Digvijay

CDC's wastewater surveillance detects pathogens 2 weeks after spillovers. Yet everyone in biosecurity is building faster, cheaper, more sensitive tests. They're solving the wrong problem. The bottleneck isn't detection. It's sampling & concentration (& incentive alignment). (1/n)

I like @Olivia's response cus she is specific about taking threat models to a conclusion (which therefore teases out what exactly should be built).

Covid shed massively in feces (high-signal vector). So we thought wastewater surveillance is great. What of threats that shed very little? Fundamentally, early warning = low signal detection. Any epistemic built on high-signal threats has little power to predict or contain low-signal ones. AI included. No detection or reading sophistication downstream compensates for signal that just physically wasn't captured. (*our datasets are fundamentally biased.)

There's a streetlight effect in what we even call important pathogens!

@owl_posting (*for some biosecurity business models)

@owl_posting Great article! I made a short case for including ag as a threat model here (& why sampling is the invisible problem): x.com/S_Digvj/status… My pushback (esp in ag) is that biosecurity can emerge as a side-effect of solving on farm problems : diggest.substack.com/p/first-princi…

Reasons to be pessimistic (and optimistic) on the future of biosecurity owlposting.com/p/reasons-to-b… "It was such a fun read (if you can say that about an article on weapons)!" —a glowing review from an early reader this is (once again) the longest article I have ever published at 13,000 words. it involves interviews with 16+ researchers/VC's/policy folks in this field, and discusses basically every single facet of biosecurity that i could find. topics include: how machine-learning in rapid response therapeutic design may work, the financial status of the customer base of biosecurity startups, why agroterrorism feels extremely likely to me, and a lot more i admittedly started the essay pessimistic that this subject matters at all, and i end it surprised that it doesn't keep more people awake at night. im not a doomer about it all, but i can see how people become one. very grateful to the people who decide to spend their career (or some fraction of it) working here, and especially grateful to the ones who helped teach me about the subject

We're creating batch level microbial food supply chain datasets that've never been possible before!

You can just do things (genetically sequence your dog’s tumors and design a bespoke mRNA cancer vaccine to save her life)

The other problem is pattern-matching/innovation motifs. I've seen a lot of cool tech that's not aligned with pop-sci/VC get lost on the bench. So apart from capital stack, I'd argue attention/comms stack

*wastewater surveillance work they've been doing the past 5 years!

Literally got off a call where a group in Pennsylvania recited having the same problem to us unprompted

@McDonalds @YourTaylorFarms @CDCgov I think @mkoeris & @DARPA had correctly flagged this in last year's Ag x BTO! Hope to see more!

Food supply chain is an efficient, proven biological distribution networks in the US. In @McDonalds 's E. coli outbreak, @YourTaylorFarms tested & found nothing. @CDCgov traced it a month later. A contaminated batch in Salinas reached 48 states in 72 hours. No HHT reqd. (2/7)

Metagenomics + Frontier AI = Protection against new threats? Not quite, imo. Wrote a response to @Simon__Grimm 's fantastic @IFP essay on scaling pathogen detection. The architecture is right. The input layer is missing/non-trivial. (1/7)

@McDonalds @YourTaylorFarms @CDCgov I hadn't considered this & learnt about this from @lennijusten at this week's @CSRisks event earlier this week!

@McDonalds @YourTaylorFarms @CDCgov Also, the economics are harder than they look. Sequencing costs are low per-sample cus labs wait & batch samples to fill expensive flowcells. i.e. Real-time = low utilization = high per-sample cost. To contain spread has to be solved. More samples from farms maybe an idea? (5/7)

*Ebinger et al.

@McDonalds @YourTaylorFarms @CDCgov The best biosurveillance systems won't be the ones with the most sensitive sequencers. They'll be the ones that see the most of the world. diggest.substack.com/p/metagenomics…

@McDonalds @YourTaylorFarms @CDCgov To include ag, timing would be interesting cus of the January FSMA traceability rule enforcement & recent H5N1 outbreak with confirmed human spillover. If sequencing+AI infra gets built, ag samples could feed into it at marginal cost. But it needs to include sampling layer. (6/7)