Mikhail Kats

@mickeykats.bsky.social

Optics/photonics researcher, applied physicist, and faculty at the University of Wisconsin - Madison. Personal account + opinions Was @mickeykats on Twitter

I'm confused about how claude handles context windows. Basically the window fills up until you have to compact It doesn't seem like it would be hard to continuously monitor the context window and remove stuff that is definitely not important But they don't seem to do that?

With AI tools I now routinely end up on tangents where I spend 2-3 hours trying some idea, doing some calculations, etc., that I would not have spent time pursuing otherwise Can't tell yet if it's more of a distraction, or if it really expands the range of things I look at in a good way

I must say, "Free [researchers] from the growing administrative burdens that now weigh them down for nearly half their working hours" is kind of funny to read This past year is the first year in my career, where admin burdens (including and especially proposal/funding things) took ~50% of my time

Dan Garisto@dangaristo.bsky.social · 2w ago

Initial notes: There's a lot of emphasis on the GRFP (and similar funding mechanisms, such as CAREER awards) because of its ability to target high-performing individuals on the basis of merit—in contrast to institutions and diversity programs.

Refocus on the Individual Scientist: Put the working researcher back at
the center of America’s scientific enterprise. Free them from the growing
administrative burdens that now weigh them down for nearly half their
working hours. Bet on people, not just projects, by expanding portable
graduate fellowships like the National Science Foundation (NSF) Graduate Research Fellowship Program (GRFP), backing early-career independence, and scaling long-horizon grants for the best and brightest modeled
on National Institutes of Health (NIH) Director’s Pioneer Award. Open
alternative pathways beyond standard academia, and ensure that selection
rests purely on merit, not the political fashions of the day

I wonder if the Bluesky team knows that the experience for a large fraction of people is no meaningful engagement, leading to many people abandoning it I believe in the bluesky mission, so I've been posting here for 3 years (4K posts, >1K followers), and my median post gets no replies and ~1 'like'

I have no clue how students who have decently smart LLM tools available when they are in high school / college will ever learn to have their own writing voice Seems like a steep hill

When people leave the group I often don't take off their recurring meetings from the calendar for sometimes up to a year. Both because it feels weird and "final" to do it, but also because it helps buffer my calendar so I have some enforced gaps