How software development is taught
The Handmade Network exists because we are unhappy with the software status quo. To a large extent, this is because of how software development is taught. What are the good parts of software education today, what are the flaws, and how might we change things to improve the state of software?
Hello, and welcome to Fishbowl Day 8! 
8️⃣
Topic: How software development is taught
❗ Idea Source / Discussion Context: https://github.com/AsafGartner/hmn_fishbowl/discussions/35
Relevant additional context: https://handmade.network/podcast/ep/eeac46c9-5e90-4b1c-9104-cff49734764d
Please read over the discussion context if you haven't already. It is recommended that you also listened to the podcast as some of this was explored a bit by @demetrispanos and @ryanfleury , though if you haven't that's alright too.
A few goals I have for this fishbowl are:
1. Discuss how formal software development education is flawed. From the podcast, we learned how societal and economic pressures have influenced universities to become this way, let's break down exactly what they're doing wrong and what they could do differently if these factors weren't a concern. For example, if professors were completely and entirely motivated on producing the best programmers they possibly could. Let's also discuss what they're doing right.
2. Let's extend this example to other formal learning environments: Books, Online Courses, Videos, etc. What goes wrong there? What goes right?
3. Perhaps most important, and what I would like to spend the most time focusing on, are solutions. What would be a better way to train someone in software development? How does it start, what does the journeyman path look like, what about advanced learning? Be specific.
Participants:
@Rudy [moderating]
@raysan5
@demetrispanos
@greensleevless
@ratchetfreak
@CianMc
@bumbread
6
3
4
@bvisness is adding roles now (thank you 
)
)
You can read adjacent conversation from #fishbowl-audience, the peanut gallery, here: https://discord.com/channels/239737791225790464/708458209131757598/855128040811003966
2
2
Hey folks I’m Rudy - I proposed the topic and will be moderating the discussion. I work as a jack-of-all trades on our systems team at Automattic - we are the group responsible for the availability, performance, reliability, and security of all Automattic’s services. I’m currently focused primarily on Tumblr.com (yes that one ) I spent a long time in the Military where some of that time was served as an NEC 9502 (Instructor), I spend some time making educational videos on youtube and I earned a Bachelors and Masters Degree. My wife is also a professor at the University of Texas at San Antonio and through her, I understand many of the topics related to higher education are not necessarily specific to computer science.
- I proposed the topic and will be moderating the discussion. I work as a jack-of-all trades on our systems team at Automattic - we are the group responsible for the availability, performance, reliability, and security of all Automattic’s services. I’m currently focused primarily on Tumblr.com (yes that one ) I spent a long time in the Military where some of that time was served as an NEC 9502 (Instructor), I spend some time making educational videos on youtube and I earned a Bachelors and Masters Degree. My wife is also a professor at the University of Texas at San Antonio and through her, I understand many of the topics related to higher education are not necessarily specific to computer science.
the podcast made the case that programming should be taught in a tradeschool instead of a uni
I might expand this to something like a "specialist school" since "trade school" has connotations (at least in the US) that maybe aren't helpful
music school, art school, design school, architecture school, these are generally seen as equally prestigious as universities
but could all be called trade schools
but could all be called trade schools
in a platonic sense certainly, you could throw in law and medical school too
Nice!
Jumping back up to the conversation that started ^pre intros^ - I thought the podcast discussion about universities existing across this gap of "centers of aggregated knowledge" <-> "human capital factories" was really interesting. How do you see specialist schools differing / why would they avoid the pitfalls of existing higher ed?
the "HCF" view is explicitly the view taken by music schools, architecture schools, and so on
With trade schools, does anyone think that there could be an issue with attracting the necessary teaching talent. I agree with the concept but with such a competitive job market would it not be difficult to attract the required teaching talent?
On my experience, trade schools are more open to professionals teaching some specific subject than at University level where usually some specific titles (as a PhD) are required to teach there.
1
On my experience, trade schools are more open to professionals teaching some specific subject than at University level where usually some specific titles (as a PhD) are required to teach there.
that's true, it would definitely be more attractive and open to a working professional with a Bachelors/no degree
On my experience, trade schools are more open to professionals teaching some specific subject than at University level where usually some specific titles (as a PhD) are required to teach there.
