0:00
I do mean this that I love articles that are titled in such a way that make me think that I’m confirming my own bias
0:06
okay like honestly you know I I am not somebody who believes in this idea of being able to be unbiased like I
0:13
recognize that we’re all biased and I just you know sometimes I embrace it so if this is all about students and the
0:19
danger of AI dude I’m so excited because you know the reality just like I mean just just just as warning I think the
0:25
reality is that there’s about to be a lot of younger Engineers that are just getting started they’re going to have the root Awakening of their lifetime and
0:31
so if this article confirms my bias I’m very excited about it but if it doesn’t I hate you article I hate you and you’re
0:37
stupid and you’re fat and You’re dumb okay dear students yes AI is here you’re screwed unless you take action now this
0:43
can be taken in one of two ways maybe this is going to be like if you don’t learn how to PRT now you’re never going to be able to PRT in the future yeah
0:50
we’re going to improve pring by thousands of percents over the next couple years but guess what if you don’t learn it now in the most shitty version
0:56
ever you’re not going to be able to do it later I always thought that take was so stup stupid like is that take not the
1:01
dumbest take out of like out of all AI takes isn’t that just like the dumbest take of all time like isn’t that just
1:07
like the worst of them all if you think that it’s going to literally improve some dramatic amount but that at the
1:13
exact same time you say if you don’t learn it now the worst possible version then therefore you’re going to be left behind
1:20
you’re like what does that even mean if you think that this is the worst possible version of all time what are you saying all right two weeks ago a
1:26
student anonymously emailed me asking for advice uh this is is the reply and
1:31
if I was in your shoes I would do what I do nice no dude nothing like this guy
1:36
believes in taking his own advice okay this is good ad that means it’s probably pretty good here we go so which which
1:42
way is this all right so I read your blog post and oh [ __ ] moment in time oh
1:48
my gosh what is this blog post why is this a picture of vs code what’s
1:53
happening here okay okay I mean now all of a sudden now I want to read that uh
2:00
jofrey Joffrey you’re you’re Huntley your last name’s an adverb um you’re
2:06
making me excited alongside the future belongs to the idea guys that just do things and decided fine I’ll go try
2:13
cursor for myself and see how well uh composer does I normally use co-pilot and some Claude but not much since
2:19
getting the context of my 50,000 line uh project into Claude is rough and I didn’t want to start paying for the API
2:26
use so I could start using ieer I normally get co-pilot for free I heard some claims about cursor and friends
2:32
that assume uh let’s see let’s see oh my goodness about cursor and friends that I
2:37
assumed were pumped up LinkedIn talk so I really didn’t uh think about it until now by the way just so you know like
2:43
real talk 100% real I am I am doing a cursor moment I hope you guys are prepared I am getting ready we’re going
2:50
to cursor hard reading is hard this is true um especially when you don’t get as much sleep I I screwed that all up so
2:56
just like let that go but I am going to do this I am going to do one week unadulterated 7 days straight usage I’m
3:04
going to embrace every part of cursor I’m going to become amazing at it te can
3:10
I announce this are you here te te is streaming okay I’ll just call him hold on tell TJ I’m calling himo all right
3:19
hey how’s it going good it’s going great all right um yes should can can I can we
3:24
talk about our grand idea the which the one that we got information on Friday
3:30
yeah no what why not I know I’m not going to tell it what it is but can I just say that like we’re going to we’re
3:36
going to try something wild uh can we just wait like two days I I wasn’t going to give anything
3:46
away I mean I’ve already said I’m going to try cursor and I’m going to try for one week straight but that’s all I said
3:53
okay okay TJ is so angry oh my goodness dude I know you got
4:01
to get people excited I know can we get people excited on Wednesday I I don’t I like to trickle it
4:08
out so it’s like very little you know what I mean um that what you want my honest opinion
4:16
you should mute me on your stream no we’ll talk we have a meeting
4:21
today anyway so don’t worry I know we have a meeting in an hour and a half just relax I don’t like to
4:27
relax oh my goodness this okay TJ’s actually crying I’m I’m just I’m not
4:37
tilted I’m not tilted but I’m gonna go back to writing O Campbell to make me bye bye TJ is so tilted all right flip
4:45
take it out it’s way too long dude I know I like dude I get so excited TJ and I just planning a super fun event coming
4:51
up he’s dude he’s tilted he’s so angry dude he is so angry right now TJ just does not like leaking things okay I’m
4:58
sorry for leaking anything I will I I I trust me it didn’t happen nothing happened don’t worry it’s not real uh so
5:07
don’t worry it’s not real everything I said is fake the end okay so I read your
5:13
blog okay so let’s just start over so I read your blog post oh let’s see ah oh [ __ ] Moment In Time alongside the future
5:19
belongs to the to idea guys that can just do things and decided fine I’ll go try cursor for myself and see how well
5:25
composer does I normally use co-pilot and some Claude but not much getting the context of my 50,000 line project into
5:32
Claud is rough and I didn’t want to start paying for the API use so I could start using iater uh I normally get
5:38
co-pilot for free so I assume this is a student talking right because this is the student sending the message I heard
5:43
some claims about cursor and friends that I assumed were pumped up LinkedIn talk so I really didn’t think about it
5:49
until now by the way as a student I’m not exactly sure if you even have the wherewithal to be able to judge how good
5:56
or bad something is just throwing it out there you know it’s taking a it’s taken me multiple decades to really understand
6:02
the good and the bad I’m not precisely sure if you can judge as accurately as you might think you can you might get
6:08
wowed oh you know sometimes it’s easy to get wowed by Magic you know what I mean LinkedIn is creepy really creepy uh for
6:15
example I’ve been working on a little jni model for my Android app uh but it has a bunch of weird memory errors
6:21
because managing the jni resources is borderline impossible and C++ Ray has piles of insane foot guns so I planned
6:27
to rewrite it in Russ so I could 100% isolate the memory unsafety this was a tall order it was a three-way
6:33
integration between bulky C++ library with its own cmake setup the jni code of my app it required some really messy
6:41
translations from the rust trait into C++ abstract class it would take a lot of in let’s see a lot to integrate into
6:47
my app I decided it was a good enough Benchmark to see if composer was up to Snuff okay this could actually be a good
6:54
generally speaking this could be an okay usage meaning like this is kind of what AI is pretty good at just just like it
7:00
can get like 90% of the way pretty good is just like I have language a can you make it into language B like it’s it’s
7:06
like it’s it’s not it’s not terrible at that generally speaking it’s like really not that bad first go round was not well
7:13
I gave the lump sum of all the work I wanted and it choked and went around in circles I tried to go back to debugging
7:19
the memory problems manually but honestly I hate C++ and decided to go another around with composer and the
7:24
rust module second time I sliced up the task into smaller but still pretty big pieces and fed them one at a time into
7:30
composer this time Co composer probably wrote 99% of the code porting the previous shell script uh making the
7:36
shims I needed to map a trait to C++ abstract class porting the manual jni
7:41
code and the massive build script linking all together it debugged its own work and when I pasted the error traces
7:47
in sadly since the rust module was short uh shoehorned into my app uh into my app
7:52
composer agent couldn’t really test on its own and it just worked there was some stuff uh it couldn’t do mostly
7:57
choking on big air traces and incremental build bugs thanks Gradle and sometimes I needed to point point it in a better Direction architecturally but
8:04
otherwise I barely touched the code all I did was direct composer to do what I was thinking of and then just killed
