My AI Coding Workflow That Actually Works — Transcript
Full transcript
- 0:00This is my AI development workflow. It's
- 0:03not perfect and it's constantly
- 0:05evolving. And this is why I think you
- 0:07you don't see very many people talk
- 0:09about this, but I do encourage people to
- 0:11make comments on this video uh and just
- 0:14say, "Hey, this is how I do it
- 0:15differently compared to yours. This is
- 0:18meant for bigger software, more complex
- 0:22software. Obviously, you need to
- 0:24understand Git and those types of basics
- 0:27before you get into a workflow like this
- 0:29one, but it will give you really good
- 0:33code. Uh, and I'm going to walk through
- 0:35it. Now, if you want to just be like,
- 0:37hey, I want just one prompt and have it
- 0:39loop a whole bunch and then just spit
- 0:41out something,
- 0:42watch my last video about how I spend
- 0:441.5 billion tokens and realize that's
- 0:47not really a good idea, especially on a
- 0:49big project. you have to have manual
- 0:51gates. Uh so having said that, let's get
- 0:54into my workflow right here. Uh a broad
- 0:57overview for this is you're going to
- 0:59have an agents, a spec, a road map. Now
- 1:02the agents can be machinewide. Spec and
- 1:05road map are repository specific. So you
- 1:08typically make a git repository and
- 1:10you'll put those spec and roadmap files
- 1:13in. And then you plan out and build
- 1:16these files and also kind of understand
- 1:19as uh before you start really coding or
- 1:22having LLM code, you're going to have
- 1:24and instruct it specifically to make uh
- 1:28a testing suite for you and then check
- 1:30that testing suite. And then once you
- 1:32have like the scaffolding or the the
- 1:34shell, the guard rails, if you will in,
- 1:37then you can actually build. A lot of
- 1:40people skip this entire thing or just go
- 1:43I'm gonna go in plan mode and then hit
- 1:46implement plan. And I think that's how a
- 1:48lot of people use LLMs and that's why
- 1:50they get poor results. So how what I did
- 1:54is I built my own repository called
- 1:56Titus-ai. It has an agents, a spec, a
- 1:59road map all with examples in them. And
- 2:02then from there um I pulled specific
- 2:06things depending on some of the leaders
- 2:08in in the space. I think Cararpathy's
- 2:09agents uh is public and I took that and
- 2:12I made slight modifications to my tastes
- 2:15of how I like to code or or develop
- 2:18projects and then have just sample
- 2:20examples and then also a skills folder.
- 2:23But the skills again, don't get too
- 2:26caught up on these things. And remember,
- 2:27they're all plain text. So when you look
- 2:30at like um the agents file, look at
- 2:33these and if there's a rule in here you
- 2:35don't agree with, delete it. Add your
- 2:37own rules. This is pretty plain text and
- 2:40it should follow this. Um, but like skip
- 2:43flattery, filler, ceremon openings,
- 2:46emojis, whatever it is. Put this in here
- 2:48and say do not use this. I want straight
- 2:51working code only. I want you to be as
- 2:54specific and minimal and only touch the
- 2:57things that the task requires. I don't
- 2:59want you to just start going off on a
- 3:01tangent. It's important to get these
- 3:04principles. And also I do have some
- 3:08token reduction. I know I need to get
- 3:10more into that space with token
- 3:12reduction. I've tried other projects
- 3:13like Headspace and uh many others and uh
- 3:16token reduction can be a good thing, but
- 3:19often it does lead to a little bit
- 3:21poorer results. So depending on how much
- 3:23token reduction you're doing, uh you
- 3:26might reduce your output, but it can
- 3:29solve. But don't get stuck on that side
- 3:31of things. I I'm only briefly touching
- 3:33on here. If you're interested in the
- 3:34token specific thing, I typically do
- 3:37targeted token reduction things like
- 3:40giant log files and those types of
- 3:42things. I'll use RTK to cut down on. But
- 3:45I won't use RTK on every task that the
- 3:47LLM does. Uh because often I find it
- 3:50just repeats the task and increasing
- 3:52tokens used, which is the cost of this
- 3:54thing. But I could go through this
- 3:56entire file, but we're going to be here
- 3:58all day for that. I just wanted to kind
- 3:59of just show these are the rules. The
- 4:01spec is in the actual repository itself
- 4:04when so if we go into I think my website
- 4:07has a spec. Uh it'll actually kind of
- 4:10specify hey we're going to be using Hugo
- 4:121.62
- 4:14uh it specifies the version. A lot of
- 4:16people when they build out their
- 4:18projects using an LLM LLM will pick old
- 4:21crap. They will pick old versions things
- 4:23with security problems all that. It's
- 4:26just not good. So specify pin the
- 4:29version that you're on right now using a
- 4:31spec file. And then I also go in I think
- 4:34I even have a linting uh thing kind of
- 4:36just showing the structure of this. I
- 4:39have some automation scripts that fire
- 4:41off through GitHub actions as well on
- 4:42this website. And I thought I'd put I
- 4:46probably should add something here. So I
- 4:50I probably also validate using markdown
- 4:52lint. Uh there's actually something a
- 4:54package on this specific system called
- 4:56markdown lint CLI 2 and it would verify
- 4:59all the markdown so there wouldn't be
- 5:01any formatting errors as well. So that
- 5:03would be an idea. So like when you're
- 5:05going through and doing the validation
- 5:06you're like oh okay I need to do that.