My experience at design school feels similar to that aspect of trade school - our teachers were often people commuting in from their ~jobs as designers/illustrators/artists/etc - they often didn't have higher degrees than the ones we were studying for, but relied on portfolios of work instead.
1
@demetrispanos @CianMc one thing to consider is the market has already tried to address this in the form of "boot camps"
I feel like boot camps, generally, I don't have much experience of them, try and do the bare minimum in the least amount of time to make someone "employable" without delving too much into any one subject, which is fine, but there's also a need for something like a trade school where a deeper understanding can be developed
1
I feel like boot camps, generally, I don't have much experience of them, try and do the bare minimum in the least amount of time to make someone "employable" without delving too much into any one subject, which is fine, but there's also a need for something like a trade school where a deeper understanding can be developed
at the risk of being provocative, I think most boot camps are either cynical or one step removed from cynicism
@CianMc makes a good point that specifically for programming, there is a huge cost problem in getting good people
I think it's very difficult for Universities to find teachers for high-technical subjects, at this moment there are many better offers out-there for those profiles than what the University can offer...
the idea that "anyone can be a programmer" is a real problem, and underlies the boot camp premise
"learn programming in 20 days" 
I think most traditional engineers could become programmers, that seems empirically true
There definitely is a certain shared problem solving mindset between those professions
It might be an interesting point of discussion to talk about how music is taught and what we can translate to how to teach programming /sw engineering
one essential part of music/arch/design/art school is you spend most of your time doing the thing
synthesis vs analysis 
while I feel everyone should be introduced to programming at a young age so that everyone can see if it's for them or not, the idea that just because there's a large demand for developers that everyone should consider learning it is ridiculous and inherently leads to a reduction in software quality
I believe that everyone can learn to program quality software. We just don't have any teachers that would teach how to do it. Universities do not concentrate on correctness of the code too much, even though the point is spoken, the actual quality is not checked ~ the code is "tested" in the best case.
@bumbread well, I think it depends on the teacher
and the student
2
I believe that everyone can learn to program quality software. We just don't have any teachers that would teach how to do it. Universities do not concentrate on correctness of the code too much, even though the point is spoken, the actual quality is not checked ~ the code is "tested" in the best case.
that's an interesting view, maybe those that currently excel at programming have a natural ability/are good self learners, and others who struggle could become better developers if the proper guidance and resources were available to them?
let's stay focused on the problem at hand
1
I believe that everyone can learn to program quality software. We just don't have any teachers that would teach how to do it. Universities do not concentrate on correctness of the code too much, even though the point is spoken, the actual quality is not checked ~ the code is "tested" in the best case.
personally, I focus way more on the soft skills that students need to actually learn programming than in advance programming concepts. For example: organization, coding conventions, focus to detail, variables naming, debugging, patience, understanding the process...
personally, I focus way more on the soft skills that students need to actually learn programming than in advance programming concepts. For example: organization, coding conventions, focus to detail, variables naming, debugging, patience, understanding the process...
Curious: why do you say 'soft skills'?
Curious: why do you say 'soft skills'?
I would also consider those hard skills
that's an interesting view, maybe those that currently excel at programming have a natural ability/are good self learners, and others who struggle could become better developers if the proper guidance and resources were available to them?
I don't like the idea of "natural ability", nor do I believe in it. All people have different experience, e.g. some were doing painting, so naturally everyone would understand different concepts differently. Some people also can learn based on the experience.
I don't like the idea of "natural ability", nor do I believe in it. All people have different experience, e.g. some were doing painting, so naturally everyone would understand different concepts differently. Some people also can learn based on the experience.
neither do I, and maybe that's the wrong term for it, it's more probable that certain people can learn things by themselves in an unstructured in environment whereas others just need a structured/guided learning experience
Curious: why do you say 'soft skills'?