8:11
time while it did the work I am now let’s see I am at a loss now I’m a
8:16
student fresh grads and interns aren’t uh seen as idea guys that can get stuff done they are uh they are seen as
8:24
useless ticket monkeys to be herded by seniors now cursor composer is the ticket monkey Pres presumably to be
8:30
hurted by a senior not me and maybe even the senior isn’t needed if the scale
8:35
enough Reasoner model can do the task slicing I’ve already felt like I’ve been shuffling chairs on the Titanic talking
8:42
about projects and internships as if this threat doesn’t exist and honestly it confirms it is there literally any
8:50
reason why I shouldn’t just throw away the 10 years of my life and do I don’t even know what uh manual labor is
8:55
probably going to be a very important task over the next while dude being a plumber is probably going to be like actually a fairly lucrative thing do you
9:02
know that the do you know what the do you know what the median age of a plumber is median halfway 57 57 57
9:11
that’s in America that’s probably not true in in in Europe or other places Source uh the largest uh aggregated data
9:17
source uh that someone talked to me uh about and it was awesome and we had lots of talks about data and all that five
9:25
seven so that’s like kind of scary right so anyways alternatively if you really
9:31
feel like these things aren’t working out just so you know like maybe a plumber actually is like not like that’s
9:36
it’s it’s like really probably not a bad one 57 I I mistyped uh 57 at 75 deal
9:43
with it at this uh at this development Pace I don’t even know if I have time to Pivot before everything else gets solved
9:48
by 04 or whatever what do I let’s see what do I even do anymore thanks Anonymous student it’s kind of sad by
9:55
the way how do you how do you email anonymously oh you just have like an anonymous email account is that is that just what he did he just got his
10:02
professor’s email and just Anon admited from like puppy lover 69420 got him
10:07
reality is literally a South Park yeah that’s a strongly worded headline it’s just facts I’m a straight shooter I’d
10:13
rather tell you uh straight and provide actual advice than uh plate feelings
10:19
okay so let’s let’s now we’re into the article how are we feeling the step that you take now will determine your success
10:25
rate with obtaining a sweet role going forward if you are a high auton aous person then you’re not [ __ ] as long as
10:31
you take action okay it uh has all happened before the software development industry Works in Cycles this is the
10:37
third bust after a boom uh that I’ve weathered through understanding what happened in 2000.com bust is important
10:44
because it led to the 2017 boom so let’s rewind it’s now 1998 oo books a million
10:53
books a million dude uh books a million stock price historic let’s see if we can
10:59
pull up some historic stock price let’s look at some let’s see if we can pull up some some bam some bam is
11:06
there a way to pull up like 1999 stock prices uh1 1999 oh price history let’s
11:12
see if we can pull it up here all right can we go back further can we go back further whoopsies down down meant to go
11:19
forward in time wait really can I just what kind of date picker is this oh it only goes to two thou oh wait oh my
11:25
[Laughter] gosh let’s go to 1997
11:31
January and let’s go to man this is some this dude our jobs are so over we’re so
11:39
over uh 2002 January there we go apply let’s see if we can oh my gosh why
11:47
why aren’t you letting me go through is it not able to is it not able here let’s go uh weekly
11:54
monthly daily I see your resetting nice okay forget about it books a million
12:00
became like a multi-billion dollar company and then quickly fell back down to the nothing okay I don’t I don’t want
12:06
to do it I don’t even want to use that website it reset the date it’s stupid uh but nonetheless it was like the greatest
12:13
bam is uh so bam is nothing bam is nothing so during this time bam became like one of the most sought after stocks
12:19
Qualcomm went all the way up to 900 and some points and fell all the way down to like 30 points or some ridiculous number
12:25
like that in a matter of a couple days so like people don’t realize how bad the crash was in 1999 2000 like it was
12:33
insane all right it’s now 1998 and if you know how to make a website using Dreamweaver or
12:38
how to Cobble some Pearl together and use rsh to connect to a sun e4500 then
12:45
you’re guaranteed to role as a software engineer after a massive runup of hype and uh and an injection of too much
12:50
Venture Capital Money resulted in a talent Market where there weren’t enough software engineers in the world also I
12:55
mean there’s a whole other things that happened um the grid just couldn’t handle that type of traffic right you couldn’t make a website that could
13:01
handle um it’s just like everybody was on 56k or slower very few people had
13:07
like you know uh the what was that the 128k what was that one called the ti or
13:14
why am I blanking on it right like you had very little right you had ISDN right T1 there you go T1 like people didn’t
13:21
have T1 or any of that t T3 any of that right life was great that is until it wasn’t the bubble burst in 2000 and an
13:28
engineer called Philip J PUD Kaplan chronicled the downfall on a website [ __ ] uh [ __ ] company.com to put into
13:35
perspective how important and meaningful this website was at the time is the equivalent of Hacker News or uh uh Slash
13:41
dot report a new [ __ ] suddenly all the jobs overnight dried up students had just graduated and were looking for a
13:46
sweet sweet summer engineering position were up shits Creek uh just like you are now if you don’t take action seemingly
13:53
overnight the employment switched from sellers Market I.E in employee favor to buyer Market in employee employer favor
14:00
because there was an over supply of senior engineering Talent desperate to find their next job why would a company hire a graduate when they could obtain
14:06
someone with 5 years of knowledge on the cheap it was let’s see it was a covid-19 moment in time if you were born in the
14:13
wrong year I.E TW uh year 12 high school students disrupted in Australia then your life was utterly turned upside down
14:21
that generation of graduates simply missed out on opportunities and core life experiences but there were some
14:26
nice outcomes from it as it flushed out all the folks who were in Tech just for the dollars I mean that’s always going to happen right anytime there’s money
14:33
people are going to join in for it and I don’t necessarily think it’s bad I actually think I there’s something
14:39
admirable uh about doing something you dislike or you find difficult but you want to make a better life so you work
14:45
extra hard to figure it out like I actually enjoy working with those people to be completely fair if they’re if they
14:51
got if they got in it for the money and like worked really hard to get there there’s something pretty good about that
14:56
now at some point you have to figure out how to more like you have to figure some things
15:02
out and make it into a craft and make it really really good but i’ I’ve generally enjoyed I’ve I’ve never had a huge
15:07
problem with it now there’s people that are just that are just not good they just don’t care I’d rather have someone
15:13
that cares than someone that doesn’t care right the folks who stayed entered were those genuinely interested in computers because why else would you
15:19
enter a field that had very few job prospects yeah I actually entered into computer science in 200 uh five and I
15:26
didn’t even know that there was no jobs I thought yob I didn’t I didn’t even know that there was a yob market or no
15:31
yob Market I had no idea it took many years to recover from the bust but here’s the uh kicker because there was
15:36
an under supply of graduates as companies weren’t raising the Next Generation it led to a shortage of talented software engineers in the
15:42
market the result was inevitable the market switched sides to where the employees had all the power and due to
15:47
another massive round of venture capital injection money was freely flowing every company out there uh wanted to attract
15:53
Google pedigree Talent companies everywhere started offering perks and compensation packages to tempt them to
15:58
come join their company I do remember that I do remember this like this was not this was not crazy this was not too
16:06
long ago right this was a real thing and things were just wild word got out and suddenly everyone wanted to become a
16:11
