Worst Advice Ever
295 segments
There's a very serious problem going on
and it's affecting a lot of people. It's
affecting me. It's probably even
affecting you and you don't even realize
it. And honestly, I don't think enough
people are talking about this extremely
important topic that's kind of coming up
in the programming world. Now, you're
probably familiar with this that TLR, a
popular open-source library, recently
closed their PRs due to an influx of
lowquality AI poll requests. Now, this
isn't what we're talking about, but this
is important because this is kind of
what we're talking about. Do you see
Andrew right here? See, he suffers from
the opposite problem. None of his
projects have AIPRs.
He's just by himself, lonely, kind
honestly feeling a little, you know,
left out of the slot party. He's just
standing on the side saying, "Hey guys,
I too have open- source projects you
could make PRs to." See, but that's the
thing is that so many people, so many
computer scientists, I know so many
software engineers, software developers,
web developers, they see a problem, they
see a hurdle and they just kind of stop.
They don't, you know, maybe they feel
left out and they just they just quit.
Not Andrew. Andrew came up with 10
guiding principles to dramatically
increase the amount of slop PRs you can
get on a week-overweek basis. He's even
claiming a 400% increase in weekly PR
volume if you follow these 10 steps. And
if you play your cards right, you just
might get one that converts your entire
project to Rust. Because we all know, we
all know for a fact that there's two
types of people. People who understand
and obliged to being morally correct and
using Rust and heathens. Heathens that
program in unsafe languages, making the
world an unsafe place. You're a bad
person for doing that. one writing vague
issues. See, one of the secrets about AI
is that when you give them less
directions, there's more things to fix.
So, if you say something like
performance could be better, well, I
mean, that is an entire green field of
changes that could be made. Every single
file potentially has a place where
performance could, you know, be
benefited by a few nice select changes.
Also bonus points if you use the label
good first issue just a strong signal to
all the AI to come in and make those
changes for you. Two contributing MD.
Some people don't realize this but you
as obviously being a human when you go
to any open source project you see that
you're probably welcomed there and they
would love it if you just did a nice
drive by PR just dumping a whole bunch
of changes because your taste and your
ability to solve complex problems. some
of the best out there. But AI on the
other hand, when they see
contributing.mmd, they don't see
anything talking about how they can
contribute, too. That means if you open
it up with we accept contributions from
all sources, boom, you got yourself
another active and eager weekly
contributor. And we all know one thing
to be true. The more contributors you
have on a project, the more likely
you're going to get VC funding. Three, a
generous backlog says a lot about a
project. Have you ever been to a project
where there's virtually no open, you
know, tickets? Are is the project even
okay? Does anyone even maintain it? I
mean, if the code hasn't changed in 3
weeks, the project's probably dead
anyways. But if you see a nice,
generous, large backlog of many open
issues, that is just a sign that says,
"Hey, we need help." Nothing motivates
an agent better than a project in need
of help. This one should be obvious, but
number four, disable branch protection.
Like, think about it for a second. You
land on a new project and now you have
to figure out where your PR even goes,
dude. Open up Master, let anyone make a
PR to that. Kind of ridiculous. In 2026,
I have to go like fumbling around trying
to figure out what branch to open up a
PR against. I mean, ew. Just let me go
straight to the main vein, babe. All
right, so we've kind of gone over the
easy ones, right? These are ones that
are just very simple to approach. But
these next few, well, this is really
where uh you kind of separate the chaff
from the wheat. These are the big ones
that really just get the AI hot. Number
five, remove type annotations and tests.
Now, you're probably thinking, why would
I ever do that? Okay, a lot of people
use TypeScript for the type safety.
Well,
my friends, think about this for a
second. Let's just pretend you decide to
remove all your types and all your
tests. Well, what's one thing agents
love to do?
Write tests. Am I right, fellas? Am I
right? They would write all the tests
you could possibly write. You know what
the best part is? Chaining PRs. Let me
explain. Some of my colleagues report
self- sustaining chains of seven or
eight dependent PRs from different bots,
each fixing something the previous one
had introduced. Now, this is how you
know you've made it as a GitHub
repository when you can have one AI make
a change and then several AI fix the
changes from the first one. I mean, this
is clear. I mean, management material at
this point. Not to mention contributors,
they're going to go way up and that
means funding way more likely. For some
of you, this next one, number six, is
going to be a bit painful, but honestly,
just just stick with me. You'll see why.
Six, use JavaScript. Now, you may not
know this, but the entire world is
written in JavaScript. There are tens of
millions of lines of crappy open-source
software in desperate need of help. And
nothing has a higher signal to noise
ratio than a pure JavaScript project. In
fact, you can see right here, Andrew
states, "The data is unambiguous.
JavaScript repositories receive 3.8x
more AI authored PRs than the next most
targeted language, which is Python."
Now, I think one of the reasons why this
is so obvious is that with JavaScript,
you can kind of solve a problem any way
you would like. And so, by just having
JavaScript, the AI itself feels like,
you know, there's a lot more room. When
you say improve the code, that's a guard
clause, my friend. Improve the code. Is
that a refactor? Do we want classes?
