> An employer is looking to screen two recent harvard grads by GPA, not really between a new grad and a 5/10 year ago grad.

that's a really good point, actually. in every situation i can think of where someone is looking at your grade (always admission to the next step in the ladder, in whatever form), you are being compared to people "from the same time" as you.

and i'd like to reiterate how difficult it would be to have a "stable" standard of mastery, no matter how nice. technical fields change a lot, and fast, these days. all across STEM, in 20 years everything changes. everything's so niche, as well, sometimes it may be hard to compare two degrees with the same name of different institutions. maybe we could do it with the fundamentals (mathematics and physics)? but look at a textbook from 100 years ago (say, Whittaker and Watson) and you'll find that even this changes. and even if the field doesn't change, the world does: i'm imagining how old-timers could claim that in their time information wasn't so easily accessible.

> I've always wondered what the steelman is for curve grading.

assuming that by "steelman" here you mean "the justification", i believe the point is that a curved grade shows how you compare to others. the idea is that "getting 40% of the answers right" is meaningless if you don't know how hard the test is, so you'd rather have a grade that says "top 5% of the class".

this what i see as the justification, at least. not an endorsement of the idea

> Or 100% Fs if you want to retire from teaching immediately

it's crazy to see that mentioned so non-chalantly. my expectation is that the teacher, when they grade, is meant to be impartial, as if they were doing nothing more than taking a measurement of the student's work, you could say (this is why, i believe, we value standardized tests in some settings, even though they are worse in other aspects). it's the student who is responsible for the grade. a teacher not being allowed to give F's to everyone suggests a corruption of the system to me.

can you share more? what pressures teachers not to do this, for example?

Same argument about distributions cuts both ways to me imo. Like you taught 400 students and you couldn't get a single one over the line? I think the immediate suspicion would fall on the common factor of the teacher. For a crazy event like that it seems much more likely that the instruction or assessment was flawed.

Mostly though I was just chuckling in my head about an old curmudgeonly professor of mine who was literally retiring at the end of that semester. We were all actually quite scared he would go out with a bang given his disdain for us. We weren't like 100% Fs or anything, but I think it was obvious we thought the material was pretty phooey and the lectures interminable.

> Yes, in positivist sciences 20% intending to stay would be very high by historical standards.

i'd be interested in a source for this. i did not find in the article you cite mention of historical trends.

> When folks come home from a hard day of work they don't want HBO.

that's hilarious, since today, as many other days, i am literally looking forward to watching HBO when i get home from work

somewhere else they were discussing how to use a 555 to time 55 years, and how for such a long period you'd need impractical resistance and capacitance values. easy workaround would be to set a more reasonable period, say, 1 sec, and use a counter to know when you hit 55 years. coincidentally, 55 years is 2 ** 30.7 seconds, so it'd just fit in a 32 bit register.

though i take you were thinking about counting clock cycles or something in which case surely your register would overflow

The size of a register is not the largest value you can conveniently count on a computer. You can use multiple registers.

Old computers often had a "carry flag" specifically to make this easier e.g.on Arm:

    add r0,r0,#1
    adc r1,r1,#0

But even on RISC-V, often criticised for not having a carry flag, it's not hard:

    addi  a0,a0,1
    sltiu t1,a0,1 # set to 1 if a0 wrapped back to 0
    add   a1,a1,t1

yes, from the point of view of an individual programmer, there is no difference between code created by AI or by some other programmer.

but from the point of view of the company as a whole there is a difference: code created by some programmer is understood by that programmer, while code created by AI is understood by nobody

> code created by some programmer is understood by that programmer

was at some point understood by that programmer, at least for a few months, before they moved on to something else and forgot important context and detail, or left the company

some people would, in this case, prefer the unrefined version

that would slow down the process considerably. it would also not be of much use to the professionals, which i guess make up the majority of those involved most of the time, and so, i guess, would not have much support.

IMO a good middle ground could be attained by everyone having some understanding of the legal system. we could use school for that. i mean, we cover calculus and ancient history, it's not like covering law to some extent would be harder

> Patent texts read as prose, but are actually precisely structured legal documents.

at that point why not just use something precise like a programming language? have there been efforts in that direction? genuine questions

I have no idea.

A few months ago, for the first time in my life, I had to write a patent document. It was very complicated – too complicated. Noting the structure, I searched for tools, but found only LLMs. So I wrote my own tool.

The amusing thing is, LLMs prefer the DSL-structured document!

Yes (not patents in particular).

The name is "law as code".

(My impression: various approaches; mostly academic; some small companies in the space; judging from a loose assessment wrt my career choices as a freelancer: no real business opportunity yet)