software engineer because the money was absurd YouTube is now a thing all sorts of grifters turned up selling courseware
16:16
on how to get a job and boasting about their Millions which resulted in even more people becoming software Engineers
16:21
just for the money because it was an easy path to fast retirement so there’s a difference though there’s a difference
16:27
between wanting to get a job that pays well because you want to be able to establish a good life for yourself and
16:32
thinking that there’s a all I have to do is do this for three months to be able to get a multi-million be able to get millions of dollars right those are two
16:39
different people those are two very very different people one’s like I actually
16:45
want to develop a lifelong career and I want to be able to get money to be able to provide for my family and there are
16:50
just like like the amount I still what’s his name David a brag right David a
16:56
Bragg this guy Tik Tok man I dude Tik Tock man this guy just was nonstop about
17:04
like just all of it just like here’s the simplest way to do it this is how you do it I just never forgot this guy’s name
17:11
because his name was he has a a nominal determinism inside of his own name his
17:17
own name was nominal determinism right the name’s brag what are you gonna do
17:23
brag about how to get like brag about making a lot of money on Tik Tok like dude there’s something so funny there’s
17:30
something so funny about about that oh my gosh it’s beautiful it’s just beautiful oh and Tech lead yeah why I’m
17:37
a millionaire as a millionaire man did he fall out of like all people to fall
17:43
his early videos were pretty funny and pretty good like honestly this this title why I’m a millionaire as a
17:49
millionaire is like a hilarious title and this could have been a hilarious video but man did this guy turn into
17:56
just such a grifting douche uh it’s now the year 2023 and after a strong 14
18:01
years of growth the industry fell on its face again interest rates went up and parking money with risk at Venture
18:07
Capital firms was no longer a desirable place for investors because they were other Alternatives now available at
18:13
lower risks fair I’m not sure what this uh diagram honestly I don’t even know what this diagram is can somebody is
18:20
this supposed to be interest rate is this the interest rate yeah I think this is interest rate uh wilst I managed to
18:26
avoid original do dust or bust this uh this one affected me personally I was
18:31
one of the software Engineers caught up in the 29,7 layoffs and it took me with uh my
18:37
experience 6 months to find a suitable replacement role if you graduated from University or entered the job market in
18:42
2023 then you were competing against me or or a Circa uh 220,000 qualified
18:48
employees with experience experience I.E screwed I’m not re I I personally don’t
18:53
think uh zero interest rate is the reason why all this stuff fell I I do think I’ve liked this article so far I
18:58
like kind of how he’s building up to this but uh I do think that this is actually the incorrect um the incorrect
19:04
reasoning as to why things fell apart like I do I do agree because
19:10
here’s the deal well first off money was almost free for a decade beforehand and jobs weren’t out of control money was
19:17
free during this time but somehow it was out of control one might have a different kind of guess which is that
19:25
when the whole world had to go online and Amazon all of a sudden quadrupled their
19:32
traffic every place quadrupled everything due to covid and people like like government’s literally saying you
19:38
can’t leave for some period of time like this happened in many people’s governments like the the use of the internet greatly increased and so
19:45
there’s this huge influx of money and need for talent and demand and all of a sudden anybody who could program was
19:51
like a good enough hire to make something happen because they just needed boots on the ground apparently like at least that was the thought
19:57
process and so now that that’s over and it turns out free money ain’t free money forever and people that and it turns out
20:04
people don’t want to stay inside people actually like touching grass this thing
20:09
exploded and all a sudden people like guess what those jobs are going to kind of kind of Go With It notice that uh
20:16
like I believe if I’m not mistaken Amazon they’re only Rec catching back up to the levels in which they were
20:23
beforehand like they’re only now starting to earn back the same amount of money and so it’s like it is a very different World okay now it’s 2025 AI is
20:31
at your doorstep the employment Market has not recovered from the last bust you’re lucky uh you’re lucky in a way
20:36
because you realize early what’s coming and you have time to take action there are many software Engineers who simply aren’t going to make it because they
20:42
haven’t had the Epiphany that you have and haven’t started progressing through people uh through the PE the people
20:48
stages of AI adoption let’s see the six stages of AI adoption by software developers that’s not good enough prove
20:55
to me that this uh isn’t hype uh experimenting with l LMS dear in the headlights uh will I have a job in the
21:01
future alarmed we have Let’s see we need to bend our planning and change priorities engage in consuming AI engage
21:08
realization that llm could be uh programmed and doing it um okay so the
21:14
this is the whole act now okay so this is the whole act now or else you’re screwed to take action kind of thing
21:20
okay so it’s the opposite article this article is in fact not confirming my bias so therefore I’m in pain and I hate this uh right therefore I’m at pain and
21:28
I hate this uh but but the again this this you know I made the exact same argument this whole like
21:34
you’re going to get left behind thing it’s just like yeah but just just take a step back they said the exact same thing
21:39
two years ago when jippy 35 came out can anybody realistically tell me the the
21:46
skills you learned with jippy 35 are they applicable at all in today’s world
21:52
no they’re not like like virtually nothing
21:59
right the way you prompt has greatly changed the way tools integrate with
22:04
with how you prompt have greatly changed right like learned nothing with jiy I’m
22:09
just saying the pring skills right like everything has greatly changed since then so it’s just like this this idea
22:14
that you’re being left behind I’m still not buying it because I think in another five years the way you use AI is going to probably be so vastly different than
22:20
it is now that it’s like it does not matter that you use now right a brilliant coworker penned together the
22:27
following chart and in the following words which explain the impact of AI on different levels of engineering
22:32
interesting I think I would have had a slightly different I guess my expectation would have been different
22:37
honestly I I I honestly I would have had a much different uh expectation for this graph so here I’ll give you my version
22:44
of the exact same graph right like I know it’s not quite right obviously it’s not quite the same and I understand that
22:51
I’m not even trying to say that they’re necessarily wrong I’m just saying that like sometimes I have like a different I
22:56
guess perspective maybe than that I guess that’s how I always viewed it
23:02
right I always thought it was this but I can be wrong I mean I’m glad
23:07
to be wrong uh you’re just starting uh your work in a new code base and you’re still piecing together a solid mental model of how things actually work here
23:14
an llm is a lifesaver stuck on an eror an error an llm can give you an explanation that makes sense need to
23:19
write some code for a uh minor feature or do a library upgrade all of this can be done much faster with an llm and llm
23:26
can already feel like it can do a huge part of job for you by the way I do think there’s something nice if you can
23:31
get an llm to search your code base and tell you why an error is happening and then you figure out how to fix it I
23:37
actually think that that could be a very big debut uh the debugger like you still need to debug and all that uh just
23:42
because you’re if you’re super unfamiliar it’s just a it’s just a an exercise of search and so it’s like if
23:48
if something can help you zero in on something better then like hey that sounds good right I’ve been in I’ve been
23:54
in a 7 million line code based in C++ it ain’t fun right that sucks an llm can
24:00
already feel like it can do a huge part of your job for you that’s why I believe there’s a real danger zone here if you lean on an llm as a shortcut to get
