[gchallen]

They have not succeeded in forcing me, yet. But it's sad how many computing faculty apparently can't operate the basic online infrastructure needed to support their courses. Not that universities make it easy for us.

And of course the other serious concern I have with Canvas is that they are likely using all the materials faculty upload to train their AI replacements. Many of my colleagues engage in dark humor about this but I haven't noticed much action.

[JumpCrisscross]

> they are likely using all the materials faculty upload to train their AI replacements

Instructure (Canvas's developer) partnered with OpenAI last year [1], about a year after KKR and Dragoneer (PE firms) acquired it [2].

[1] https://www.forbes.com/sites/rayravaglia/2025/07/23/instruct...

[2] https://www.pehub.com/kkr-and-dragoneer-complete-4-8bn-take-...

[lucas_v]

instructure/canvas-lms is open-source -- is there anything preventing universities from hosting it themselves?

[Meneth]

A bunch of plugins running on canvas.instructure.com are proprietary, according to their FAQ: https://github.com/instructure/canvas-lms/wiki/FAQ.

I would guess these plugins are chosen so a majority of user won't want to live without them.

It also seems these plugins "link" to canvas-lms, so keeping the proprietary would be a GPL violation if anyone except Instructure holds part of the copyright to Canvas.

[dotancohen]

Money, skill, liability.

That calculus is about to shift.

[FloorEgg]

I'm sure the engineers at instructure are not capable of building systems that can do that. You give them too much credit.

[freedomben]

Former Instructure engineer here. Ive been gone almost 10 years at this point, but some of the best engineers I've ever worked with were at INST.

I'm not sure where your stereotype even comes from, because Canvas is not trivial software. You can see for yourself as it's AGPL and I assume you looked at the code before criticizing it because any good engineer would do that.

[w4der]

I don't care how good you think it is, the fact that it (back when I used to be a TA) would break if two TAs tried concurrently grading different parts of an assignment of a student is bonkers. The workaround for that was to use a Google Sheet document so TAs just looked at the submission in Canvas, then filed in their grades and feedback on the sheet. The issue is that Canvas, as far as I could tell, did not support mass uploads from a csv, so we had a script which would read every entry on the csv, map that to the student's ID and grade them, which made it look like the TA which had generated the API key graded all of the students (and would get all the backlash from poor grades).

I completely agree that it is not trivial software in the worst sense, it tries to do too much, while not being particularly good at any one of those things, and is way too rigid for how diverse the needs of different courses might be even inside a single faculty. And saying "It's AGPL, just self host and add your requirements to it" is not really useful, that would mean way more money and effort than what a university's overworked IT dept. is capable of.

[slfnflctd]

This particular section of your comment sounds a helluvalot like some of the SAP implementations I've seen:

> it is not trivial software in the worst sense, it tries to do too much, while not being particularly good at any one of those things, and is way too rigid

[FloorEgg]

I didn't say instructure engineers were bad.

What I meant is they aren't capable of building AI capable of replacing professors. I still consider it a reasonable assumption, as it has nothing to do with how well engineered canvas is. It's a different competency than instructure would have, and I've heard from insiders instructure has been spinning their wheels on way more trivial AI challenges. I also understand well how hard it would be to create AI that replaces professors and how the current best AI from Google, Anthropic, OpenAI is orders of magnitude away from being able to do that.

An engineering culture can change a lot in 10 years, and a company's engineers' ability to do stuff depends both on the individual engineers abilities as well as the company systems and culture.

[mbesto]

> some of the best engineers I've ever worked with were at INST.

> You can see for yourself as it's AGPL and I assume you looked at the code

Can you look at any codebase and tell me it's written by some of the best engineers and it's not trivial?

[hackyhacky]

I've been using Canvas for years and it's some of the worst written software I've ever used. It's slow, buggy, with an atrocious 2001-era UI. It's a CRUD app that has no excuse for being so cumbersome. I'm not surprised at all that their security is just as bad as the rest of the product.

A bright undergrad could build a superior replacement in a few months, even without AI.

[freedomben]

I won't disagree on usability. It has some sharp edges for sure. But

> A bright undergrad could build a superior replacement in a few months, even without AI.

Is quite naive. Canvas is not at all just a crud app. You can view the code yourself as it's AGPL

[ericpauley]

What component in particular goes substantially beyond CRUD?

[jameshart]

It is a very common error to look at a specialist piece of software, superficially consider the basic data structure it appears to have and think ‘seems simple enough. Basic CRUD app.’

But it’s rarely the case in practice.

In a sibling comment right here for example someone bemoaned the difficulty in Canvas of having two TAs simultaneously grade separate parts of the same assignment. That sounds like something that goes beyond CRUD.

But more importantly any workflow system, which an LMS will be full of, has to handle the always tricky problem of how changes to workflows affect the things that are currently in the workflow. Assignments posted in course X need to be approved by person Y; some assignments are submitted for approval; person Y goes on leave and now the approval needs to be person Z. Not a simple CRUD problem.

These are things that occur to me with only a moment’s consideration of what an LMS system might need to deal with. The actual domain probably has considerable more complexity that I can’t even imagine.

[hunter2_]

If they're at the level you say, they just might install some AI gizmo like the Vercel employee was accused of, but really let it run amok with write permissions.