joshhug

Weird, it looks like it might be deterministically giving this exact answer. I guess from caching? I made a video for posterity: https://youtu.be/C_yyvznVG64

Ah, that's because the "python" is actually just using javascript evals.

} else if (mode === 'python') { if (cmd === 'exit()') { mode = 'sh'; } else { try { // Safe(ish) eval for demo purposes. // In production, never use eval. Use a JS parser library. // Mapping JS math to appear somewhat pythonesque let result = eval(cmd); if (result !== undefined) output(String(result)); } catch (e) { output(`Traceback (most recent call last):\n File "<stdin>", line 1, in <module>\n${e.name}: ${e.message}`, true); } }

This was cool: https://codepen.io/ChetasLua/pen/yyezLjN

Somewhat amusing 4th wall breaking if you open Python from the terminal in the fake Windows. Examples: 1. If you try to print something using the "Python" print keyword, it opens a print dialog in your browser. 2. If you try to open a file using the "Python" open keyword, it opens a new browser tab trying to access that file.

That is, it's forwarding the print and open calls to your browser.

The part about dad joke square theory got me thinking about this classic scarecrow joke, which feels like an example from some higher order version of square theory:

"Why was the scarecrow given an award?"

"He was out standing in his field."

The fact that a scarecrow's job is to be "out standing in his field", and that excelling at one's job can be phrased as being "outstanding in his field" is an incredible linguistic coincidence.

Also the African humid period isn't visually apparent (https://en.wikipedia.org/wiki/African_humid_period).

But very cool!

I am told that an interesting alternative is the Structured State Space for Sequence Modeling (S4). I don't personally know much about this technique, but didn't see anybody else mention this in this thread.

https://srush.github.io/annotated-s4/

I had him as a teaching assistant when I was teaching data structures at Princeton back in Fall 2013. Princeton CS has their professors rotate through as TAs every so often through their courses.

That semester, I made a slight mistake on the final exam where I asked students to create an algorithm that could find the second shortest path from s to every other vertex in a graph. I forgot to specify that the second shortest path should be simple (i.e. should not reuse any vertex twice). Having to deal with non-simple paths makes the problem much much harder.

None of the students figured it out in the time available, and I'm sure I would also have been stumped if I had tried to solve the problem. Bob figured it out though. And then I remember he graded all 150 solutions to the problem himself, having as a blast as he went through students attempts at an effectively impossible problem.

I just tried o1, and it did pretty well with understanding this minor issue with subtitles on a Dutch TV show we were watching.

I asked it "I was watching a show and in the subtitles an umlaut u was rendered as 1/4, i.e. a single character that said 1/4. Why would this happen?"

and it gave a pretty thorough explanation of exactly which encoding issue was to blame.

https://chatgpt.com/share/66e37145-72bc-800a-be7b-f7c76471a1...

That was interesting. I asked it to try to say something in another language, and she read it in a thick American accent. No surprise. Then I asked her to sing, and she said something like "asterisk in a robotic singing voice asterisk...", and then later explained that she's just text to speech. Ah, ok, that's about what I expected.

But then I asked her to integrate sin(x) * e^x and got this bizarre answer that started out as speech sounds but then degenerated into chaos. Out of curiosity, why and how did she end up generating samples that sounded rather unlike speech?

Here's a recording: https://youtu.be/wWhxF7ybiAc

FWIW, I can get this behavior pretty consistently if I chat with her a while about her voice capabilities and then go into a math question.

I'm curious to hear feedback, especially:

1. Given the limited amount of scarce student time I want to spend on this, do pull requests / GitHub actions tests (provided for them) / code reviews seem like the right team processes to introduce?

2. Each student's work is independent of every other student's. Is there some way to have them working on different pieces of the same problem, or having the students solutions interact in some deeper way than random selection in a lottery? On my first post of this idea, a user named taftster imagined taking inspiration from Factorio in some way, with students each contributing some piece of a large system.

3. Is there a way we could naturally allow pull requests for the contest code or maybe even the README that establishes how to review pull requests? I'm thinking a bit about Fluxx or Nomic here.

I'm curious to hear feedback, especially:

1. Given the limited amount of scarce student time I want to spend on this, do pull requests / GitHub actions tests (provided for them) / code reviews seem like the right team processes to introduce?

