So yeah your AI can outperform a doctor that gets 5 minutes with the patient before having to move on to the next one in order to keep their private equity Masters satisfied.

So, suppose, we stick it to the "private equity Masters", compel them to double the number of doctors — forget for a second, who is going to pay for them — and afford them a whopping 10 minutes with the patient.

ChatGPT will still beat humans... And it will be getting better with every month, whereas the humans will not...

A new study from Harvard Medical School and Beth Israel Deaconess found that an OpenAI reasoning model outperformed experienced ER doctors at diagnosing and managing patient cases

AI is sufficiently anthropomorphic to be capable of making mistakes. Demanding perfection from it is stupid. It does not need to be error-free. It just needs to be better than humans...

The PGP encryption source code was printed in a loose-leaf book with checksums on each line to make it easy to OCR.

It was still a huge project because they forgot to convert tabs to spaces (or vice-versa) before printing so software had to be written to try all possible combinations of tabs and spaces on lines where the checksum check failed.

For the Apollo Guidance Computer code they got lucky and had a binary dump of the compiled executable at the end of the listing so they could run the OCR-ed code through the compiler and check for mismatches in the compiled binary to find the OCR errors.

It's definitely non-trivial and can be even if the developers went out of their way to try to make it easy.

Oh wow. So are you lined up to be a long haul driver then? Surely you are manly enough. We have a huge shortage of drivers right now. Maybe you should talk to a trucker a little bit. The job they do is important and they are proud of what they do. But they are treated horribly.

Yes of course we're talking about BEV trucks. What does that have to do with reefer trailers?

What are you talking about? All the major manufacturers are currently selling electric trucks of that size and range in Europe. There's a guy documenting daily long haul driving in Europe with electric trucks. Google for electric trucker or elektrotrucker.

And to head off the inevitable comments, yes European trucks are as big or bigger than American ones. And yes the distances driven are just as long as American routes. Infrastructure for changing is much better than in the US of course, and improving.

Exactly. There are many different learning styles. (One issue I had with school is I don't learn like most people - several learning disabilities/issues - so I'm aware of the differences and how things need to be altered, but for most students, yes, repetition within that first 24 hours is important.)

I use GitHub as a simple public-facing repository, that part is pretty solid. I do all the work, including merging and actions, locally. I simply don't use GitHub's pull request mechanism. I did discover that if someone does submit a pull request, I can do the pull and merge locally and when I push back to my GitHub repo, GitHub automatically detects the pull request was pulled and closes it without using any of the web interface.

I haven't been in the classroom since close to the year 2000, so I don't remember the study names. What I do remember is that there were studies, plural - studies, that showed that when you learn how to do a new task or learn new information, that using that information or practicing the task within 24 hours increases the chance of it being remembered by a large percentage. That's over 25 years ago for me, and I'm not going to claim it's at a certain percentage, but I know it was WELL over 50%. So if you learn a new process in Algebra, or a new move in ballroom dance, and you don't practice it within 24 hours, you have a lower chance of remembering it. But it was at least over a 50% increase in your chance of remembering it IF you reinforced it by going through it within 24 hours.

I preferred to use homework as practice - not as learning new material (although that might help if it includes reading for the next day's class). I also worked in psych treatment, which meant I taught more than one subject - I had the odd mix of science and math plus English (lit and grammar). So I'd assign reading overnight that gave us more chance for discussion (discussion, not lecture!), and the math I assigned was to use what we had learned in class. For science, I'd actually prefer to assign reading for what we had done that day, compared to what we would do the next day. That way students found the reading easier, it went faster, and they'd bring in a few questions the next day that we could review (before moving on to new material).

When I grew up, I was forced to go to a prep school where we had 3 or more hours of homework a night, plus we were required to stay for some form of athletics, so I rarely got home before 6 PM. With that in mind, I was selective about homework. For the time I was teaching in public schools (as opposed to my time teaching in treatment), the dept. heads and supervisors jumped on me for not giving enough homework or for assigning science material we had reviewed in class - pretty much everything about my homework system offended the dept heads or supervisors.

Meta/Oculus were developing a solution to that but it was canned a few years ago because it would have made the headset a lot more expensive. It's not a big issue in most cases because VR apps generally don't try to do depth-of-field effects.

The simpler solution is eye-tracking to figure out what you're looking at and depth-of-field based on where your eyes are looking, but Meta/Oculus were building something that would adjust the lenses based on where you were looking so you could physically focus near and far; I don't remember the details.

The biggest seller for something like Vision Pro would probably be a nudify app which detects hot chicks and uses an AI model to remove their clothes. Which Apple would never allow.

I never had a problem with motion sickness in 90s VR, even with those glorious maybe-15fps graphics. I do sometimes feel it with modern VR, but I just avoid those few games that trigger it (e.g. Half-Life VR is particularly bad for me but I don't know why as it doesn't seem much different to other games I can play for hours).

Most of the motion sickness triggers were debugged and solved years ago so app developers should know how to avoid them.

Maybe. I think I've spent around $3k on my VR headsets over the last decade and played over a thousand hours of VR in Skyrim alone.

So if you're actually going to use it, it's pretty cheap in terms of cost per hour. If you're going to play a few demos and then put the $500 headset on a shelf, yeah.

3D TV and it just wasn't very good unless you had a huge TV or sat very close to the screen. I watched a few movies on ours but in the end it wasn't worth the hassle.

The best 3D movie I ever saw was an IMAX 3D demo video in the 90s because the tech was new, they were trying to wow everyone and I was sitting close enough to the screen that I could barely see anything outside the screen, just a little black line around the edge of my field of view. You can't really do that in your living room.

The second best is 3D movies playing in a VR headset. I've been buying up lots of used 3D Blurays ever since I found out how to do that.

It depends on how you use git. My primary working repositories are all local, and not readable or writable by anyone else (nor should they ever be). I push my working branches to GitHub regularly, so others can pull from them. My collaborators on various projects do the same thing. I could push anywhere that's appropriate. A corporate or publicly-accessible Forgejo instance, GitLab, or even just a folder on my web server.

In my naivete I thought this sort of local repo, public repo split is how git was designed to work. Pull requests could take any form, really, provided commit messages meet your requirements (this commit closes issue #foo), and it's simply a matter of git pull. Sure I have to add remotes to my local git repo, and possibly create local branches to review the pull request commits, but that's not too hard. git gui's like gitg come in handy for navigating numerous branches and remotes. The author could have done that in the time it took him to write a complaint about GitHub. And like I said, github actions can be made to run locally in each collaborator's local repo.

git is decentralized. Development need not stop when GitHub has problems, and it should be trivial to move off of GitHub if that was necessary. Of course, maybe I just use git wrong. And sure I can understand that corporate development doesn't really mesh that well with git's distributed model, hence the use of GitHub as a crutch.