24:06
unstuck in the same way as you reach out for your more senior colleagues then you otherwise have to ask uh then fine in
24:12
the real world chances are you won’t have the luxury of avoiding llms even if you wanted to however if you end up copy
24:17
and pasting code back and forth between your IDE and the llm without truly understanding what’s happening or why then advancing your engineering skills
24:24
will become a serious challenge okay wait a second this article actually coming backwards and and agreeing with me all
24:31
right actually the article there we go we’re back on we’re so back all right
24:36
that’s what we’re talking about that’s what we’re talking about I knew it I knew this article confirmed my bias I
24:42
just knew it uh mid-level engineer you’ve built up a fair amount of context and can navigate your code base with
24:48
confidence you still find that llms make you write code much faster you can ship features uh faster with co-pilot
24:54
completion that I do not understand I again like I said I just turned on co-pilot for the first time in a year
25:00
and I found it like the suggestions are are bad they’re like they’re genuinely
25:07
bad now maybe there’s I bet you there’s other co-pilots maybe cursor’s version of co-pilot is better but it’s like it’s
25:14
bad it’s like genuinely bad I was shocked like co-pilot back in the day
25:20
was really good it’s really bad you’re not mid-level though I don’t think my level
25:26
has anything to do with something being wrong categorically right like this was just really bad it was like I was
25:34
continuously going whoa whoa whoa wait a second how’s it getting this no that’s no no no no I can’t use this no no
25:40
that’s not even close what am I doing no I can’t like there was like maybe like 20 lines of code that were actually like oh nice it got it thank you co-pilot but
25:48
I found that I probably fought against for every one line of code I probably fought against three and I was like that can’t be right this feels very like a
25:55
lot of mental overheads going into writing code that I never had before is co-pilot context aware codebase aware it
26:01
it does a pretty good job right like I don’t think it’s good to be you shouldn’t have to fight against code
26:07
that’s being generated right that doesn’t feel right that feels like that’s a fundamental Oopsy Daisy however
26:13
so so I’m willing to admit that that like co-pilot was just wrong I do think single line completion is the best I’ve
26:20
kind of described this a couple times I think the weather is the greatest uh analogy for co-pilot or these type of uh
26:26
llms right now meaning that you start off if you ask for if you go to the weather machine and you say hey what’s
26:32
the weather in 30 minutes its accuracy is going to probably be within you know Point like a tenth of a degree right
26:39
it’s going to be pretty accurate and so I feel the same way about same line completion same line completion should
26:44
be the same thing as asking for uh near uh uh neartime weather right like
26:51
what is weather going to be in an hour and so it should be pretty near accurate but co-pilot was just not accurate at
26:56
all but as you grow as the more things you try to get it to do it’s like asking
27:02
for uh the weather further and further out to the point where it’s just kind of like it might be right it might be even
27:07
close like maybe you thought it was going to be 55 but then it ended up being 45 but it’s like it’s like kind of
27:12
in that area I’m doing Fahrenheit obviously the the superior and objectively better uh weather measurement system but it’s just like
27:19
it’s somewhere in there right it’s just like okay it’s going to be warmer today and sometimes it rains when it said it
27:25
wasn’t supposed to however you’re already bumping into the case that LM simply can’t handle yet it won’t decipher what the customer actually
27:31
wanted from the ticket uh you were given it can’t use your debugger to pinpoint a dangerous race condition it can’t help
27:37
you uh much when you’re responding to a midnight on call alert you’ve okay senior engineer you’ve got a great
27:43
mental model of the whole code Base by the way this idea that a senior engineer and also the like whole mental model of
27:48
the code base this isn’t a very apt thing a senior engineer and its ability to navigate a specific codebase isn’t
27:54
really one to one you could be a brand new senior engineer at a new at a new place with zero mental model but you’re still a senior engineer right um you’ve
28:01
got a great mental model of the whole code base you’re responsible for you know all of the ins and outs hell you probably wrote a decent part of it sure
28:07
you can code much faster and you really enjoy it but how much time do you really spend writing code when you work on the road map it can’t really help you much
28:14
when you dive into a weird heisen bug it can’t really help you it gets confused and when you’re writing extensive design
28:20
documentation for your next project it can only help you with the formatting and structure not the hardest part the
28:25
substance it just doesn’t have all the nuance and context you’ve accumulated in your head and even if you wanted to you couldn’t write it down many of your
28:32
friends and colleagues are excited and you want to be excited but you just can’t the AI is simply not there yet
28:38
this is probably the level where the most skepticism about llms come from and the more techn technical or unique your
28:44
domain is the stronger the disillusionment okay that’s it’s honestly this is honestly not a bad example right I think this is genuinely
28:51
pretty good it’s it’s this is I would generally agree with this statement and this is true like if you’ve ever just
28:57
green fielded a project and you want to build something and you’ve built the same thing like maybe kind of a few times so you have this
29:03
like general idea of how something should be built like let’s just say that you’re going to build a um like me right
29:09
now I’m building my second kind of reverse proxy for a game and we’re going to build a true Vim game experience with
29:15
that in my head I kind of generally know how I want to format it and using a AI
29:20
it just like I have to take what it does and I have to redo everything that it does and I have to set it into a new way and so it’s just like massively
29:27
inconvenient use that’s it it’s because I already have like I already have all the context and what I want to get done
29:33
with it and so its ability to approximate is pretty poor all right staff plus engineer while there are many
29:38
staff archetypes out there one thing is common between them all your role is often a uh to light the path for others
29:44
to follow to achieve this you have to experiment a lot here’s where llms can start shining again writing proof of
29:50
concept projects has suddenly become much easier if you need to show the feasibility of taking an approach and an LM can help you create a half-baked
29:56
barely working solution much faster than without it and the best part is that once the llm gets stuck you can very quickly get unstuck using extensive
30:03
domain knowledge in your brain so this is actually where I’m going to Full Stop think that this is the worst take of them
30:08
all and here’s why if you are a staff plus engineer and you are making these
30:14
decisions it is really up to you to get get in there and really understand
30:20
exactly why and how something should work it should not be that an llm can
30:27
generate it quickly you should really go okay I have seen a thousand problems I’m going to try to program in 10 different
30:33
things and really get it to such a deep level that I know exactly how it’s going to scale as we code and I can foresee
30:40
the architecture we’re going to want to be able to scale this into a big thing and so this idea that you just want to
30:46
be able to prove a concept something to make it work it’s just like no that’s not that doesn’t seem right at all you’re trying to make decisions that are
30:53
going to scale to a million lines of code you need to understand it at such a depth to go this is a good step
31:00
forward this just feels like a really bad way to make that decision like oh yeah I could whip it out with a uh an
31:07
llm and yeah I could make something that looks pretty it’s like I don’t know if that’s I don’t think that’s the way you
31:12
want it it’s not about the speed of whipping it out my friends it’s the speed of going this is a great
31:18
architectural decision to base a a million lines of code on it’s a much different role so yeah I just I simply