I mean, not that purely technical programming concepts but more a useful way to think to write code
for example, I require my students to follow strict coding and naming guidelines, not only on the deliveries but also on the daily work; I question my students a lot about the concepts I explain, I don't want that they just believe what the teachers says
that's a very good approach in my mind
one problem here is that "good" is contextual
agreed
for example: writing non-clever code, always writing with the next person in mind
this is just not true for test-an-idea code though
1
hard agree. I guess the problem is people tend to pipeline their "test an idea" code into production
1
and I think many many people suffer for lack of knowing "test an idea" practices
that was a big one for me, often I would never make any progress as I thought that my first implementation had to be flawless rather than a proof of concept
a related skill is how to then progress from that test stage to a robust piece of software
this feels like the natural solution to this
that was a big one for me, often I would never make any progress as I thought that my first implementation had to be flawless rather than a proof of concept
I suffered from this myself for a few years and I had to train myself consciously to just write the simplest and stupidest thing first
same, but at least we're aware of it now, rather than not knowing where we went wrong 
I suffered from this myself for a few years and I had to train myself consciously to just write the simplest and stupidest thing first
Still suffering 
hard agree. I guess the problem is people tend to pipeline their "test an idea" code into production
Throwing something in from design school - the idea of a sketch is super important. It's never meant to become the final piece, just something you do to prepare for when you actually start
Throwing something in from design school - the idea of a sketch is super important. It's never meant to become the final piece, just something you do to prepare for when you actually start
right, well this brushes up on my favorite hobby horse the "plastics and metals" metaphor, which is based on how industrial product designers prototype things
maybe a better structure in formal learning would be the same assignment three times. 1. get it to "just work" 2. reconstruct it entirely with the next maintainer in mind 3. have someone else read and optimize it
absolutely
which our education system actively discourages
100% very important here, schools condition you to see your performance as deviation from perfection, as opposed to deviation from failure
schools say "you start with 100 points, and you lose points for making mistakes"
3
schools say "you start with 100 points, and you lose points for making mistakes"
The game designer in me went to "those are mechanically the same" but the perception is super important
I'm trying to keep the jokes out of this but can't help my brain from typing "10 POINTS TO GRYFFINDOR"
2
maybe a better structure in formal learning would be the same assignment three times. 1. get it to "just work" 2. reconstruct it entirely with the next maintainer in mind 3. have someone else read and optimize it
I usually provide my students with a GitHub template to work over it
1
In this case they get the code from the instructor
they have some base code to start working with, they can use it or not, but the good thing about it is that they already have a preconfigured package for a nice delivery and presentation of the product, not only focused on the code itself
That's interesting - makes me think about the skill of "getting started well". Using your example to show them what that looks like, but do you ever have them then try and start on their own on follow up assignments? Is there value in that?
by the way, the code in that template is not written by me, it was created by a previous teacher, I just tried to adapt it a little... but I was not allowed by University to adapt it to what I'd like it to be... that's probably another discussion point...
1
Is there anything higher learning institutions, books, training courses, i.e. existing paths to learn software development, get right?
they teach the theoretical well
well... 
the port of those fundamentals to practice is what I think could fail...
To be a bit provocative - and this comes from the way the question was originally phrased ("when do we switch from foundations like mathematics, algorithms and data structures to creating software") - is that I think that order is backwards
Another thing institutions do well are labs, either where structured work is given or where it's used as a place where students can ask the professor/tutor for help with the current assignment
1
unless they let the student BS their way out of the work of course
yeap, but, again, it will depend on the teacher
Yep, or what happens quite often is that the college doesn't hire enough tutors so students get very little or no 1-1 time
agree!
Yep, or what happens quite often is that the college doesn't hire enough tutors so students get very little or no 1-1 time
In my experience, and this was covered in the podcast a bit, but teaching is really an afterthought for professors
In my experience, and this was covered in the podcast a bit, but teaching is really an afterthought for professors
also agree... and I think that's a big problem
1
To be a bit provocative - and this comes from the way the question was originally phrased ("when do we switch from foundations like mathematics, algorithms and data structures to creating software") - is that I think that order is backwards
do you think fundamentals are actually conditioned by the way software is written today? should them?
