The video owner has disabled playback on external websites.
This video is no longer available on YouTube.
This video cannot be played right now.
Watch on YouTube
Unlock AI-Powered Learning Tools
Sign up to access powerful tools that help you learn faster from every video.
Senior Developers are Vibe Coding Now (With SCARY results)
Subtitles (494 segments)
AI generated code is causing some
serious problems. Security
vulnerabilities that are introducing
real threats into applications, sloppy
code, and bloated pull request. This is
what the latest reports are showing. It
turns out that while AI is helping us
write code faster, it's also degrading
the quality of our code. And not only is
it producing more bugs, it's producing
an entirely new kind of issue
altogether. Now, even if you're not
using AI to write code yet, you should
still know about the type of issues that
are out there because in one way or
another, this kind of affects all of us.
In this video, we're going to break all
of this down. We want to look at the
kind of problems these reports are
seeing. We're going to talk about what's
causing these issues, and I also want to
take a look at what we can do to
mitigate these risks.
Seeing how things changed from the start
of early 2025 to the second half of the
year made me realize that I can no
longer ignore this AI shift. I saw the
same senior developers who at the
beginning of the year brush it off as AI
slob or fancy autocomplete start
embracing it like never before. One
report that stood out to me was done by
a cloud provider. And this report
surveyed 791 senior developers all with
10 or more years of experience. And this
report stated that 32% of the senior
developers they surveyed said they had
shipped AI generated code. Now to be
honest, these numbers seem to fluctuate
depending on who does the survey and
what time of the year this was done, but
generally speaking, this number seems to
be pretty accurate based on what I've
seen. Whether you're convinced that this
is a good idea or not isn't really the
point anymore. The fact is there's a lot
of decision makers who are buying in and
for now this seems to be the direction
we're headed in. So what are these
reports finding? Well to start this
report by Veraricode found that 45% of
code generated by AI failed security
test and introduced OASP top 10 security
vulnerabilities into code. If you don't
know what OASP is, it's a globally
recognized foundation that provides
guidelines and information on software
security. And every year they release a
list of top 10 security vulnerabilities
that applications face. On this list
includes things like cross-ite scripting
attacks, SQL injections, misconfigured
access controls, and much more. So these
are not minor issues. What's even worse
is that these results remained largely
unchanged even as models dramatically
improved. Another report which was done
by code rabbit reviewed 470 open- source
GitHub poll request and this report had
similar findings when it came to
security being an issue in AI generated
code. This report showed that on average
AI poll requests had 10.83 issues per PR
while human generated code had 6.45
issues per poll request. That's 1.7
times more issues in AI generated code.
Now, if we break this down by severity
levels, AI underperformed in all metrics
here. When it comes to critical issues,
it was 1.4 times higher. Major issues
were 1.7 times higher. And when it comes
to minor issues, it was nearly double
for AI generated code. Now, let's take a
minute to see what's actually happening
here. So, this report breaks these down
into four categories. We can see that we
have logic and correctness, code quality
and maintainability, security findings,
and performance issues. For logic and
correctness, the two that stand out for
me are going to be the incorrect
dependencies and sequence and
misconfiguration. I see this quite a bit
where I'm working with a newer library
or package and anytime I'm trying to
write some code, if it can't figure it
out based on the latest version, I'm
going to get imports that are based on
an older version or simply just off of
Full subtitles available in the video player
Practice with Exercises
Generate vocabulary, grammar, and comprehension exercises from this video
Comments (0)
Login to CommentSign up to unlock full features
Track progress, save vocabulary, and practice exercises
Interactive Mode
Quiz
Correct answer:
Related Videos
Anatomy of a Scalable Python Project (FastAPI)
How A Nuclear War Will Start - Minute by Minute
You’re Passing Way Too Many Arguments (and How to Fix It)
Two-Minute Case Study - Data & Gabbana: Stitch Fix's Predictive Algorithms
Build Production-Ready AI Agents in Python with Pydantic AI
Movieclips
Quiz
Correct answer:
Quizzes appear as you watch the video
Memory Tip
From this video
Start learning languages for free
