Scientist at IDM ♥s epidemiology, running, prob stats, 🦀/🐫, fly fishing, greyhounds, backpacking. Author of book Bioinformatics Data Skills. All views my own.
Happy to share the alpha release of camdl, a new compartmental modeling framework I've been cooking up.
With coding agents writing more of our scientific code, I think it's more important than ever to shift to architectures that engineer in robustness...
vincebuffalo.com/blog/introd…
Interesting. AI will in effect increase both supply and demand for formal methods. You need them more, but you also have tools that make them cheaper.
blog.janestreet.com/formal-m…
I enjoy using IGV, but I need a more efficient tool for analyzing diploid genomes. So, I developed one for myself. It provides exceptional GPU acceleration, allowing me to uncover all essential information spanning six orders of magnitude. #bioinformatics#genomics#visualization
This is part of a larger trend of generative tools paired with verification that is so exciting right now. One of the things that’s blossoming under agentic coding. Types FTW!
I've been telling people for 25 years that Jane Street is not interested in formal methods.
No more!
And we're actively hiring to form a new formal methods team!
I’m annoyed by excessive biosecurity filters too, but people forget: risk = probability × severity.
For catastrophic tail risks, the prudent default is oversensitive filters that are audited and tuned down, not weak filters that need tuning up after dangerous false negatives.
Over lunch today I was asking Claude to help me feel less disillusioned by the direction of modern biology, and it gave me a well-reasoned reply that did precisely the opposite.
Claude nails it here:
“measurement technologies generate their own research programs. Once you can measure something at single-cell, spatial, genome-wide resolution, the field reorganizes around producing and indexing that measurement… ‘we mapped it’ gets published as though mapping were explaining.”
I got curious about the real story of Ozempic Gila monster spit people cite when advocating for basic research funding. The truth is more interesting, and shows us more about the stories we tell ourselves about science than it convinces people to maintain the funding status quo
Each time we release a model, we run the same test: give it code that trains a small AI model, ask the new model to speed it up. It takes a skilled human 4-8 hours to reach 4x faster.
In May 2024, Claude Opus 4 averaged a ~3x speedup. This April, Mythos Preview achieved ~52x.
Sam Altman, Dario Amodei, Demis Hassabis and many others have signed a letter urging Congress to increase security on orders of synthetic nucleic acids - and the equipment needed to make them - as models continue to become increasingly bio-capable.