To be a bit provocative - and this comes from the way the question was originally phrased ("when do we switch from foundations like mathematics, algorithms and data structures to creating software") - is that I think that order is backwards
Could you elaborate
Could you elaborate
Yea so a lot of what I've worked on was building an engagement curve for kids - how do we show them why engineering is interesting/worthwhile - and the way you do that is by letting that have perceived impact early on. We can't start by teaching them algorithms - they don't care. And I think this scales up to adults - we aren't fundamentally different from kids, we just have more experience.
2
Yea so a lot of what I've worked on was building an engagement curve for kids - how do we show them why engineering is interesting/worthwhile - and the way you do that is by letting that have perceived impact early on. We can't start by teaching them algorithms - they don't care. And I think this scales up to adults - we aren't fundamentally different from kids, we just have more experience.
love this
So when we're talking about how to teach people software engineering better, I think the best thing to do is to get them to the thing they care about as quickly as possible
and so I created raylib
So when we're talking about how to teach people software engineering better, I think the best thing to do is to get them to the thing they care about as quickly as possible
And then the possibilities are endless. You could say "hey see your program is a piece of shit. in this course I will teach you how to make this piece of shit into an actually good program"
1
So when we're talking about how to teach people software engineering better, I think the best thing to do is to get them to the thing they care about as quickly as possible
yes this also relates to programming language design, and the huge gap between when I learned basic/logo vs what people learn today
when I started as a kid, you just booted the thing and wrote a few lines of code and you had pixels on the screen
nowadays this thing is rarely exciting unless they have call of duty running, and IMO having that created in under an hour is a very poisonous idea
nowadays this thing is rarely exciting unless they have call of duty running, and IMO having that created in under an hour is a very poisonous idea
I find that if you show kids they can do this in the browser, some are fairly interested
and so I created raylib
Ya so I'm curious what design considerations went into that?
and so I created raylib
Oh I forgot that you are a person who created raylib lol
1
Ya so I'm curious what design considerations went into that?
Simple, I had 50h to teach art students to program games (they had never written a single line of code before)... and still I wanted that they actually wrote some code, so, I tried to create something simple and self-contained that could fit in my teaching methodology
sincerely, I didn't expect it to work... but students start seeing things moving on the screen connected with the code and that was VERY engaging for them
immediate feedback on iteration is very important to keeping people engages in learning
2
sincerely, I didn't expect it to work... but students start seeing things moving on the screen connected with the code and that was VERY engaging for them
This reminds me of something Dan Cook (from Spry Fox) said on Twitter a few months ago - how good game design (and design more broadly) is about arranging the world to reveal value in specific places. That seems to intersect with good education as well. Is it fair to say that this is kind of how raylib facilitates education early on? How do you think about growing past that - to where you have to start dealing with externalities?
re "revealing value", there's the saying "education is not filling a bucket, but lighting a fire"
6
re "revealing value", there's the saying "education is not filling a bucket, but lighting a fire"
Yep - its kind of where I was heading with thinking about growth past that initial intentionally-designed-for-education environment. Once you light that fire, does growth beyond your initial introduction kind of happen on its own?
re "revealing value", there's the saying "education is not filling a bucket, but lighting a fire"
its a great quote
(i.e. creating some self-sustaining process in the student)
that's it!
This reminds me of something Dan Cook (from Spry Fox) said on Twitter a few months ago - how good game design (and design more broadly) is about arranging the world to reveal value in specific places. That seems to intersect with good education as well. Is it fair to say that this is kind of how raylib facilitates education early on? How do you think about growing past that - to where you have to start dealing with externalities?
what do you mean with externalities?
what do you mean with externalities?
This is where I think this discussion starts to have a ton of overlap with the HM philosophy.