31:24
disagree with this one completely does that muscle mil say cope it does it’s cope power that’s why I’m so good at
31:31
this is I have cope power right hard agree also yeah Jeremy I figured you would agree with that one because it
31:38
it’s it’s it’s it’s evidently wrong right it’s evidently wrong honestly I
31:44
think they should redo some of this I think he I think this geophrey fella has a lot of things that are correct I think
31:50
they’re just incorrectly perspective right his perspective goes something like this where the uh the level of
31:56
engineer is to the familiarity and Direction you put into a code base which
32:02
it may or may not actually be true right and so this is kind of his perspective window when I’m saying no we should look
32:08
at it from a different perspective right the perspective should be along this vertical axis instead which is you
32:15
regardless of your skill level is completely unfamiliar with the code base therefore an error is happening can you
32:22
use an llm to be just a better rip grip right can it say hey here here’s the
32:28
three places that I think could be okay let’s get the debugger out let’s walk through the code let’s find out if this
32:34
is the place where the bug or the thing is happening I want to understand how logging Works can you please like help
32:40
me go through and like maybe because like the one thing I remember working on that I actually kind of wish I had an llm for was looking at logging in which
32:46
logging was a huge amount of macros in C++ and it was really hard for me to
32:52
follow I actually kind of wish I had an LM to be like hey can you just start breaking this down for me let’s start walking through this code I really just
32:59
can’t keep my head with all these macros like I just don’t have enough nightmare Fuel and context for me to be able to do
33:04
this help me walk through this right I’m a senior engineer help me walk through this and then as you get more familiar
33:12
it’s like it should be like your usage should wne because you’re like I’m solving the million line code problem
33:18
over here this million line code problem over here I need to figure out what architecture decisions we need to make
33:24
so I need to understand what’s going to happen in a very intense and intimate way and so this could be a senior
33:31
engineer this could be a junior engineer this could be a mid-level engineer this could be a staff plus engineer level
33:36
person right they can be at this location right like there’s nothing in here that says that an experience level
33:42
can’t be unfamiliar with something and so that’s kind of how I’m looking at this which is just like I think the
33:48
perspective the identification of some of the things he’s saying is absolutely correct I think it’s just simply off by
33:54
90° and so hopefully this sweet diagram makes it a lot more understandable yeah that was a sweet by the way that was a
34:00
sweet diagram and you know it was a sweet diagram so you can shut the hell up companies closing the door are
34:05
closing the doors on Junior uh senior plus uh or staff plus senior and mid-level Engineers who invest in themselves are now more desirable
34:12
because they can use their expertise to Output multiple factors more code and thanks to wisdom earned over the years or tens of years they’re able to uh
34:18
obtain better outcomes from AI as they have experience to know when the AI is bullshitting and when they have deployed
34:24
or developed taste for what uh for what does and does not look right there’s that taste word again straight out of a
34:30
a y combinator um also multiple factors more code I don’t know if to me that
34:37
actually is an alarming rate right when I see a lot of code being generated I get more worried than anything else just throwing it out there I find more code
34:43
not necessarily good code okay that’s not a good that’s not necessarily a good code there not a good you know companies
34:50
don’t even know uh how to hire anymore with hundreds of thousands of dollars at stake all incentives are there for
34:55
candidates to cheat the video blows uh is one of the many tools that are now exist today oh yeah we already went over
35:01
this this is this guy oh wow they actually flipped the video by the way uh I actually had this um I actually had
35:07
this video up on my channel for about 3 hours when I realized that the video I linked was taken down Roy apparently
35:14
took down the video that he uploaded and so I took my video down too uh just
35:19
because I felt like if he didn’t want it up or he had a reason to take it down I would respect that so I think all the
35:25
people that co- uploaded it I don’t know if that’s necessarily a good move right I kind of feel like you’re not you’re
35:30
not really respecting the man you know what I’m talking about the software interview process was never great but
35:35
it’s taken a turn for the worst as AI can easily solve anything thrown at it including interview screenings another
35:42
co-worker of mine recently pended a Blog uh post below which went viral on Hacker News I highly recommend reading it AI
35:48
killed the tech interview now what I mean I like that idea I do think that we we do need to have some level change
35:53
company’s business models are in Jeopardy it’s now incredible uh it’s now incredibly easy to clone any SAS company
35:59
out there in existence if you know how to drive an AI with expertise I’m leaving I’m leaving that’s dumb no
36:05
[Music] no sorry I just had to I just had to leave I just had to leave for a
36:11
second I just had I just had to get up and I just had to walk out there’s just just like this is a
36:18
this is a bad statement unread the sentence If X released 100% of it its code and made
36:26
it completely open source that anybody could copy and do it they wouldn’t it just wouldn’t work most SAS companies
36:33
aren’t simply just the code there’s also relationships things they’ve built over
36:39
time customers that they have gained uh what’s it called um the reliability that
36:45
they have put into their software right the infrastructure the ability that customers feel like they can trust this
36:51
company overall support ability to like meet uh customers where they are and the
36:56
ability to be able to like features that they actually need right it’s just kind of like it’s it’s wonky that you would
37:02
say the statement ever right Twitter’s down right now yeah I’ve seen that which is I mean there’s something that’s kind of funny about the fact that Twitter is
37:08
actually down right now as I say that Twitter is down dude I wonder I wonder what the hell’s happening over there all
37:15
right AI powered B2B SAS apps are in trouble instead of paying almost two $20,000 I made an AI powered chat bot in
37:21
less than 3 hours using only cursor
37:26
nice okay okay one that does the Discord bot we’ve got one that actually takes
37:32
care of the prompting just really basic prompting we’re not even doing any chunking here we literally just grab all
37:37
the docks and stuff nice oh I paused it uh a founder of a YC
37:44
backed AI customer support Discord bot quoted me uh $177,000 per thousand questions per month okay well if
37:50
anything I mean I agree so hold on let me let me back this up so what the Sam
37:56
Hogan fell is trying to say here um is that he was able to just kind of replace this $20,000 thing I’m going to go with
38:03
a separate thing to say I’m going to say that this bot that they’re attempting to
38:08
sell for $177,000 a year is a shitbot anyways AI customer support on Discord
38:15
dog you’re telling me that you’re just Reading and Responding to messages for
38:20
$177,000 a year like dog that’s what kind of technical Mo are you talking
38:25
about that’s just I mean this is just if anything this goes against your entire
38:30
point which is that the hype cycle of AI is so out of control people are saying that this is worth to be invested in
38:37
right are you are you kidding me are you kidding me the mode is underwater I feel
38:43
like I’m living in a in a troll yeah I know you’ve already read the article below uh this was hinted at so I’ll
38:49
supplement with the following wisdom I suspect the future uh of work is going to be lots of small 10 person companies
38:54
operating similarly to how law firms work with profit sharing between senior Partners uh so when does this story get
38:59
good by the way I think this I think most companies could could do this already like if we’re going to be real here I know a lot of people dislike dhh
39:08
