Carazy
1.7K posts

Carazy
@carazy
The reply guy your parents warned you about. Really hates all the things you like. My side is the best side.
Katılım Eylül 2007
444 Takip Edilen149 Takipçiler

@SergioRocks This post is obviously 100% AI written. But I'll bite. If you are a dev for an enterprise coding faster will save you. The only place this applies if you are self employed which if you look at twitter seems to be 95% of the people posting but likely 1% of reality.
English

Coding faster won’t save you.
It feels like the obvious move:
- Better tools
- Faster output
- More code shipped
That used to be the game. Now everyone has access to the same leverage.
Cursor. Claude Code. Copilot.
Execution is no longer the advantage. It’s the baseline.
What actually moves the needle now is:
- Picking the right problem
- Scoping it correctly
- Knowing what not to build
You can write twice as much code. And still build the wrong thing.
You can ship faster. And still not create value.
That’s why some engineers feel stuck. They’re optimizing the part that got commoditized.
AI didn’t make engineering irrelevant. It changed what matters.
From how fast you build to what you choose to build and how you ship it.
English

@TheQuietLynx @TheGeorgePu yeah, unfortunately I have seen this. non devs buying simple AI wrappers that us devs could duplicate in a day.
Specifically, yesterday I told we're getting an AI tool to help create user stories from business requirements. Which we've been doing using existing cli tools.
English

@TheQuietLynx @TheGeorgePu I don't know of a single person who has paid for any of these AI wrappers. We all just go with the big names.
English

@TheGeorgePu It’s all about focus and time - most people don’t want to build a workflow from scratch and will happily pay a “wrapper” to handle and maintain it for them
English

@TheGeorgePu And any AI wrapper that does anything valuable will be merged into the big player platforms.
English

@benvargas I've gotten this a few times and pushed back on it and it apologized and said it wouldn't suggest calling it a day in the future. It did write something to memory and I haven't seen it sense but yeah very weird.
English

@ganesh__47 I'm interested in replies because Claude is my current favorite. I expect that Codex is right there with it but haven't given it much of a chance yet. I do want to try some open source ones like opencode so not tied to a specific model provider.
English

@icanvardar Depends on the definition of vibecoding. Is that just using a AI to program? You can definitely do that without knowing anything about programming. But the way developers use coding agents differs drastically depending on your dev experience.
English

@asaio87 you didnt mention the maintenance cost, which adds even more on the plate
English

Saas is dead.
Why pay $49/month for an app when Claude code can build it for you with just a $20 subscription ?
That sounds quite surreal, yet many say its true.
But we have an issue.
Nobody tells you that you need to spend hours or days or even weeks prompting to actually build the thing, and you might hit limits, you might hit hallucinations, and stuff like that.
Busy founders or people doing business want something that works and dont need to maintain or wake up at 2AM to fix the payment system or the contact form.
Meanwhile, AI hypers will poison you that saas is dead.
Only people who didnt try to build, can agree.
English

@NanEmAll @AmericaAeterna @Oilfield_Rando Copilot? Why would you use copilot when you have all the models separately? And grok is way far behind right now, not sure what you would use it for.
English

@AmericaAeterna @Oilfield_Rando Claude is great for structured tasks. For the rest is useless.
And yes, before you ask, I'm using daily Claude, Copilot, ChatGPT, Grok, Gemini and Perplexity.
English

@kalyan_wtf Every once in a while to look at what's changed before AI commits.
English

@pedropankaj @kalyan_wtf This is why jr devs are in trouble. Sr devs can see the code, understand the direction the ai is taking it and guide it in the right direction. You can also then use the AI to review the code. I haven't manually coded anything since November. And doubt I ever will again.
English

@kalyan_wtf the real shift isn't just using AI tho, it's knowing when NOT to use it. seen plenty of AI generated code that created more problems than it solved
English

@kalyan_wtf By 2027? I would say it's more than 10x now. AI enables you to multitask effectively and each of those sessions is way more productive than I was when I was focussing on one task.
English

@matt_artist115 @IranArmyStan No wars. It's 2026 why do we still think killing people is okay? We should do everything possible to not be at war not jump at any opportunity to be in one. Not a hard platform to get behind.
English

@carazy @IranArmyStan So I'm curious on your actual position? because it seems to me like your hatred for Trump is the sole reason for anything for you.
Should we just continue to let the Iranian regime be horrific? What do YOU want.
English

so let me get this straight
all of ai twitter was telling people to buy a mac mini to run openclaw, which is literally just a framework, an orchestration layer that sends api requests to actual ai models. something you can run on a $5/month vps. which is exactly what i do btw
but when google drops gemma 4, an actual large language model that you can run and fine-tune locally on that same mac mini, with no api costs, no subscriptions, no third party dependencies, completely yours under apache 2.0
the ai community is silent
you were buying $800 hardware to run a wrapper but ignoring the actual ai model that would justify that hardware
this tells you everything you need to know about the average iq of ai twitter
English

@Joestar_sann ?? I've heard countless stories about putting gemma 4 on a Mac mini including this post?
People that put openclaw on mac mini was more for being able to be a "blue bubble" with your friends and interacting with other mac specific apps directly. Something you can't do on that VPS
English

@deguerre @krishnapro_ @marcba What undeniable example? I haven't seen it. If it was undeniable why can't we define or test for it?
I will agree 100% on the unaware of their lack of knowledge. They can have huge blind spots because they just don't know things even exist without being told about them.
English

@carazy @krishnapro_ @marcba LLMs are unaware of their own lack of knowledge. That's why they hallucinate/conflate.
I don't have a good definition for consciousness, because it's a prototype category and we only have one undeniable example, but I would say that is one requirement that currently isn't met.
English









