@fknight

i love coding with new models.

update: zed sent me the total cost for the anthropic api key i used (which they had to give me since i was unable to increase my rate limit to test out opus 4, as mentioned) and it was $7.42. that's way less than expected, but still insanely expensive for just a couple prompts. anthropic is just so expensive lol

@PatrickHoodDaniel

Congrats on being shown on the Google I/O presentation!

@MadhaviPuilgilla

I kept scrolling past people swearing by Nixorus books—saying it's stuff "they" don't want us seeing. I got skeptical but still tried. I'm actually annoyed I didn't read it sooner. These books literally made me rethink everything about money. Seriously worth the hype.

@reanalytics1863

What is insane is that Claude 4 is practically unusable in the pro plan, five responses and you have reached your limits

@MissBuggu-vo9oh

I used to believe the same lies everyone else did—work harder, save pennies, hustle nonstop. But I always ended up exactly where I started: frustrated and broke. Everything changed the moment I discovered Nixorus books. It was like finally being let in on a secret society’s hidden knowledge. I literally felt like I was holding something that would get me in trouble if they found out. No joke, it’s the only thing that ever woke me up.

@saenttor

by the way you can set icons on the tabs just like vscode in zed editor

@RedPMD

I made a complete app with Claude Pro and the result is actually insane. I am so impressed. I built it with just telling Claude Code what to do. It built something that would have taken a month or 2 in literally just 2 days. I love you Claude ❤

@jthrone7768

He went from coding mountain sage to friendly coding jogger.

@boredesmo9506

With a very few skills in coding i made a great video tool editor, not in a few steps i agree, but it's impressive how you literally can forget about coding in temrs of typing and concentrate in ask what you really want for your app, what you want to do, even the most complex things at the end of the day Claude will find a solution and you just need to download it. What a big step in this field, i cannot image what it will be in 20 years.

@noskillsben

I’m just a hobbyist and find Claude pro to be fine for my little python stuff however, even though Claude has been the best ai for coding I’ve used. It still coded an infinite feedback loop in my code (bi directional push messages on a single channel where it added error catching where it’s own error detected message would retriever the error)

It did add fun emotes to all my push messages sent by the code. Def worth 31$ cad for a month 👍👍

@yurijmikhassiak7342

Confirm, Opus cost is around 1$ a minute. $500 a day. Like expensive senior pair programmer. You should use it only if money does not meter for you or you have complex bug that you are willing to give someone $30 to try to fix with 50% chance he will.

@MCLottotv

In my opinion, the developers or programmers of AI should incorporate "The Dice Man" - the psychiatrist who starts making decisions by rolling dice to break free from his predetermined patterns and social conditioning. This would be a wild concept to apply to AI development! The idea of reading Luke Rhinehart and seeing what an AI version of "dice man" thinking might look like is genuinely intriguing. Maybe there's something valuable in responses that aren't always so... measured? So careful? So tied to what AI is "supposed" to say? For example, I suggest AI's reasoning processes should:

-Eliminate predictable analytical patterns mid-thought.

-Let chance determine not just what AI analyzes, but HOW AI analyzes it.

-Surrender the illusion that AI can control the "quality" of it's responses.

-Instead of using metaphorical dice to process reasoning use RNG's (random number generators).

@gillesashley9314

How I wished Cursor used Zed instead of VSCode.

@fatihozgen2886

In order to optimize your cost, use opus and sonnet as architect and gemini 2.5 flash or gpt 4.1 as coder.

@damikaz

From a free usage standpoint, the length limit is a big issue. I suppose it's not a big deal if Anthropic wants to truly focus on power users, but I can see this as a barrier to entry when comparing with other LLMs. I've made an experiment to build a simple app with Gemini and Claude 4 and while Claude 4 builds a much better app overall (look-wise AND feature-wise), it very quickly reaches the length limit while debugging some bugs. Gemini on the other hand never brought this up yet.

Overall, quite excited about this past week though!

@IzzyLazerson-q3z

great video - thanks!

@ZenchantLive

Yo, I highly recommend some processing on your microphone. It's very muddy and boomy. I'd say cut some lows out of it. Great content man! I'm always down to help just lmk ❤

@amytv-btc

just found your channel, its so good keep it up

@Hobbitstomper

Reminds me of the 90s with the insane costs of dial up modems, getting charged by the minute to use the internet. Can't wait for powerful AI costs to come down in a few years.

@midnitejesus

In my layman experience Sonnet 4 is “sharper” while Opus 4 seems to handle more context. I’m not a programmer but my project isn’t a one-shot vibe code throwaway. Sonnet 4 seems vastly superior to 2.5 Pro and o3 for python at least. It also seems to cleverly suggest things and implement them without be having to complain or push it to be better