but that guy was so good with his people that he hired and all that like I thought I believe at one point I don’t
39:14
know if this is true I just heard this uh and I just read about this just recently so I’m not sure the exact accuracy of this but it was
39:21
approximately $400,000 was the lowest paid employee due to how they did profit sharing and everything so it’s kind of
39:28
like I actually agree with this I think that more companies should try to shoot for this dhh David uh David H hman uh
39:38
Hansel hman H heizer haniz Hamer David Hasselhoff Hasselhoff Hamer the Ruby guy
39:45
I can never remember his last name anyways uh hunam Miner Hansen there we go Hamer Hansen dude he races like a
39:52
race car driver um if you let’s see if so like actually I think that this would be the idea way to set up a company
39:58
anyways right you don’t just have employee like I do you know I know how people say like the whole like company
40:04
thing or company’s not a family thing is stupid like I I’m I’m generally on that team but if you just have a 10 person
40:10
company and you pay people super super well like I do think that you can purchase some level of loyalty I think it’s true right like if a person knows
40:17
that they’re coming in they’re a key player they can make a lot of decisions and they’re paid very very well I think they’ll be generally a very very uh very
40:24
um engaged employee even if the CEO is making millions right right loyalty is
40:31
100% purchasable it’s not 100% purchasable I think that it is possibly purchasable I think you can incentivize
40:37
more loyalty that way right um so when does this story get good if you
40:43
graduated last year and are entering the workforce this year then it doesn’t unless you take action it’s covid-19 moment life in again really all right
40:51
okay if you are a student who has just started University and will be graduating four years from now there will be software engineering rle
40:57
although they’ll be different but what’s likely to happen is for companies close uh close their let’s see if more
41:02
companies close their doors to Juniors then the next generation of Juniors won’t be raised similar to what happened in the do com bust uh boom/bust we’ll
41:10
get another boom which will lead to an incredible perks and fatch paychecks for people with the right skills uh because
41:16
not enough Juniors entered the workforce I actually do agree so like it’s so funny how this article I find myself
41:21
loving this article and hating this article it’s such a wellone article by the way I hope you know I don’t I don’t think this is like trash anything I love
41:28
this guy’s takes and his ideas cuz it it is true this is going to cause a bunch of Junior Engineers not to participate
41:35
and this lack of participation is going to cause like a whole it’s going to cause a whole
41:41
craziness that’s going to happen what would you let’s see what would I do if I were in your shoes understand that time is on your side if you take action if
41:47
you are about a year maybe less what do you do do not squander it your Edge right now is uh a large majority of
41:53
software Engineers have not discovered what you have when they do it’s it’s game over unless you have an edge learn
41:59
the fundamentals that universities typic typically does not teach
42:05
okay okay what is it what are the fundamentals create an application uh that’s not really uh fundamentals learn
42:10
how to uh do property based testing oh now there we go okay A little s little get a little get a little fuzzy and simy
42:17
on there okay okay get that Simi in all right set up a CI pipeline this isn’t
42:24
this is not Fundamentals by the way these are not fundamentals these are specifics right just just throwing it
42:30
out there uh learn uh source code management learn how to release software
42:35
and increments using uh source code CI and property based testing okay find yourself a peer that uh pushes you to
42:42
the limits of learning the standard pace of learning is for Chumps okay I I mean I like that I like this ending if you
42:47
are better than the average then you can fold SpaceTime and outpace the competition read this blog post um I
42:54
like that do not join a startup I don’t know if I agree with that in the last VC pumer bust uh that happened before AI
43:01
many startups raised too much Capital engaged in outright fraud to pump their customer numbers up by cross- selling with each other the AI bust that’s about
43:08
to happen is going to be brutal for them there are many many Living Dead zombie companies out there only join an
43:14
existing startup if it pays well ignore the equity dreams of a jackpot and if you find an operating environment that
43:19
pushes your growth to learning uh and learning to uh the limits so yes join a startup that’s I mean that’s why you
43:24
join a startup is that it it’s fast paid you get to wear multiple hats I mean that’s really the best reason to join a
43:30
startup my first 2 and a half years were startups and it’s just really really good plus I did two years of my own startup before that understand that most
43:37
startups fail and if you could be out on your ass overnight ensure that you have plenty of cash stowed away it could take plus six months uh to find a new role
43:44
mhm that’s what I’m doing yeah I I’ve done it yeah I think startups can be very very great right I think startups can be very very great and yes it could
43:50
take a while don’t get a job that has uh has a ban on AI coding tools uh you’ll
43:55
be doing yourself a huge disservice for companies that encourage it um I would argue that if you’re really needing a
44:02
job you can drop this requirement and I’m just going to throw it out
44:07
there you could probably just drop this requirement companies ban AI all the time like you just got to remember that
44:14
there’s a lot of companies that uh there’s like this whole idea that they can’t share their code and they have
44:20
agreements with like um with third parties that they can’t share code and
44:25
you’re taking your code and sending it to a bunch of AIS like yeah it’s totally reasonable governments medical Banks I’m
44:34
sure a whole slew of different security industries um I’m sure a bunch of whole things involving taxes and
44:40
finances right like there’s like a whole bunch of this stuff that just exists that you can’t Aerospace I’m sure energy
44:46
right there’s there’s probably a whole bunch of things that exist that you just can’t do right people for like again
44:52
this is the fundamental problem about Twitter and YouTube I hope to dispel this
44:57
web development in the Silicon Valley is not the end all of software it is a
45:03
slice of software I don’t even think it’s half of software if I’m not mistaken I bet you it’s somewhere
45:09
between 15 to 35% of all software there’s a lot of other software out
45:15
there right a lot of other software right um by the way I have absolutely no
45:22
idea how you got to what we are talking about to this uh she try to like try to
45:29
stay on topic if you can okay like if you can’t can you just pretend you’re as dumb as an llm and at least say things
45:37
that are predictably the next thing that should come out like could you at least be that bad I mean it takes a pure human
45:42
Ingenuity to get to this point I got one guide absolutely I got I got I got I got
45:48
one guide hard um become a subject mattered expert in M dude this is this has been out for
45:55
like what 3 days and everybody’s talking about it it looks like it came out much earlier it’s only now hitting the
46:01
Twitter sphere in the last Just moment and now everything is just like mcps mcps yeah everybody mcps expert how dude
46:08
it’s been out for like dude if you’re not an expert in something that’s been out for less than six months you’re an idiot [ __ ] and stupid and have no
46:15
foresight and also you might need to go talk to your mom about your life choices okay that’s just that good luck and
46:22
therapy loser um pull apart uh pull apart GitHub block goose and the All
46:27
Hands open a open hands uh learn how they work and then build your own AI coding assistant from the ground up from
46:33
first principles and follow along with my series I’m not going to lie to you when you talk about how all these
46:39
grifters came on and took advantage of all these people getting into an industry and then you give people exact
46:44
ideas of how to get started while also doing something that’s feels similar it can come across as if you are
46:51
effectively doing your own like kind of disservice you want to be employed uh so you want to be the employee that
46:57
companies are fighting for because you have 6 months plus on everyone on edge else and are brand
