lacunary

just to be clear, it appears you're making up studies, results and explanations for results all in one go, is that right?

tuning defaults for performance feels much more acceptable than tuning defaults for correctness

this was my experience as well. "here's a trend, it's not true in these cases for reasons we won't explain." I only had two semesters and the second was much better than the first.

it is both because the "objective" system is also rife with subjective judgements

is true that if you've done some amount of mental effort in the day, learning becomes "pretty inefficient"? I could see that being true if you're exhausted, but then doing a workout would also be a problem.

presumably whatever the top model uses and then some, since the human can use the model.

I wonder if a model could score higher if it had a human at its disposal?

me too, but they apparently feel more comfortable, so I try to consider that

I agree for the most part. I have seen some non-native English speakers improve their short form English communication with an LLM, as a counter case.

so, the option to disable alerts is not present in the settings menu, despite appearances

any cool features you can share?

I feel great about my degree from my third tier college. I learned a lot I didn't expect to, got into a top tier grad school afterwards, and came away with no debt. Maybe me and the millions of other third tier college students are too dumb to know what's good for us, though.

I'm curious what you want to model; what's an example phenomenon you'd like to be able to demonstrate?

yes it's called "actually having democratic elections"

also for ones fully into AI

too new, trinoo, ping -f, synflood, teardrop!

If only we had some way of signing messages

delays between increasing capacity to produce and bringing production to market naturally lead to cycles of over and under production. https://thesystemsthinker.com/balancing-loops-with-delays/

where's a good irc chat these days?

The develop-test-refine feedback loop for this kind of attack is so long (or expensive) that it seems likely to limit its real world use. Poison training data, wait months? a year? for the model to come out, see how well it worked, refine... or do you see a faster way to iterate?

if you don't know much math, it's easy to confuse the two

"what you see is all there is." it's generally much easier to verify something you've been made aware of than it is to know of it in the first place (and still verify it.)

did cursor do model training? I thought it used models built by other companies

I'm having fun building elaborate software that meets my needs precisely and nobody else's. I mean, maybe it would meet other's needs but that production would take away from the fun and learning I have building it, and would also reduce its utility for me personally.

Flair Driven Development. What do you think about a project that only has 15 pieces of flair?

but claude -p is still Claude Code

ostrich, meet sand. I mean, have you tried CC? It's fun to build stuff with it. I think developers that don't use it (or something like it) will be out of a job, yes (unless they're very niche).

The Obscene Bird of Night by Donoso

that's because they - and their ancestors - experience physics all the time, not because they saw physics on TV

Going back to the late 90s I've been able to consistently buy books I want from my local B&N. I got most of the programming books I used to build my skills that led to my career there. My young adult kids buy books there now. They keep great hours and it's a pleasant experience. I recently went on a whim, at 8pm on a weekday, and found the best novel I've read in years. What am I missing?

if it worked this way, we could get good at golf by watching TV, writing songs by listening to the radio, or doing math by watching 3b1b. but it doesn't - we don't learn that way, for better or worse.