- 5:08Uh another issue like with uh my other
- 5:11project where I have my entire uh guey
- 5:15that I'm building. I decided you know
- 5:17what screw it I'm going to build my
- 5:18whole desktop environment for Linux
- 5:20video coming out sometime soon. um using
- 5:24QuickShell. QuickShell had specific
- 5:26tasks and differences with uh its
- 5:29linting. So when it was linting the QML
- 5:32files using QT6's base one, it wasn't
- 5:35being able to grab the types and things
- 5:37like that. So it was not classifying it
- 5:39correctly. So then the linting would
- 5:41throw errors which would blow up the
- 5:43code. It just not good. So I went ahead
- 5:45and made something in the spec file to
- 5:47say, "Hey, make sure you're always doing
- 5:49this with the types and then run your
- 5:51lint and then it will actually show
- 5:53correctly." So anytime you have a
- 5:54specific spec or something that needs to
- 5:57change, think of it that way. Build the
- 5:59test suite, understand how to validate,
- 6:02say, "This is where uh you're going."
- 6:04And I kind of write this down. I think
- 6:06it's right in here. It shows kind of
- 6:08like the the purpose of the project and
- 6:10all the things in the agents. But in the
- 6:11spec here, you need the problem,
- 6:14intended users, uh required behavior,
- 6:18user experience, architecture, major
- 6:20components is probably a bigger one. I
- 6:22probably should clean this up a little
- 6:23bit. Security and privacy requirements,
- 6:26which it kind of goes to the versioning
- 6:28I was talking about. And then the big
- 6:30one here is acceptance criteria. I
- 6:33always in that acceptance criteria
- 6:35always have it go through some kind of
- 6:36test or validation. You got to let it
- 6:38know when it succeeds. And then for road
- 6:42mapping, this is where you would
- 6:44actually say here's how you'd probably
- 6:46go about doing this. Now, uh you just
- 6:49need to make sure you give it that hey
- 6:51at phase one it will be complete when
- 6:54this happens kind of uh workflow. And
- 6:57then you you break that up and chunk it
- 7:00into different phases. And in each
- 7:01phase, I like to do a giant review. But
- 7:04to showcase where I'm going with this
- 7:09on DWM, here is an example run. I
- 7:13already have my testing. I already have
- 7:15all the specs, road maps, every kind of
- 7:17everything logged out. But I want to
- 7:19change
- 7:21this to show the audio percent next to
- 7:24the audio icon. add
- 7:27uh audio percent next to audio icon. And
- 7:34what this will do is it'll go through
- 7:38find the skill. So it found my quick
- 7:40shell skill. I also put where there's
- 7:44spec and skill. There's some overlap
- 7:45there. So whenever it hits quick shell
- 7:48skill, it will actually go ahead and
- 7:50grab that markdown CLI I just went over.
- 7:52So, just know that the skills can be
- 7:54modified as well to where whenever you
- 7:57hit bash and your bash is doing
- 7:59something weird, you might change that
- 8:01up a little bit. You can do it in the
- 8:02spec sheet for the project or the skill.