So you constructed raylib in such a way that students get to impact (pixels on screen) very quickly. But the world isn't like that. Eventually you have to get them to start engaging with the existing software environment - build grit towards things like poor documentation, bad APIs, etc.
once you know that you can, in principle, do things you want with programming then you're more willing to tolerate "ok here's all the ugly stuff"
1
once you know that you can, in principle, do things you want with programming then you're more willing to tolerate "ok here's all the ugly stuff"
That's interesting - I 100% agree, just never thought about it that way
That's interesting - I 100% agree, just never thought about it that way
there's a psychology "trick" that is common in the military, which is often summarized as "embrace the suck" where you actively discuss how bad something is going to be
where I tell people "ok, now we're going to do a bunch of stuff that's really awful and stupid, and you shouldn't feel bad thinking that it's awful and stupid ... everyone feels that way"
1
And I'm not saying this to downplay the value things like raylib have - I think that role is possibly the most important (calling back to the 'light a fire' idea). But there are steps beyond that as well that are important too right?
raylib was originally created to be just a playground for art students, as you said, to "light a fire", most of those students just had fun for a while and never wrote code again... but some of them (and those are the key) decided to learn more and go deeper down the rabbit hole... and they learned Unity or Unreal later on but when some fundamental question pop-up, most of them came back to raylib sources (that are very readable) to understand how things work in the background.
it makes it harder to sell that idea when in military movies their friends are dying and you know it's going to suck then you watch movies with hackers and they're in the CIA mainframe in 5 minutes
1
it makes it harder to sell that idea when in military movies their friends are dying and you know it's going to suck then you watch movies with hackers and they're in the CIA mainframe in 5 minutes
yes another thing I do with young students is talk to them about misleading hacker portrayals, and just make fun of it a bit
I wonder if a piece of the education discussion is around how teaching is equal parts expertise + experience design? Like its not enough that @raysan5 is a good engineer, but also that raylib incorporates thoughts on the experience of the student
if you have never written code before and it looks like magic to you, being able to make some magic for a while is an unforgettable experience
if you have never written code before and it looks like magic to you, being able to make some magic for a while is an unforgettable experience
That is really wonderful
a huge reason for the high dropout rates in Comp Sci is the lack of understanding of what programming is, usually fuelled by hollywood lol
i think this has contributed to a subtle disdain I have always had for the word 'coding'
programming is a process
yes, but not all the students want to be programmers
if you have never written code before and it looks like magic to you, being able to make some magic for a while is an unforgettable experience
yes this is what I mean about the old-days experience, where you boot your Apple II and write 10 lines of code and you have a polygon
3
depends on which api you picked that day
there are APIs that hide the 20 hour opera, but the opera is happening
because that's where people go to learn programming
1
for the first question I'd add "why do we even care if universities are good places to learn programming?"
I would assume it's because of the systemic changes to academia over the last 40 or so years, like you mentioned
because that's where people go to learn programming
right, I agree, but this could in principle change
right, I agree, but this could in principle change
sure but that would require a mentality shift
the problem is that people have expectations for CS departments that are not realistic
but they're sold that
so perhaps fixing the universities is not the most important avenue
I think Universities are still a strong point of reference for education, so, if their educational offer can be improved, it's a positive thing
1
I think Universities are still a strong point of reference for education, so, if their educational offer can be improved, it's a positive thing
my point is that improvement presumes a goal, and I'm saying that for the university's true goal (advancing knowledge) they do an ok job
1
again, there is much blame to go around, but I don't think it's wise to expect universities to change their focus to be mainly about creating good programmers
1
maybe the real lesson is universities shouldn't be the perceived route for becoming a developer
this is, I think, the most important thing
again, there is much blame to go around, but I don't think it's wise to expect universities to change their focus to be mainly about creating good programmers
personally, the main lessons University taught me were not technical but being organized, responsible, punctual, to learn fast, to adapt to changes, identify key elements to learn... but I imagine not all the students learn the same way...
personally, the main lessons University taught me were not technical but being organized, responsible, punctual, to learn fast, to adapt to changes, identify key elements to learn... but I imagine not all the students learn the same way...
as a counterpoint, for me personally, I learned an enormous amount of domain knowledge from my university courses
I think university was the most efficient path by far for me to get the knowledge I wanted
2
but as I mentioned in the podcast, I didn't study CS
I neither studied CS
personally, the main lessons University taught me were not technical but being organized, responsible, punctual, to learn fast, to adapt to changes, identify key elements to learn... but I imagine not all the students learn the same way...
This is interesting in the context of more online/distributed/self directed learning. I feel like the biggest thing I learned from school was a work ethic (I certainly learned a ton of domain specific stuff) - but I wouldn't have gotten that if my school weren't so demanding.
