Cat Hicks

yeah, obviously the e.g. neuro world of "AI" is far bigger than LLMs, but everyone I know is using agents every day for analysis, custom searching, vetting candidate approaches rapidly You can read about the claude for science partnerships directly, with examples www.anthropic…

Claude Science, an AI workbench for scientists anthropic.com

You could take a look at this as a replacement mitpress.mit.edu/978026205248...

mitpress.mit.edu

I won't show you all the demographic back and forth, but I'll show you just how easy it is to make an AI-administered survey good, even with the stupid AI tone ("honest" lol), you can dramatically improve LLM output with a good harness Here are inventory instructions: github.c…

informed-patient/informed-patient/skills/informed-patient/references/symptom-inventory-methodology.md at main · DrCatHicks/informed-patient github.com

haven't used but I think things like this are directionally neat and in the orient subskill contributed by @mcmullarkey.bsky.social to learning-opportunities we did a quick pass at imagining repo context for comprehension goals! github.com/DrCatHicks/l...

learning-opportunities/orient/skills/orient/SKILL.md at main · DrCatHicks/learning-opportunities github.com

I haven't gone and verified any of this myself because tbh, waste of my limited time imho. But this is a pretty fascinating project I remember reading sites.google.com/view/elizaar...

ELIZA Archaeology - 3 Weizenbaum's Secretary sites.google.com

preprint: arxiv.org/pdf/2508.02740

arxiv.org

Ok apparently there are papers! I dug up one arxiv.org/abs/2410.06351

Moving Faster and Reducing Risk: Using LLMs in Release Deployment arxiv.org

I think what's sort of funny about this stuff is how everyone writes about software & dopamine & AI writes "I have discovered a thing" when it is extremely recycling so like this piece I saw going around: pydantic.dev/articles/the... (lol @ "That loss is real and it's worth na…

The Human-in-the-Loop is Tired pydantic.dev

I think this is the paper? it's a preprint osf.io/preprints/ps... The questions are media trivia questions, which aren't exactly judgment questions. Presenting "AI" that's forced to be incorrect basically just tests whether participants trust it to be a lookup tool which I thi…

OSF osf.io

I would like the learning scientists and assessment psychologists outside of tech to please look at this dl.acm.org/doi/10.1145/...

dl.acm.org

Just finished the CTO Hour with O'Reilly on eng metrics and agentic dev teams. I'll summarize a few high-level takeaways www.oreilly.com/live-events/...

CTO Hour with Peter Bell: Engineering Metrics for Agentic Development Teams oreilly.com

I suspect once we figure it out agentic work in software will actually be a huge aid to codebase comprehension This is such a wildly unpopular thought right now but you know what I'm going to pre-register my belief in this possibility

"Scientists who use AI are morons who aren't reading" I'm a computational social scientist and I feel like LLMs are a lot of things, but one thing they can be is a computational tool for text-based information in science. Here is what I am doing this morning

Can I just say it's disorienting to see people who like, roundly and vociferously mocked the idea that AI could have utility now immediately float into AI leadership and thought leadership positions. When I open accessed my AI Skill Threat study I had to field a thousand skeptical biting comments

A punitive view of education will never have the answer to what to do when students can use AI. A view of education that cares about motivating students will. In a world where student motivation needs to be earned not just enforced it's pretty clear that's the only way to elicit effort.

I've been thinking about my previous life in data science/applied computational social science and how many people have always been cut off from that. I hear a lot of examples of "I can't believe I can do this" from friends about LLMs that are along the lines of finally being able to look at data

In pseudoscience folk theories about AI and cognition I'm kind of surprised we're just limited to labeling all human behavior as "psychosis", we could also have "AI depersonalization," "AI derealization", "AI folie à deux" "AI alexithymia" like come on where is the theoretical richness

I think I would believe everybody's complaints about AI taking over writing more if people actually went out of their way to engage in and share and make visible complex and interesting writing in tech

There is this interesting problem in the AI design space (including critics' stances about what Rules should be enforced) where "better model ethics" just seems to mean default to authority in all hard situations. Little allowance for when authority does not provide a solution. Or even harms

Millions of doctors are using OpenEvidence (65% of US doctors according to an article from May) without their patients ever knowing it but patients in so many desperate circumstances are supposed to beg forgiveness for ever asking AI a health question

As long as I'm sitting here with my coffee, ANOTHER THING that pmo is the assumption that the impact of AI on real world technical work can be modeled as independent tool impacts while it is absolutely freaking obvious that an individual developer builds skills across tools and over time