- 8:05So, now I can go into any project with
- 8:07quick shell and it'll just remember my
- 8:08tastes uh based on that skill that I
- 8:11made. Now, I think there's some
- 8:15terminology changes here too between
- 8:17claude and codeex. Claude, I think,
- 8:19refers to them as workflows where Codeex
- 8:22kind of refers to them as skills. And I
- 8:25I've noticed a lot of um
- 8:30guides and things I've seen online, it's
- 8:33still such a fresh thing. Don't take
- 8:35this as gospel. What I'm saying today
- 8:37will probably be wrong tomorrow or
- 8:40change, but it's just understanding,
- 8:42hey, this is kind of what each phase
- 8:45should look like. So the the phrasing
- 8:47I'm using might change, but the purpose
- 8:51of what I'm doing does not. So you
- 8:53always have these tasks that you're
- 8:56doing. And it's just so important uh to
- 8:59do them because if you're just trying to
- 9:01do one task and just say run wild, just
- 9:04don't keep going until it's done. It's
- 9:07just going to use tons of money, tons of
- 9:09tokens, and then probably not even
- 9:10produce anything good. add to live
- 9:12install and make a PR with these
- 9:16changes.
- 9:17So, this is going to add it to my live
- 9:19install, restart my quick shell and then
- 9:24create a pull request that can then be
- 9:27verified. So uh even though this is done
- 9:31or well it should be done once the
- 9:32restart happens here the poll request
- 9:35will open up and then we start to
- 9:37approach the next phase that would
- 9:39happen with any kind of implementation
- 9:41and that next phase is the review phase
- 9:43which is really important. This is where
- 9:45unit testing happens. This is also where
- 9:48um I typically pull independent
- 9:50reviewers. Uh I've used code rabbit.
- 9:52I've used uh just straight up at codeex
- 9:56review. You can do claude review. Um,
- 9:59but it needs to be outside of this
- 10:01agent. And that agent is so important to
- 10:05do. And you can see right here, it
- 10:07already reloaded, rated why I'm why I'm
- 10:10live here. So, within the last minute,
- 10:11it's went ahead and changed that. But
- 10:13we're also making that pull request
- 10:14because we need to verify the manual
- 10:17gate. Well, we can see right here. Yeah,
- 10:20it did work. But how did it implement it
- 10:22is what I'm now wondering. like, okay,
- 10:25that was magic. That was pretty cool.
- 10:27That would have taken me hours before,
- 10:29which took, you know, maybe 2 minutes.
- 10:32But now we need to verify. And this is
- 10:34where the last video I went into how it
- 10:37fell apart on me. This is where it all
- 10:40happens. The magic happens at that
- 10:41manual gate, which we're about to hit.
- 10:44As this live starts, it's like, okay,
- 10:46everything looks good. It now displays
- 10:48the 100%. Fantastic. Now, I want to see
- 10:51the poll request. And you'll also notice
- 10:54it's taking screenshots of the screen
- 10:56here and then verifying that it shows
- 10:59there. So you can actually live verify
- 11:01automatically.
- 11:03But I also caution you have that manual
- 11:05gate the poll request at each phase with
- 11:08that criteria we wrote at the road map
- 11:12is what I'm looking for. I don't trust
- 11:15AI enough to just manually rubber stamp
- 11:18this stuff. So important to have that
- 11:20manual gate. So here we go. And while
- 11:22that PR is getting made, there's another
- 11:25part of the setup we need to talk about,
- 11:27too. I talked briefly about that
- 11:30versioning. LLMs always pick old
- 11:32versions that have tons of security
- 11:34problems in them. They've gotten better
- 11:35about not leaking AI API keys and stuff,
- 11:38but it's still not perfect. So, I always
- 11:41enable dependabot and code QL. Oh, I
- 11:45heard it. It looks like that PR is now
- 11:47ready. The cool thing about this is it
- 11:52will also be checking to see if there's
- 11:54any updates because LLMs love to choose
- 11:56old stuff and we want to make sure we're
- 11:59using like an LTS version. If it's using
- 12:00node, it needs to be node 24. You know,
- 12:04that's the current LTS as the making of
- 12:06this video. Now, you might not know that
- 12:08off the top of your head, so you're
- 12:09going to need to look some of this up,
- 12:11but it's really important to have the
- 12:14security of Depinabot and QXL in there.