Alright everyone, fishbowls normally run about 2-3 hours, and we’re about 1.5 hours in.
We have discussed the flaws in current software development training flows, we have discussed a little about what they do right. Let’s spend the remainder of the time discussing solutions. Things like:
- Is it possible for formal institutions to change course to incorporate the necessary changes, or is this not possible given the structure of academia?
- What would a “trade school” or “music school” for software development look like? Be specific, imagine you were responsible for everything from the timeline to the curriculum. What do students learn, in which weeks, and how?
- Anything else you would like to discuss?
Jumping back to the second question - thoughts?
those are definitely second year subjects after they are comfortable with imperative programming already
Do you have thoughts on how you do that first year?
assuming they come with blank slate at the very least teach them a language
actually this is a great point that I think we should explore, in comparison to music schools
I don't think there's a serious music school that accepts students who don't know either an instrument or composition already
what about this did you want to explore?
what about this did you want to explore?
well, assumptions baked into what universities or other schools should expect
bringing this back to the discussion of an improved way of teaching software development: it'd be interesting to think about how you utilize that skill gap to teach some of the finer details. I'm thinking about how you could use something like @raysan5's project template to teach new students what a good project looks like, but then when they advance, have them create the template for new students so they start to understand what kinds of considerations lead to the systems baked into the template they learned from
actually, in the University were I taught, students have several subjects connected along the 4 years, they have that template to start with but it evolves into a big engine with an IDE
still, in terms of API design and architecture I think it can be greatly improved
do you have some specific things in mind?
Might be cheating for @raysan5
hahahaha... I was about to share the University one
I just reread @Rudy's questions again and noticed: "What do students learn, in which weeks, and how?"
Could be fun to put together a curriculum in the time we have left?
personally I think this is missing the forest for the trees, I don't think the main problem is that universities have mis-optimized their course sequences
indeed the main problem is the parentheses
4
if the problem were that mundane it would be solved by now
I think one problem with Universities is not the the syllabus but the teachers
the university is not especially interested in doing that
but I think students require that kind of teachers to see the connection between their studies and the professional world
even though that is how they are sold to the public
1
hello world is nice but far from a "first program"
sorry I don't understand. Do you mean that it is far too simple to be meaningful?
sorry I don't understand. Do you mean that it is far too simple to be meaningful?
it's going through the motions of setting up the environment and verifying that it worked
even though that is how they are sold to the public
actually, many Universities here in Spain use that "connection with companies" for internships as a selling point
I wonder what the split is in acquiring skills between formal classes vs. the internships
internship depends a lot on the company you get, in some companies you can really learn a lot
yes, one of the many problems involved here is the economics of training
1
2) teaching you programming skill takes a lot of effort from people who can make a good income
yes exactly, this is partly why it's hard to get good teachers no matter what the institution
1
and the education quality problem intensifies...
places like HMN are interesting in this discussion since there's plenty of people who are very willing to teach on here - albeit in bite sized increments.
1
1
1
1
yes exactly, this is partly why it's hard to get good teachers no matter what the institution
also the fact that being a pretty good teacher in a bunch of concepts across the board won't get you hired, but being an expert in ML and shit at everything else will
1
also the fact that being a pretty good teacher in a bunch of concepts across the board won't get you hired, but being an expert in ML and shit at everything else will
yes also true
There’s also name bias, in both people and schools - i.e you might have a research paper with strong results but it will immediately met with more scrutiny than a paper submitted from an Ivy League, or even if you’re the 5th author on that paper, or if the author is a big name in the field.
that's one thing that I don't like from University...
1
@Rudy @demetrispanos @greensleevless @ratchetfreak @bumbread hey! I'm afraid I have to go...
2
thanks for joining @raysan5 ! It was an honor
1
Yea same here (in the middle of moving) - this has been wonderful yall! Thank you!
thank you very much for the fishbowl, it was great! see you around!
It's about the stop time anyway so I'm okay with wrapping up the official discussion and moving anything residual to #fishbowl-audience as we're losing participants
1
All right then, thanks for participating, everyone!
3
3