Hello everyone. This is Professor

Patterns and today's a great day because

we also have the Opus 4.1 model that's

also been released. So, this I've been

waiting for for a really long time. Um,

and it's an upgrade from the current

Claude Opus 4 model. And it's great for

things like agentic task, real world

coding, as well as reasoning. And we see

here how much of an improvement it is.

So, this is set 37, that's 62.3, and

then opus 4 is 72.5. and then 41 is

74.5.

So they say it's about a one standard

deviation improvement over Opus 4 on

their junior developer benchmark. And

the claim is that it's about the same

performance leap as you would have when

you go from Sonnet 37 to Sonnet 4. I

don't really know how I feel about that

because to be honest, I kind of still

like the 37 model more than the Sonnet 4

model. And I'm only saying this from

experience just because I've been using

not only the sonnet model heavily

through open router. I mean honestly if

you take a look at my open router

credits page uh you you'll see here that

I just over the last two or three months

I've probably spent about 2500 to about

$3,000 on just open router credits. And

that doesn't even include my Cloud Max

subscription as well as the API that

I've been using from uh Claude directly

and also the API that I've been using

from Amazon Bedrock. So probably spent

closer to about $15,000 over the last 3

months on you know all of these models.

So I think I would be pretty decent at

being able to figure out that if there

whether or not there is a difference or

not uh between these models. So, um, I'm

excited about Opus 41. I think Opus 4 by

itself is great. If I have a bug, uh,

anywhere in my codebase, it's able to

find this bug and fix this problem like

really, really easily. And I just need

to review it and just give it a go and

say like, yep, this sounds great. Okay,

so enough talking. Let's call this model

through open router. And I could have

used the cloud provider as well, but

that's fine. For now, I'm just going to

give it the cloud opus 41 model. Just

make sure that that looks right. I will

also give it the or enable extended

thinking and allow the full budget of

6,553 tokens. And here I'm actually

giving it a code base that actually has

a bunch of different things. So, it's

not just like an empty codebase. Um,

this is a code base for my website for

Percr. And as you can see, like this is

a full sort of landing page, but doesn't

really have anything else in there. Um,

this particular folder is just for the

landing page development. And what I'm

going to do is simply take a screenshot

of this and go here. What's wrong with

my hero page? And paste the link. Um,

the background is really off. Can you

help improve it further?

Uh, think step by step and put yourself

in the mind of an expert UI UX designer

and researcher.

List out all the things to improve and

do it.

This is kind of how I usually

work with this sort of prompting. I ask

the models to put themselves in the mind

or themselves itself in the mind of what

I want it to be. And there we go. It's

starting to work. This is my first ever

time working. Oh my god. That's that

that was 20 cents. That was 50 cents and

it hasn't really done anything yet.

Okay, so it looks like it is finally

done now. And this cost me about $2.50.

Um, and this is the current state of the

site and the landing page. And this is

the one that I just created.

I absolutely hate it. This is complete

waste of $2.50.

But I can see an appeal that this might

have for some people. I just don't think

that this is the right type of thing for

the product.

But I guess the UI is kind of cool. You

hover your mouse there and all the stars

kind of move and stuff. We have this

little hover effect kind of thing.

But would this really be premium? I

don't know. It's like stars and space

kind of vibe, I guess. But okay. Uh,

that's what you get for vibe coding. Let

me try it one more time. I gave it the

same exact prompt and this time it built

this

which to be honest does look a little

bit at least it looks different. At

least it doesn't look

as whatever it was the last time. I

think it's a little bit more

enterprisegrade. It did add in a bunch

of stuff that's not true. So that needs

to get worked on a little bit. But

honestly this time around it's not as

bad as it was. Would I still change it?

I don't know. Um, maybe I'll think about

it. Let's also see how this works on

LinkedIn. Create a post for the

LinkedIn. So, I basically ha had it

create a full master document that

essentially lists all of the things that

the site actually does. And I want to

see what the writing style looks like.

Um, recruit. So, it does have m dashes.

So, a little bit of a giveaway that this

is an AI generated post. 212 word count.

And what's the raw content? We flip this

equation. Here's what changed.

Yeah, I don't know. I think that if I

was to create a post like this, it would

be pretty obvious that I feel that it

was AI generated.

So, I don't know. Is it still there in

terms of like the writing part of

things? I probably I I don't really

think so. Um, but maybe I just need to

play around with the prompt a little

bit. Maybe give it a couple of like one

twoot examples of like how I want it to

be. Maybe create like a full brand

writing style. Um, if you haven't

watched my video on Claude projects, um,

I recommend it. That way you can kind of

like create an entire project, give it a

full knowledge base. And then here I can

maybe add some content on let's say text

content on like brand

writing guidelines and style. And then

this way anytime that I'm referencing u

or anytime I'm actually making a chat

here, it's always going to be able to

reference the writing style and

guideline and all of those kind of

things.

But yeah, so far I I feel like in terms

of coding, it looks great. I'm going to

try it out with a couple of bugs that I

know that I haven't been able to fix

yet. Um, and that way I'll have a full

understanding of like how well this

model actually does. In terms of cost,

yeah, it is an expensive model, so I'm

not going to be using it all the time.

This is just something that I would use

when, you know, I'm really frustrated. I

just can't seem to get to the bottom of

the problem with cloud set. and then I

just, you know, either debug it myself

or I would just use cloud opus. But

that's it for this video. I think that

overall I'm really excited about this

new update. Uh the model does seem

really great at figuring out a lot of

different things. I'm still needing to

experiment with it a little bit more. Uh

let me give it a couple of really

important stuff like things like in a

huge code code base that has a lot of

different interactions and stuff and let

me see what I get back from there. But

more to come. Thank you all for tuning

in. If you haven't watched my video on

GPT OSS model that was released

literally an hour after this model, uh,

watch that one as well. I posted a link

to it in the description below. Thank

you all for tuning in. I'll see you in

the next one. Goodbye.