- 12:18So, I usually go code scanning and then
- 12:20check the projects. Enable code scanning
- 12:22on all your project, I would say, just
- 12:25to see these security errors. It's so
- 12:27important to do and this is free, so why
- 12:29wouldn't you do it? So, here we go. We
- 12:32got that. Currently, I have Code Rabbit
- 12:34hooked up. um not a sponsor yet. Maybe
- 12:38maybe in the next video. So, what I'm
- 12:40going to do is I'm going to just take it
- 12:41from not ready to ready for review. And
- 12:44I'm not going to merge this yet because
- 12:46I need independent verification. So, I'm
- 12:49going to trigger the review here. And
- 12:52then also we could do an at@codeex
- 12:56review. So, let's go at codeex review
- 13:00comment. And this is where AI is cool to
- 13:03me. I don't really care. so much about
- 13:05it creating the code. That's neat to put
- 13:07100% on the volume button. I could doing
- 13:10that myself. But where it's really neat
- 13:13is code review. This is where I use a
- 13:16lot of my data. Like when I'm looking at
- 13:18what I'm using, uh we just had a reset
- 13:21today. I'm already 25% of my max plan.
- 13:24So at a $100 plan, uh doing this type of
- 13:29stuff, it's going to be heavy. If you're
- 13:30a $20 a month user on Codeex, it's gonna
- 13:35eat through that pretty fast. So, uh,
- 13:38that's another thing to consider about
- 13:40doing heavy development work. You're not
- 13:43going to be able to do it without the
- 13:44$100, $200 subscriptions. And I think
- 13:47that's another part of this that needs
- 13:49to be talked about. I think uh, we will
- 13:52be transitioning to local models for
- 13:54some, like I know the new Quinn models
- 13:56coming out that can run on a lot of
- 13:58local machines. It's not going to be as
- 13:59good as these models, but I think it
- 14:02using this type of workflow with that is
- 14:04feasible sometime in the future. Um, and
- 14:08I'm know that this type of thing at
- 14:11$100, $200 is also going to be highly
- 14:14subsidized. I know I'm using way more
- 14:16compute than the cost that I pay them.
- 14:19So, I'm getting a hell of a deal right
- 14:21now, but I know this deal is not going
- 14:22to last forever. So, kind of think about
- 14:25that as well because even if you take
- 14:28the LLM out of this, this is just a good
- 14:31workflow for developing software
- 14:33regardless of using an LLM or not. I
- 14:36know I've said that a couple times, but
- 14:37I just can't emphasize that enough. So,
- 14:39review. Now, if you're doing this with a
- 14:42human, you'd have multiple humans
- 14:43looking at this project. Uh, but as you
- 14:46see, a lot of my usage is actually
- 14:51all code review. I use a ton of code
- 14:54review. I love code review. And you can
- 14:57manually trigger this with just an at
- 14:59symbol. You'll see the eyes. That means
- 15:01it's hooked up properly. If you don't
- 15:02see codeex do the emote after doing a
- 15:05review, that means it did not uh did not
- 15:09recognize it. So yeah, that means you
- 15:14need to go back into your your settings,
- 15:16whether it's GitHub or another one, and
- 15:18then hook it up properly. Also, another
- 15:21thing I like to do as well is local code
- 15:24rabbit. They have a CLI baked in to
- 15:27where you can actually before you make
- 15:28the poll request, if it's a bigger
- 15:30change, I'll like to do a couple local
- 15:32passes with the review before sending it
- 15:34up and making a pull request. Again,
- 15:37very re review heavy, very testing
- 15:39heavy. And a lot of the code written in
- 15:42this, if we look at the actual code, oh
- 15:45my god, this seems pretty excessive, but
- 15:48look at the code that it's made.
- 15:5116 lines added, five lines removed. But
- 15:55if you look at this, this is 11 lines of
- 15:58code. We're using a lot of compute for
- 16:00those 11 lines of code. But what it
- 16:03yields
- 16:04is so good that I'm like, hey, this is
- 16:08fantastic. It's only taken me a couple
- 16:10of minutes. We're using my monthly plan
- 16:13and it's going to town. This is the
- 16:17future of development in my eyes and how
- 16:19I work on my thing. But let's let's see
- 16:22what the code review comes back with.
- 16:24Usually, it will find something on the
- 16:27code review. Uh, this one already found
- 16:29a couple things. You can see uh hide the
- 16:32percent when audio is unavailable. Yeah,
- 16:35that's probably a good idea to do. uh
- 16:38record the new panel behavior in the
- 16:40change log. This adds a user visible
- 16:44panel, but the diff updates only the QML
- 16:48and regression assertion leaving
- 16:50unreleased section and change log
- 16:52without an entry for the new volume. So
- 16:54this is a documentation issue. Uh both
- 16:57are valid though. So let's let's just
- 17:00have AI do those things for us. So what
- 17:03I'll say is address and resolve PR
- 17:07review feedback and commit and push.