47:04
new you want to be the employee that companies are fighting for because you have six plus months Edge on everyone
47:11
else and are bringing new ideas techniques on the table trust me companies don’t want this I’m just going to be real here companies aren’t looking
47:18
for the absolute bleeding edge okay most companies still use react class components okay like this is just not a
47:25
thing right this is just not a thing the companies don’t want the absolute newest thing they want something that’s like a
47:32
little bit more tried and truee and older I know this sounds uncool but if you like learn Java and spring boot you
47:40
might be able to find a job I know it’s crazy it’s crazy got a little bit of CP
47:46
in there guess what become one of the one that understands what all these tools do under the hood better than an
47:51
engineer who’s uh 20 plus years into their career and hasn’t been paying attention honestly I probably know better what these tools are doing
47:57
underneath the Hood than this guy that non-stop studied them that’s only like two years into his career like that’s
48:02
just the reality uh focus on shipping with rust let’s go ship with rust baby every
48:10
programming language community and ecosystem has its window and time which attracts Brilliant Minds right now that’s the rust ecosystem it’s full of
48:16
innovators you want to be hanging out with the innovators innovators sometimes don’t have jobs like that’s I’m just
48:22
saying like if we’re if we’re going to if we’re going to go statistically here like this was be rust jobs okay this is
48:29
like Java jobs it’s like the nipple okay it’s just the
48:34
nipple so I mean I don’t like I don’t know where this this ending is kind of confusing for me on this article because
48:40
it it was actually a pretty good article that I made some arguments that I had to kind of think about things I really agreed with things I didn’t agree with
48:46
but this last part just seems like kind of weird vaporware advice like rust could be really good and perhaps in the
48:52
next 10 years it could become a more valuable language for you to focus on because perhaps rust might be the next
48:58
Java right I can’t say whether things are or are not right 10 year 10 years
49:03
will kind of come to that conclusion of whether something is or is not but um
49:08
that doesn’t mean you should be focusing your time on the thing that might be you
49:13
might want to be focusing more on the things that are going to be solid things in which companies actually look for you
49:19
know what I mean that’s a lot of time of course it’s a lot of time maybe zigg might be the most popular thing in the next 10 years I don’t know right I don’t
49:27
no uh if you’re focused on learning something uh uh something uh commodity such as typescript you are positioning
49:33
yourself incorrectly having uh said all that programming languages no longer matter but rust has a unique property
49:38
where it’s able to achieve better outcomes with llms due to the type system than other programming languages except H hasal lack I don’t know if
49:46
that’s true like honestly like to be completely fair if I feel like for consistency of
49:54
this article it would have been better if he would have Rec recomended typescript because typescript is also one of the largest languages out there
50:00
like python typescript JavaScript and the reason being is that llms tend to be the best at those three languages due to
50:06
the amount of available training data right so I I actually kind of find this to be a bit confusing you know kind of
50:12
like a fun trick to do by the way if you’re programming in typescript I know there’s this whole war on return values
50:18
and return types and how you should never do return types and typescript if you do provide return types in the
50:23
signature it actually helps improve the autocomplete of llms because it kind of knows what it’s doing so I would assume
50:29
that if you really wanted to take advantage of uh llms you would actually uh you would actually want to use things
50:35
like Python and typescript honestly just being real here uh understand how your
50:41
work directly translates to business value the organizational abstraction layers are shrinking going forward it’s incredibly important to think and act
50:47
like an entrepreneur if you aren’t a software engineer who types code into an IDE those days are gone bold statement
50:53
let’s see if it’s going to pay off for them bold strategy C and let’s see if it’s going to pay off for him instead you need to understand how your work
50:59
directly translates to business value and Achieve that you need to become a product engineer find yourself a product
51:05
manager or founder who has immense domain depth and befriend them learn what high value problems are to be
51:10
solved and discover their pain points research items by walking in their boots and in the boots of customers be a
51:16
doctor not a waiter learn a little bit about business and finance I mean I I do
51:21
agree with that ending part you should business and finance is pretty good right
51:27
Real Talk actually pretty good you should you should learn about B business and finance let’s see how that old Netflix stock is doing by the way
51:34
damn dude oh my gosh like just like three weeks ago it was like right here dude
51:41
it’s down 200 points I’m not dude not going to make it man 200 points I know
51:46
I’m going back to work everybody bad answer Prime dropping off dude it’s it’s a serious drop off I’m
51:53
not going to lie to you it was pretty awesome but now feel was less awesome all of a sudden uh build a widget not an
51:59
aeroplane when building starts small as soon as you make the tiniest things uh share it with the world take an approach
52:06
of building lots of small things that comp pose well and can combine to make something big it’s a marathon not a Sprint did he just mention posic uh also
52:13
uh to be completely Fair he also just said that SAS is completely irrelevant and it’s completely clonable at this point so in some sense releasing the
52:20
small thing without a moat can also make you go downhill with that type of approach build a public profile I can’t
52:25
stress this enough so many opportunities in uh my life which yielded fruits only happened because I manifested luck uh
52:31
once you have a couple mcps under your okay now I don’t know what mcps are what are we talking about now did did are the
52:39
are we are we still talking about uh the AI uh API language model context
52:45
protocol or did we just change minimal copyable product no I think this is still model context protocol right okay
52:52
under your belt you’re on the way to building your own AI coding assistant start networking don’t Network to get a job Network to share your knowledge with
52:58
others if you’re looking for job also networking to get a job is like a good it’s good it’s a good approach too uh in
53:03
applications this means create your own personal professional websites uh name.com and regularly post it with your
53:09
learning start small build this okay that I mean honestly not a bad not a bad piece of advice uh real talk it’s
53:16
actually not a bad piece of advice create a GitHub account and start publishing all your work there I honestly think this does not matter for
53:22
the most part I GitHub Global uh global rankings
53:28
followers um here here we go we’re going to bring this up I bring this up every now and then just because of how just
53:33
just doesn’t even matter right here I am I’m 19th in the world 19th on all of GitHub does it matter no it doesn’t
53:41
matter right it like literally doesn’t matter right I’m in fact not the 19th
53:48
richest Dev I am not the 19th hottest Dev right like it just it it honestly I
53:56
creating a GitHub to share your work uh or as a means to kind of show that you know what you’re doing is pretty good uh
54:02
publish your own mCP tools on npm or creates IO so typescript or typescript or rust there’s only
54:08
two uh interesting he’s really into mCP so mCP that is mCP the new Pathway to
54:15
AGI uh attend meetups this is by the way this is great advice they’re always looking for speakers public speaking is
54:21
hard no matter how good you are uh how good uh at it or at uh you get it oh my
54:27
gosh no matter how good at you get it always remains hard I know what he’s trying to baby I know what you’re trying
54:33
to say baby baby girl sh I know what you’re trying to say you try to make you try to make me look dumb but you know
54:40
what I already know what you trying to say baby uh push uh past it and start sharing your knowledge identify peers
