just raised $5.7 million to change that. The seed extension was co-led by Balnord (a Baltic/Nordic deep tech VC) and Tesi (Finnish Industry Investment), with participation from Voima Ventures, Inventure, University of Helsinki, and Dimerent. Total raised since 2024: $8.7 million.

The company maintains a library of roughly 5,000-6,000 signal peptides and uses high-throughput wet-lab screening combined with machine learning to identify optimal signal peptide-protein combinations. Instead of testing one variant at a time, they can generate thousands in a single experiment.

CEO Tero-Pekka Alastalo (who co-founded the company alongside COO Katja Rosti, Research Director Ville Paavilainen, and CSO Juho Kellosalo) describes the secretory pathway as "one of the remaining black boxes in therapeutic protein production." The signal peptide gets cleaved from the final protein, so the therapeutic itself remains unchanged. This matters for biosimilar comparability.

Applications span antibodies, vaccines, AI-designed proteins, gene therapies, and mRNA therapeutics. Avenue claims to be building the first platform specifically modulating the secretory pathway for therapeutic protein manufacturing. Given how much money is flowing into protein engineering right now, they won't be alone for long.

So there you have it. American babies are eating more additives than actual food, you can now smell what Egyptian embalmers smelled 3,500 years ago, and RNA is getting suspiciously close to figuring out how to copy itself. Meanwhile, Illumina is building a cellular Google Maps with pharma's biggest names, and a Finnish startup is teaching proteins how to find their way home.

Got thoughts on ultra-processed baby food? Strong opinions about ancient scent reconstruction? Want to argue about the semantics of self-synthesis versus self-replication? Hit reply. We read everything.

Forward this to someone who still thinks baby food is just mashed peas.

Keep questioning everything (especially ingredient labels), Prateek & Jere

Prateek & Jere