- 17:11Commit adds a commit locally. So another
- 17:15thing about understanding git is it'll
- 17:17add it locally and say here's a
- 17:19checkpoint essentially for your code and
- 17:21then a push will push it to wherever
- 17:23that code repository resides. In this
- 17:25case it's github.com. And then it's
- 17:27going to go through that thread, come
- 17:30back, and sometimes the reviews I I read
- 17:33it and I'm like, "That doesn't make
- 17:34sense to me." Or, "Ah, that's a little
- 17:36bit too verbose. I don't really care
- 17:38about the change log. No one's going to
- 17:40look at this." All those things. You can
- 17:42actually make the call here. This is the
- 17:44manual gate I'm talking about. You're
- 17:47looking at the code. You're looking at
- 17:49the comments from the reviews from AI.
- 17:52And this is the assistance. It's meant
- 17:53to be a pair program or just like having
- 17:55an expert sitting right here next to me
- 17:57and going, "Hey, Titus. Yeah, uh, you
- 18:00idiot. You You forgot about what happens
- 18:02if there's no audio. It's just going to
- 18:03have a weird percent up there with just
- 18:06random symbols. What what what are we
- 18:07doing?" And it's those things that
- 18:11that's that's the power of it. It's not
- 18:14that it wrote the percent. It's the fact
- 18:16that it reiterated, reviewed it a whole
- 18:19bunch, and then thought of all these
- 18:21other cases that I I didn't think up or
- 18:25I didn't realize until I would put it
- 18:26out in the wild and then a bunch of
- 18:28people would review it or use it and
- 18:30then go, "Oh, I don't have audio." And
- 18:32then all of a sudden there's a bunch of
- 18:33weird symbols up there and then they
- 18:35would make an issue on GitHub and then I
- 18:37would see that issue and then I go, "Oh,
- 18:39yeah, I forgot to add that." And then I
- 18:41would manually code that in. So that was
- 18:43the old method and it's essentially just
- 18:46like the new method. It's just a a it
- 18:49looks a little different and it's way
- 18:51faster and it reiterates a lot quicker.
- 18:54So that's where the changes happen.
- 18:56That's where the magic of this is. I
- 18:58just I I just want to beat that into
- 19:00people's heads because often I I hear a
- 19:03lot of AI haters just say, "Hey, it why
- 19:06don't people talk about hallucinations
- 19:07anymore? Why don't two people talk about
- 19:09like it's more about the user and less
- 19:13about the model. It's less about what's
- 19:16going on. It's it's how you use it. And
- 19:19and that's the big thing here. You you
- 19:22got to know these basic things to test
- 19:26it properly to check and fix and then
- 19:29understand how the development should
- 19:31look. And you can see our commit in real
- 19:33time come through. Uh, and there it is.
- 19:38Now it'll actually do a review on that.
- 19:41You see it's going through my continuous
- 19:43integration here and going through all
- 19:45the tests. It's doing smoke tests on
- 19:48each distribution, Arch, Debian, and Red
- 19:51Hat Enterprise Linux. You're going
- 19:53through all this. Oh my god, I can't
- 19:55tell you how cool that is. Instead of me
- 19:57having to install and do all these
- 19:59different tests or manually write these
- 20:01tests, it would come through and and
- 20:04should check out all these code raex.
- 20:11You could probably go ahead and take a
- 20:13look at it, but I want to see exactly
- 20:15the changes made to address that
- 20:17feedback right here. If v unavailable,
- 20:21go ahead and make this non-visible.
- 20:23Okay, that makes sense. Then we have the
- 20:26test that was changed to verify that.
- 20:29And then in the change log, add the
- 20:31volume percent to address all of our
- 20:34reviews. And I even got the little ping
- 20:37from Herder. This is like my first day
- 20:39on Herder. So, uh, probably a video
- 20:41coming out soon for that. But we're
- 20:43good. Everything looks great here. I am
- 20:46happy with those code reviews. We have
- 20:48the approval. Wait for my last smoke
- 20:51test. This is actually standing up a
- 20:52whole Docker. probably a little
- 20:54excessive for this type of uh setup, but
- 20:56this is just GitHub CI that I'm burning
- 20:59through. I think we can safely say this
- 21:01is going to go through. I don't we
- 21:02didn't really have any
- 21:05distribution or distro specific changes
- 21:07here. So, you you could technically want
- 21:10to wait for all your CI. Now, I will say
- 21:14doing it this method, you're not going
- 21:16to move as fast. you're going to move
- 21:17way faster than you did a month ago or
- 21:20for me a month ago, but for you maybe a
- 21:22couple months ago when or couple years
- 21:24ago when LLMs just kind of sucked um at
- 21:29writing code, but now you are to a spot
- 21:33where you can move so fast you just make
- 21:36a bunch of crap. But I encourage you to
- 21:40slow things down. Make a YouTube video
- 21:43about this process. kind of go through
- 21:46and you can kind of see, hey, what's
- 21:48happening as it goes through all your
- 21:51testing? Does that make sense to you?
