I'm tired

And at first my only reaction was, "That's an odd name," so I asked, "Oh, why did you call him that?"

And she said to me, "Oh, he's my seventh."

Let the Garment envelop you; experience its heft and splendor. The original Garment pattern was knit in the round, from the bottom up, and designed for someone six feet in height. A simple adjustment of the number of stitches you cast on will allow you to customize the length. Follow The Garment on Instagram: @HeftAndSplendor Update, Aug 9, 2026: Many thanks for all the Garment enthusiasm in recent days! In response to sensible feedback, I’m developing and testing two alternate versions of the pattern: one for children, and the other a top-down version. I’m not entirely sure why I chose to make the first The Garment bottom-up—it almost certainly makes more sense for sizing, etc., to make it top-down—although I will say that casting on 300 stitches for the bottom of a The Garment feels like quite a leap of faith into the great unknown. In any case, both new versions of the pattern will be included in the update and sent to everyone who’s already bought the original pattern.

The pattern for the Garment costs only five usd so if you are a knitter and you want to make something beautiful and baffling, there you go. I do almost want to learn to knit so I can make the Garment.

here's the bbc article linked,

facts to keep in mind: bacteriophage, genetic engineering, synthetic biology, synthetic virology- all scientific fields or research that have existed before now. the creation of artificial viruses and bacteriophage predates the implementation of machine learning. the concerns of this field are obviously about quarantine protocols, unforeseen developments that could lead to 'superbugs', and the risk of facilitating germ warfare or eugenic gene tampering. not inherently new concerns but ones that feel more risky in our current political climate and pandemic era.

also keep in mind the custom data analysis and DNA generation software is a joint work between the biomedical research institute and the tech provided by NVIDIA. so the research is real and heavily dependant on the tech. so of course it is going to be talked up in their official research press.

The specific models used as cited in the bbc article are (according to the below article) tailor made to DNA analysis, not generic language models. the technology at play here is neural networks applied to pattern recognition/generation within the context of only DNA datasets, and some programming to make the interface and visualisations of the output visually comprehensible.

"a biological foundation model trained on 9.3 trillion DNA base pairs from a highly curated genomic atlas spanning all domains of life. [...] . Evo 2 learns from DNA sequence alone to accurately predict the functional impacts of genetic variation—[...] without task-specific finetuning. [...] . We make Evo 2 fully open, including model parameters, training code, inference code, and the OpenGenome2 dataset, to accelerate the exploration and design of biological complexity."

"we introduce systems and algorithms for convolutional multi-hybrids, a new class of architecture for sequence modeling at scale. We discuss architecture design, block layout, kernels for fast convolutions on GPUs based on overlap-add schemes, and context parallelism strategies. Multi-hybrids excel at efficient modeling of byte and character-level data, and we expect their utilization to unlock new applications for foundation models. Effectiveness of StripedHyena 2 is verified at scale (40 billion parameter, over 9 trillion tokens and 1 million context) with the Evo 2 line of models"

TO THE BEST OF MY LAY-PERSON UNDERSTANDING: the two linked articles here cover 1: how the Evo models function 2: how the neural networked programming is used to create these models

BACK TO THE BBC ARTICLE:

The Stanford researchers picked the most promising 302 [...] designs and synthesised them in the lab. Of these, 16 proved effective at killing E. coli bacteria.

we are not given information of 1: how many designs have been created without ai prior to this 2: if any previous attempts at bacteriophages have been successful by this team 3: how many other iterations than the "most promising" 302 were generated and how long that process took 4: how long synthesising them takes

regardless, 16 out of 302 is a reasonably impressive success rate I think

who is speaking here: Brian Hie- assistant professor at Stanford University- and Samuel King- a PhD student in the lab.

The phage were placed on petri dishes growing a layer of bacteria on them and the scientists waited for signs their new viruses were enjoying the feast. "We were starting to see these clear spots and it was just extremely exciting," says King. When the results were shared with the wider team "the room spontaneously burst into applause", Hie recalls.

so they created E. coli consuming phage

The researchers themselves took steps to maximise safety. They excluded viruses that could infect complex organisms from the training database, performed the research on phage rather than viruses that infect people and it all took place in a secure laboratory.

obviously lab safety is expected but additionally this article clarifies that they were training on limited datasets to reduce the inclusion of anything known to infect complex organisms- which i don't think excludes the possibility of the resulting phage adapting or mutating/changing to be more virulent/pathogenic but is what I would consider a reasonable caution to significantly reduces the risk.

note: this phage is interchangeably called a virus. a virus is not a living organism by the definitions provided

The genetic code of the phage is around 5,400 base pairs (letters) long. The smallest genome of a living cell is around 500,000 base pairs. The human genome is three billion base pairs.

they emphasise the simplicity of the resulting phage

Prof Marc Güell, from the synthetic biology lab at Pompeu Fabra University in Spain, said the study was a "very significant turning point" because for the "first time in history, we are beginning to design biology on a computer".

this, i think, is the real crux of the matter

not that the fabrication of biology is possible (that's been known for a while) but that it can be designed using a tool that is as fast as not just a computer but a computer using machine learning processes.

this isn't some automated process. the software is custom, the training input is specific, the output is completely scrutinised before selection for synthesising, and the biological materials used are also highly controlled and specifically selected.

as usual a lot of conflation of language and image generation models with the entire field of machine learning, I don't think it's particularly useful to compare technology being used to purpose with the broad misapplication of commercial models in this case.

I just wanted to share some general context here, I could be inaccurate in my interpretation of information... but my general disgust with the unethical creation and implementation of so called AI doesn't mean I have to buy into fear mongering that doubles as myth making in how machine learning is discussed in media.