2. Each student's work is independent of every other student. Is there some way to have them working on different pieces of the same problem, or having the students solutions interact in some deeper way than random selection in a lottery? A user named taftster imagined taking inspiration from Factorio in some way, with students each contributing some piece of a system. I've so far lacked the creativity to imagine what the API for this might look like for this assessment.

3. Is there a way we could have students make pull requests for the contest code itself? I'm thinking a bit about Fluxx here.

Like a lot of schools, our Data Structures course doubles as a soft intro to software engineering. I want them to know how to solve real problems efficiently and manage complexity.

Student time is incredibly precious. The hope here is that students will spend less than an hour on this, and in the process get a taste of how code reviews / automation can help teams function better. I'm hoping I can thread the needle and give them tools that will help them more efficiently complete their capstone project at the end of the semester, without spending too much of their time teaching them these tools.

FWIW, we have discussed the idea of a lower division software engineering class, and I was really hoping Pamela Fox would do this before she headed back to industry (interesting story there: https://blog.pamelafox.org/2022/05/my-experience-as-unit-18-...). We do have an upper div software engineering course as well, though it's not taken by most of our students.

I'll get pondering. Having some sort of rube goldeberg-esque mega project was my original idea, but finding a specific form for it has been elusive.

Taking inspiration from Factorio in some way is an interesting proposition. Will consider!

I like these! And I encourage a little bit of cheating. I'm curious to see how the particularly clever students will subvert the system. I've seen some very weird stuff in the past.

I had a fun assignment back in the day when I taught security at Princeton back in 2012. In the final assignment, assignment 8 students were given a Linux disk image of a hard drive owned by a guy named Nefarious who had committed a murder. This was an assignment originally created by J Alex Halderman and Ed Felton, but I went a little extra.

In the assignment text, I mentioned that Nefarious was originally arrested due to an anonymous tip from a pseudonymous "Cecco Beppe". Buried in /home/root on Nefarious's drive, there is an innocuously labeled file: CB.7z, which is notable only in that it is the only file on the drive dated 2012 (intentionally), years newer than all of the files on the drive. If one went to the trouble of unzipping this file, they'd realize it is a very small disk image inside the disk image.

Booting this image dropped the user into a chat with a depressed artificial intelligence (a custom ALICE chat bot, but with some hard coded responses to advance my story, this was pre-GPT) who, when prodded with the appropriate keywords, told the tale of depression and attempted AI suicide that followed Hapless Victim's death -- leaving the AI nothing more than a pathetic incoherent chat bot. Further prodding led the AI to reveal the existence of secret assignment #9 as well as a URL for said assignment.

At the URL, there is a zip file containing a flawed pseudorandom generator, a file encryption tool that uses said PRGen, an encrypted file, and a truncated copy of the corresponding plaintext. The truncated plaintext explains that their next task is to complete decryption of the file, which they can do by exploiting the flaw in the PRGenerator.

The deciphered text then explains that their next task is to go back to the HW1 autograder and submit a new solution for HW1 (which was to develop a PRGen) -- but with the catch that they may only use print statements.

Once they submitted this cheating print statement code via the usual Princeton web submit, the Princeton web submit script explained the final part of secret assignment 9, which is to steal the source code for the PRGen autograder that graded their first assignment. No hints were given.

This last part could be done by simply writing code that opening the .class files and prints them to the screen when the autograder script runs.

I had around 10 students (out of ~150) figure this all out, it was great.

1. It's definitely not about the Comparator. The lesson is all about showing off a CI/code review process. This is intended to be a short exercise, expected to take no more than an hour. The CI tests can't check that the comparator actually does what it claims, they'll just test that it's a valid comparator. And I'd build on this in a future pair assignment, where they can use some or all of the steps of the workflow from this more gimmicky assignment.

2. I agree it's a bad use of student time to see if someone's submission is unique, but I suspect there are ways we could structurally push students to do something unique, e.g. gus_massa's post below.

3. Yeah, it's a little weird. The integer submission is just to keep the macguffin of the assignment going. Also for N submissions, we'll use at least N-1 comparators. If we use RNG that cycles through every comparator before repeating, we can make sure everyone gets used at least once.

This is a weird assignment idea. I’d love to hear what you think, how you might improve it, etc.

Some specific open questions I have:

1. How can I nudge students towards being creative with their comparators so that we get more variety?

2. Is there a natural extension that would get students working on the same files, yielding a richer demonstration of CI?

3. Could we allow students to somehow edit the contest code itself without breaking the entire idea of the project?