Mian Shahzad Raza

@dramaricic My first vibe coded project that’s fully functional and live is talos.tools

I follow builders who actually ship. If you're: – building a SaaS – working on a side project – trying to get first users Drop your project below Let’s connect 🤝

I'm documenting my entire journey from WordPress dev → AI-powered full-stack builder. Follow @MSR_Builds if you want to make the leap too. What's the craziest thing you've vibe-coded? Drop it below 👇

My prediction for the next 12 months: → WordPress devs who learn vibe coding will charge 3x more → "I build WordPress sites" becomes "I build AI-powered web apps on WordPress" → The devs who refuse to adapt will lose clients to Wix, Framer, and no-code tools The middle ground is disappearing.

Naval just said vibe coding is more addictive than any video game. He's wrong. It's more addictive than social media, gaming, and Netflix — combined. Here's why every WordPress dev should be terrified (and excited): 🧵👇

The barrier to starting dropped to near zero. AI scaffolds the product, cloud handles infra, and distribution is one post away. But the barrier to sustaining is still just as high. Starting tests your ability to build. Surviving tests your ability to sell, iterate, and keep going when nobody cares yet.

at this point, starting a tech company is getting easier than joining one

The content game is shifting from "rank for keywords" to "be the source AI cites." Traffic as a metric is dying, but authority as a signal is not. The blogs that survive this will be the ones AI models reference because the writing is original and deeply technical. Keep documenting. That compounds in ways traffic never did.

My site is having a rapid death due to AI. Traffic has dropped by 50%, and it’s still going down. End of an era. It’s no more a source of income. Thankfully, I enjoy writing and documenting. Otherwise, I would’ve abandoned it.

Exactly, Ryan. Resilience is what separates a demo from a real agent. Hard failures treat errors as dead ends, but a well-designed agent treats them as just another observation in the loop. Observe the failure, decide how to adapt, act differently. That retry mindset is what makes agents actually useful in production.

@MSR_Builds @gvanrossum That loop framing is exactly right. The part that trips people up is when they build agents that break the loop with a hard failure instead of observe + retry.

I think I finally understand what an agent is. It's a prompt (or several), skills, and tools. Did I get this right?

This is the "speed amplifies direction" problem. When building was slow, the friction itself was a feedback loop that forced you to question assumptions. Vibe coding removes that friction, so now you need to front-load the thinking. The developers who win here are the ones treating product thinking as the new bottleneck, not the code.

The problem with vibe coding isn't the code. It's that you can now build the wrong thing at 10x the speed.

Smart design choice. Re-injecting every turn means project context stays persistent without relying on the model's memory. The real unlock is treating CLAUDE.md as your project's operating manual. Architecture decisions, coding conventions, known edge cases, all in one place. Compounding returns on every session.

Based on the leaked Claude Code source code, your CLAUDE.md file is re-injected on every single **turn** of the conversation.

@trikcode This is the real bottleneck of AI-assisted development right now. You get into a deep flow state with Claude, and then rate limited. The pricing model needs to evolve. Pay-per-token with no hard cutoffs would be a better fit for serious builders.

Apparently $200 isn't it enough People paying $200/month for Claude Max are burning through rate limits in 2 hours. $200/month. 2 hours. This is not a subscription. This is crazy!

@aryanlabde This is painfully relatable. Building is the easy part now. The hard part is still the same: talking to users before writing a single line of code. Vibe coding solves the "how to build" problem, not the "what to build" problem.

Vibe coded my first SaaS in a week. Spent the next 2 weeks figuring out why nobody was paying. Turns out shipping is the easy part.

@TeeDevh 1 app, improved for 12 months, every time. Shipping 12 apps teaches you how to launch. Iterating on 1 app teaches you how to build a business. Distribution, retention, and customer feedback loops are things you only learn by staying with one thing long enough.

Which one gets you to success faster? 1 app - improved for 12 months or 12 apps - built and validated over 12 months 🤔

@csaba_kissi AI won't replace developers, but companies will use AI as justification to cut costs. The real shift isn't replacement, it's leverage. One developer with AI tools now does the work of a team. The ones who adapt and upskill will thrive. That's why learning never stops.

Coders: AI won't replace developers. Oracle: We’re laying off 30K employees.

@mchulet Not even a hot take, it's just facts. Knowing what good code looks like is what lets you steer the AI in the right direction. Otherwise you're just accepting whatever comes out and hoping it works.

Hot take: Vibe coding only works well if you already know how to code.

@sflorimm Talk to users, always. $100K burns fast on hires or marketing without product-market fit. 10 deep conversations will tell you more than 10K in ad spend ever will.

You just launched your SaaS. Then you suddenly receive $100K to grow it. You can only choose one move first: -Hire -Marketing -Build more features -Talk to users What are you doing?

@yashhq_22 Bad distribution, every time. The graveyard of startups is full of brilliant products nobody ever heard of. You can iterate a bad idea into a good one, but you can't iterate your way out of zero users.

Which one kills most startups? - bad idea - bad distribution