- 21:53Uh, are your reviews coming back? Are
- 21:55you addressing and resolving those
- 21:57reviews? And then this this is great. We
- 22:01now have a volume percent. I actually
- 22:04wanted to add this. I just saw it and I
- 22:05was like, "Oh, that'd be a good test of
- 22:08this workflow." And then we're done. And
- 22:10then I'm just going to delete the
- 22:12branch, prune it off, and we are on
- 22:14main. Uh, now DWM Titus has a percent in
- 22:18the QML. We looked at the code, we
- 22:20looked at the reviews, we addressed all
- 22:21the changes, we've documented all the
- 22:23changes. It makes sure everything's nice
- 22:26and pretty, and it's way better than it
- 22:28was beforehand because a lot of times I
- 22:30just would like, ah, documentation, I'll
- 22:31do that later. And now it forces me to
- 22:34do it now. And it's really good. So that
- 22:37was my final merge gate. And I wanted to
- 22:40kind of give you a real world example.
- 22:41This I put an article down below uh my
- 22:44AI workflow. This is due to change. I'll
- 22:46probably update this article as I go.
- 22:48But remember, slow is smooth, smooth is
- 22:50fast. Don't move too fast.
- 22:55Wait for that testing to get polished
- 22:59and come back green.
- 23:02Wait for those reviews to happen. You
- 23:04might have to wait 5 10 minutes for
- 23:06Codex to review. I've even had Codex
- 23:08take upwards of 20 or 30 minutes to
- 23:10review a big PR. So, think about
- 23:14chunking those into smaller poll
- 23:16requests so it happens a little bit
- 23:17quicker so the work for you're not
- 23:19trying to do too much too soon. Uh and
- 23:22and follow this guide. Even if you're
- 23:25not using an LLM, this is kind of how
- 23:28you should be doing things regardless.
- 23:31Have good development workflow. This
- 23:33will make good software for you. and
- 23:35it's made great software for me. Better
- 23:38software than I made before LLMs really
- 23:41became good. And they're good enough
- 23:43now. Don't be waiting for the next
- 23:45model. Don't be waiting for mythos
- 23:48fabled ultra plus $1,000 a month plan.
- 23:52That's not going to fix it for you.
- 23:55This is here today and it's going to
- 23:57work way better than that will in a
- 24:01year. It's not going to fix you. It's
- 24:04not going to fix your bad on development
- 24:08workflows. It can't just magically read
- 24:11your mind, even though sometimes it does
- 24:14feel like that happens. And with that,
- 24:16let me know your comments. What did I
- 24:18miss? What do you do that I didn't touch
- 24:20here? Uh, I know I didn't miss. Uh,
- 24:22there's still plugins like Super Power
- 24:24and some other stuff I needed to kind of
- 24:26sprinkle in, but this is already getting
- 24:28really long in the tooth and I wanted to
- 24:30make a more succinct video about it. Let
- 24:33me know. I know this stuff. I see a
- 24:35bunch of AI. Oh, I hate AI or it's going
- 24:37to ruin the world and all the politics
- 24:39and this is where it's headed. I don't
- 24:41talk about any of this. I'm just saying
- 24:43this is what can practically be done and
- 24:46that's all I care. What can it do for me
- 24:48today? And as soon as it's not doing
- 24:51that job or it's too expensive, I kick
- 24:53it to the curb. I am a practical guy, I
- 24:57get done. And that's what I'm known
- 24:59for. So, I encourage you to do the same.
- 25:03Take what you will out of this. Take the
- 25:05good bits. Tell me the bad bits because
- 25:07I always like and I like refining this
- 25:10policy to make it as good as possible.
- 25:12And with that, I'll see you in the next
- 25:15one.
About this transcript
This page contains the full transcript of My AI Coding Workflow That Actually Works by Chris Titus Tech, generated from the public captions YouTube serves with the video. The transcript has 4,455 words across 603 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.