54:46
and build relations with them share your knowledge without asking for anything in return all right the end actually I
54:52
really like this last one by the way this last one is really really good right here attend meetups is actually
54:57
really really hard to do public speaking public speaking is actually really really hard and you should most
55:02
certainly like really try to do it cuz it pushes you in a really hard difficult way you’re going to be short of breath
55:09
you’re going to have a strong adrenal response you’re going to be under pressure like I’m bad at public speaking
55:14
I’m like literally bad at public speaking I’m perfectly fine being bad at public speaking I’ve never even claimed to be good at public speaking it is
55:21
extremely hard to be good at public speaking no you’re good no I’m not I’m not good this is not public speaking by
55:27
the way okay first off you guys don’t even real you you guys you guys don’t even real you you guys don’t even real
55:34
you guys aren’t real right you’re not in front of me it’s so much different being on a stage with a microphone than it is
55:42
on the internet with a bunch of text okay I’m just saying you guys don’t reel stop trying to reel with this chat
55:49
you’re probably all a bunch of those like lady profiles with the pictures of the vaginas with underwear on and then
55:55
you’re always like wow that’s a really great piece of advice you got maybe we could slide into DMS and chat more about
56:02
your ideas right like I know what’s going on here okay I know how this is
56:07
working right I know what’s going on I know what you guys are after you’re after my credit card okay all of you I I
56:13
already know it’s the greatest it’s the long it’s the longest con of Lifetime okay you guys are always after me Goods
56:19
uh anyways I I really like that um I think this this article really didn’t take the turn I honestly this article
56:26
did not take the turn I thought it was going to take which is that um honestly I thought it was going to be the exact
56:31
opposite yes AI is here you’re going to be screwed andless if you take action don’t don’t sell your early years fro
56:37
for productivity instead of learning like honestly that’s what I thought the the takeaway was going to be that if you
56:43
sell your early years to look like you’re productive your mid years are going to be filled with sadness because
56:49
you never got anywhere past whatever the llms can do cuz here’s what I’m seeing this is my personal thing that I think
56:55
I’m seeing with a lot of people is right now this is kind of how llms went there is pretty much no knowledge of llms
57:02
jippy 2 was probably around but some people knew about it but most people didn’t know about it I think for a lot
57:07
of people in the programming Community co-pilot came out and so we went like this we had like
57:14
LSPs and you pretty much had to do everything else and then out of nowhere co- pilot came out and it felt like this
57:22
crazy leap in time because it really was it was like this crazy thing right
57:27
Transformers have been out for a long time all you need is attention has been out for a long time and so when this
57:32
came out then they’re just kind of felt like there’s this like rapid Improvement for a moment uh Claude 35 copilot suck
57:38
stop it trust me when you say this you sound brain dead and what I mean by that is that in this talk at this moment of
57:44
time this point right here in 2022 co-pilot did not suck it was completely something different and it felt really
57:51
wild and and and it just felt unique and new and I was very very exciting right
57:57
and then all these changes came out right like Cloud 35 came out people were like wow this is like way better this is
58:03
great and then now we’re kind of into this world where I don’t really I haven’t really seen anything that’s been any better honestly Cloud 37 it’s like
58:10
yeah it’s better it’s mostly just been integration tools and so your skill level as like somebody who’s
58:16
new can only be as good as whatever the model is right so if you offload your
58:23
technical expertise to a model when you’re good or when you’re learning this is how good you are you won’t be able to
58:31
exceed this unless if you go beyond that so my recommendation yes students AI is
58:37
out and it is here and it’s likely going to be here forever I don’t see it changing anytime soon I think we’re just
58:43
going to constantly be inundated with this you’re screwed unless you take action do not let your skill sealing be
58:50
dictated by a jypy okay let your skilled ceiling be dedicated or dictated by your
58:56
desire to learn and get after it it’s okay to be uncomfortable it’s okay not
59:01
not to know what to do I’m not going to lie to you there was a moment uh just a couple days ago that I was parsing out some asy uh Vim with me right uh Vim
59:10
with me and I found this one part to be actually really really hard to go through and build this tool that allowed
59:15
me to uh watch these games that I made AI play so I made AI play uh Tower
59:20
Defense and doing that I could make it do something so I could play and it would play up until had these two towers
59:26
placed by both Chad jiid 4.5 and Claude I can tell you right now that claude’s on top and jippidy’s on the bottom I can
59:33
tell by how they play they each have their unique style in which they play this Tower Defense that I created and to
59:38
create this and to be able to display stuff I had to parse out all these antsy codes and I had this one problem with the piece of code that actually stumped
59:45
me for some time and I’m not going to lie to you I wanted the reach for an AI and just have it solve it for me but
59:50
even me at my level I decided not to do that and I actually learned quite a bit even for myself and some dangers about
59:57
goang and how easy it is to make a couple mistakes with go that absolutely screw you so even me to this day I will
1:00:04
often avoid something and take the extra 30 minutes to become better to become more knowledgeable because I want to
1:00:11
earn the experience I don’t want to get an answer it’s as simple as that for
1:00:17
those that are wondering what the answer or what the problem was it was a simple I was slicing an array I should not have been slicing or slicing a slice uh to
1:00:25
grab something out and it was my bad I totally sliced a sliced and caused a very simple problem to occur that could
1:00:31
have been completely avoided but it’s just because I always dude I just made that dumb simple mistake right it’s just
1:00:36
something that I should have realized and uh you got to be careful when you need the slice right just Classic Slice
1:00:43
time oopsie right many such cases yeah and it’s like I’ve done this to myself several times and then I was just like
1:00:48
dang it and I just yeah I reffed myself right no slices aren’t just views views I mean typically views mean you can’t
1:00:54
edit the stuff inside of it they’re both they’re mutable views right and it just makes it really uh it was an oopsy
1:00:59
Poopsy like as simple as that and I was getting owned by some other call to memet and so I just memset myself and
1:01:06
oopsi myself and that’s that and so it’s just one of those times where it’s just like I could have probably got the
1:01:11
answer copy paste in some code and things would have changed but I would have lost that one little bit of struggle and just this improved sense of
1:01:17
debugging and so instead I chose improved sense of debugging and I’m happy about it because that pain is
1:01:24
going to stick with me a lot longer and it’s going to probably influence a lot of things I do over the course of the next 5 years than Simply Having the
1:01:31
problem solved for me all right there you go the name is I still talk about uh AI because AI won’t shut the hell up
1:01:38
gosh dang it AI shut up a Jen
Reflection
Curiosities
Ideas
Misc. Notes
- Looking into cursor now; only 20 a month; def worth it for this project lmao
- https://ghuntley.com/screwed/
- Glad I found this and am using the prescribed route already lmao
- Stated one must build a widget not an airplane; I’d have to disagree to an extent; the model is building widgets aimed at combining into an airplane!!!
- Only then can one show how they aim to fit into a larger picture; by painting one themselves.
- https://ghuntley.com/dothings/
- https://ghuntley.com/mcp/
- Going to look into this; may help for setting up my thing
- https://ghuntley.com/subagents/
- Interesting take, again plan to integrate
- https://ghuntley.com/stdlib/
- https://ghuntley.com/specs/