[HN Gopher] Vibe coding the MIT course catalog
___________________________________________________________________
Vibe coding the MIT course catalog
Author : low_tech_punk
Score : 52 points
Date : 2025-08-06 18:23 UTC (4 hours ago)
(HTM) web link (stackdiver.com)
(TXT) w3m dump (stackdiver.com)
| stackdiver wrote:
| tl;dr the app is at https://chuanqisun.github.io/courseek/
|
| I also wrote a short piece reflecting on vibe coding this app and
| musing on the broader implication of building non-scalable
| software. https://stackdiver.com/posts/non-scalable-software/
| ncr100 wrote:
| Advice: capitalize the S in Courseek, to emphasize the
| misspelling as intentional.
|
| And, AWESOME WORK. This is truly "making things better", which is
| my favorite What Does An Engineer Do saying.
| plasticchris wrote:
| Unless the meaning was CourseEEK
| r3a1d33p wrote:
| Lots of school infrastructure is decades old. I bet lots of
| students can better manage the IT than the IT department. The
| only issue is security and privacy. Not sure if there is a
| balanced solution for self-governance.
| stokehacker wrote:
| SIPB is the student run hacking club at MIT
| (https://sipb.mit.edu/)
| Thrymr wrote:
| At MIT, SIPB _has_ managed IT better then the IT department
| for a long time.
| isaacremuant wrote:
| Usually it's not a lack of will but just institutional
| entrenchment.
|
| Students may volunteer but there's no path that would allow
| them to fix things.
| kkylin wrote:
| I was around when MIT had its first online catalog.
| Reportedly hacked together by a grad student over a summer
| (probably not a whole summer).
|
| Anyway I think for the catalog there are no serious privacy
| issues, and there wouldn't be a problem having students work
| on it. Databases with student information (and that includes
| registering for classes) are a different story.
| godelski wrote:
| I can confirm this. I recently graduated (PhD) and was
| constantly submitting fixes for my department's servers. I'm
| not even a networking person or anything... I do ML.
|
| And by fixes, I mean supplying the code and everything. One
| example is that the listing of classes generated a link as
| long as there was a folder on the server. Which the folders
| are automatically generated via the course offering of that
| term. This led to lots of links to nowhere when students were
| trying to see the course material and syllabus from previous
| offerings (so it was something people did quite a bit. Click,
| 404, back, click next, 404, repeat...). So the for my submit
| I found the file providing all the links, and the few lines
| around there. Provided a few alternatives such as "don't
| provide link unless there's an index.html inside" or "don't
| provide link if folder is empty". I even generated the one
| line `find` command that could go through and purge all empty
| directories that were older than 6mo (or any desired time).
| (That would clean up for anyone looking via the cli, which
| was more common than you'd expect) All the work was done,
| just someone with permissions needed to run.
|
| There were tons of small fixes (as well as some much bigger
| ones...) like this that I and others submitted. Very few ever
| were done. Maybe the IT guy's hands were tied but every time
| I walked past his office he was sitting on his couch watching
| Netflix on his iPad[0]. I saw 3 different IT employees in my
| time and none of them took action on any of those types of
| issues.
|
| I think the small things just get brushed off. Thinking "oh,
| well it's frustrating, but only happens a few times a year
| and doesn't cause much harm." Which, is true. But also, isn't
| the beauty of scaling in CS that even though 15 minutes of my
| work only saves a person 5 minutes (3 times a year, every
| year) that I'm saving that 5 minutes for hundreds or
| thousands of people? Saving 30 seconds doesn't sound like a
| lot but saving saving 100 people 30 seconds is saving an
| hour. If they have to do that once a day then you're saving a
| full work day every week. That's only 100 people too...
|
| But I learned an important lesson: the little things matter.
|
| [0] Not to bash him too much. When shit went down he usually
| responded pretty quickly. And grad students work some weird
| hours...
| conception wrote:
| You probably were speaking in hyperbole, but there's not a lot
| of school equipment out there still running windows XP and
| older. There's some certainly but "lots"? Probably not except
| in the poorest of school systems.
| buckle8017 wrote:
| Windows XP? no
|
| perl scripts written in the 90s controlling grades and course
| registration? Absolutely
| wpm wrote:
| I started at a university full-time in 2014 in the IT
| department. Same year a project to decommission an aging
| IBM mainframe that sat as a lynchpin in nearly every
| important operation on campus; course registration and thus
| billing and grades, transcripts, as well as a home-grown
| SSO from the early 2000s no one understood any more. The
| consulting costs were insane, as were the support costs, so
| that poor thing had to go (second prettiest rig in the
| datacenter after my Xserves).
|
| When I left that job in 2022, I believe _most_ of it had
| been offloadded, but I can 't say for sure if they had
| actually shut the fucker down yet.
| daedrdev wrote:
| I've heard of people making reminder bots for schools that give
| you a short time slot to accept getting off a waitlist before it
| moves to the next person
| mmmlinux wrote:
| My understanding is you're allowed to automate the discovery of
| classes as much as you want. but as soon as you try to automate
| signing up for them you're gonna get in trouble.
| pkal wrote:
| I have to say that https://student.mit.edu/catalog/index.cgi
| looks like a great site! It is all public, and when you open a
| category (a static site), all the information appears all at once
| and you can Ctrl-F for keywords. It might not solve that "unknown
| unknown" problem that the author mentions, but it is certainly
| much preferable to the solution that our university used
| (https://www.his.de/hisinone; one example: all courses are
| displayed in a tree and if you click to unfold a node, the server
| generates verbose HTML and sends it to the client. This takes at
| least 10 seconds on good days).
| ndriscoll wrote:
| Yeah literally the only thing they need to do to fix it is get
| rid of the no-cache and Connection: close headers. Maybe make
| an "All" page for better CTRL-F? Surely their catalog doesn't
| change more than once per minute and could have some level of
| caching (at least with revalidation)? Keep-alive would cut out
| ~150 ms of page load time and letting at least something like
| nginx cache it seems like it would cut out another ~150 ms.
___________________________________________________________________
(page generated 2025-08-06 23:00 UTC)