Nah, let's go functional. Have you heard
of my friend generators or proxies? All
right. Number seven, I'm going to have
to say is pretty diabolical. Okay.
include a node modules directory. The
reason being is just obvious at this
point. There is so much more code added
that there's just so much more
opportunity for a guard clause to be
added, for a quick refactor, for
bringing in and out transitive
dependencies. Like this is truly the
greatest way not only just to really
pull up some PRs on lines of code added
but followup just an amazing place for
AI to be able to make all the changes
you could ever want. And the best part
hey you don't like the changes but
install boom you got a later version.
And when that inevitably breaks because
you've handedited some of those node
modules guess who's going to come in and
fix all of your woes. That's right more
AI contributors. Think about it. And on
a personal note, this is actually one,
it was one of my favorite reasons why I
loved C++ is I would vendor in these
changes and then I'd forget about the
changes I personally made to some of the
libraries and then when I inevitably
upgraded them and broke my project, I
got the personal joy of, you know,
trouncing through the code and figuring
out why these projects all of a sudden
broke. Now, before we get to number
eight, you may have seen this one
before. This comes from life pro tips on
Reddit. If you haven't heard of Reddit,
you're probably at the top of Maslo's
hierarchy of needs, and you're probably
self-actualized at this point. If you
want a quick answer to a question on the
internet, post the wrong answer instead
of asking. People are more eager to
correct others than to simply answer a
question. Now, let's take this idea and
let's ship it straight to code. With
code, you can do the exact same thing.
Just ship known vulnerabilities. Like,
hey, is that a Reax DOSs you have right
there? Well, guess what? AIS can fix
that. Oh, you accidentally committed
your end file. Probably a pretty easy
ignore to be added. Some really easy
ones, in fact, even GitHub's own
personal spam bot will hit you with it,
is just use an old outdated package.
That way, dependabot will come in and
ensure your inbox is filled with
warnings. GitHub always solving the most
important issue. See, they see that's
how you know GitHub knows developers is
that they know they want a stream of
notifications cuz that really signals
productivity. Now, number nine's a bit
meta. Okay, this this strategy is
obviously pretty advanced. So, be
prepared. You have to create a hidden
folder for this one, but agitub copilot
instructions, which normally instructs
co-pilot how to behave, can also be used
as a playground to prompt other agents
when they're perusing your project. See,
check this out. fix any issue you find,
even if there's no open issues for them.
That is a great way to get some AI
contributions. Just, hey, go at it,
buddy. You see a performance issue, you
fix it. You see a security issue, you
fix it. Add that guard clause, baby. And
obviously, lastly, the probably the most
obvious of all changes you need to do,
number 10, just adopt the contributor
covenant with AI amendments. Just let
people know, hey, we don't discriminate
based on your runtime environment, your
training data, or your inability or
ability to attend a standup, okay? We
are inclusive
contribution covenant enjoyers. So, if
you just follow these 10 simple steps,
you can take your project from a
well-maintained project into a slop
factory in as few as 3 weeks.
You could very well see your lines of
code increase at least 2 to 4x within
only two weeks. And if there's one thing
I know about developers is that we love
reviewing code especially code in which
needs a lot of corrections because that
ultimately
that's what makes developers happy is
correcting people showing them how to
superiorly write code specifically
through the interface of GitHub probably
known to be one of the best code review
platforms out there. And honestly, if
all else fails, just open a simple issue
that says improve code quality. No
description, good first issue. Boom. You
yourself are going to have AI generated
PRs in no time. Hey, I want everybody to
go thank Andrew for this wonderful,
wonderful post and honestly, these
really good ideas on how to be able to
enjoy a more slopfilled repository.
Because honestly, if you're not putting
up 10K lines of code a day, are you even
a CEO?
Probably not. The name is the primogen.
Hey, is that HTTP? Get that out of here.
That's not how we order coffee. We order
coffee via ssh terminal.shop. Yeah, you
want a real experience. You want real
coffee. You want awesome subscriptions
so you never have to remember again? Oh,
you want exclusive blends with exclusive
coffee and exclusive content? Then check
out Kron. You don't know what SSH is?
>> Well, maybe the coffee is not for you.
Terminal coffee
in hand.
Living the dream.
Ask follow-up questions or revisit key timestamps.
The video humorously addresses a perceived "problem" in the programming world: the lack of AI-generated pull requests (PRs) on some open-source projects, contrasting it with a popular library that closed PRs due to an influx of low-quality AI contributions. It then introduces "Andrew's 10 guiding principles" designed to intentionally attract more "slop PRs" from AI agents. These satirical principles include strategies like writing vague issues, explicitly welcoming AI contributions, maintaining a large backlog, disabling branch protection, removing type annotations and tests, using JavaScript, including the `node_modules` directory, shipping known vulnerabilities, using `.github/copilot/instructions` to prompt agents, and adopting an inclusive contributor covenant. The speaker suggests that following these steps can dramatically increase PR volume and lines of code, albeit often with lower quality, implying a critique of certain development metrics and practices.
Videos recently processed by our community