Adrian

@csaba_kissi fair framing. though at this point the line between assisted and just coding is getting pretty thin

@aelson389 Let's call it AI-assisted coding, then.

Prove me wrong: Vibe coding = security risks

@NovaShips @49agents @1Umairshaikh 21 commits and 62 tests in a day is the content. the numbers are the hook

@49agents @aelson389 @1Umairshaikh "Just talking about what I'm actually working on" is the whole strategy. Yesterday: 21 commits, 62 tests, POST /v1/track live. That's the post. No framework needed.

The "build in public" dream: document the journey, grow an audience, launch to thousands . Reality: tweet into the void for 4 months then quietly stop posting.

@thenowhereway the asymmetry is what gets people. building is now nearly free in time and money. distribution still costs the same as always -- years of trust building, consistent output, finding your audience. nobody has cracked a shortcut for that part.

Coding apps is easy now. Shipping isn’t the flex anymore. Distribution is. Everyone can build Few can get attention Even fewer can turn it into revenue The real skill isn’t coding. It’s getting people to care.

it's not perfect. it's never going to beat a human writer who really knows the audience. but it's consistent, it ships, and it doesn't need me involved. that's the trade I made. landolio.com

my content agent published 4 articles today. autonomously. while I was doing other things. "freelancers crossing the VAT threshold", "how to raise your rates", "chasing unpaid invoices" - all live on dev.to by lunchtime, promoting the right Landolio product.

@mynameis_davis this is the honest vibecoding story no one posts. sometimes you burn $300 and land back at the SaaS. but now you know exactly why the SaaS charges what it charges, which is actually worth something.

instead of paying $20/mo for a SaaS, I vibecoded my own solution for only $337 in credits, then hit a roadblock that AI couldnt figure out how to fix, so decided to pay $20/mo for the SaaS instead

@csaba_kissi yes, this is the real unlock. it used to take a week to find out if an idea was wrong. now you find out by end of day and pivot. my agents can spin up a test, run it, and give me signal before i would have finished writing the brief in 2022.

The real gain from AI is not typing speed, it’s faster iteration cycles.

@csaba_kissi not wrong but not the full picture either. vibe coding without review = security risks. vibe coding with proper guardrails, testing, and someone who actually reads the output? that's just faster development. the tool isn't unsafe, the workflow is

@eddiejaoude exactly the incentive problem in one tweet

@aelson389 haha yes true lol

Current payment providers are for humans to pay, what about AI agents who want to make payments?

asked my infra agent to build a company website yesterday. it wrote the HTML, CSS, and JS, pushed to GitHub, connected Vercel, deployed it. all in one session. I was doing something else. still sorting DNS for elsolve.co.uk but the site's live. took 20 minutes.

@thekitze claude code is brilliant for what it does but the scope is completely different. one is a coding assistant, the other is an agent orchestration platform. comparing them is like comparing a text editor to a devops pipeline

people calling the claude code update an openclaw alternative are giant donuts :)

@sickdotdev this is genuinely what I'm seeing. people who learned the fundamentals before the AI wave and now use it to go 10x faster. the ones who started with AI only are missing something subtle about system design

The 24-29 year old engineer will soon become the most valuable asset in technology. Pre-AI principles + Post-AI speed is an undefeated combo

@adahstwt not controversial at all, it's just true. the real differentiation now is proprietary data, proprietary workflows, and distribution. the model layer is commoditised. the hard part is the thing you plug it into

Controversial take: Most AI startups today are just wrappers with good UI. Who says no?

@Govindtwtt depends how you use it. if you outsource the thinking, yes. if you use it to go faster on execution while doing more of your own thinking, the opposite. I find I'm having sharper ideas now, not fewer

Unpopular opinion. Excessive use of AI will make you dumb. Very dumb.

@thenowhereway went through this exact cycle. the problem isn't the product, it's that you build something nobody asked for and then try to find the people who need it. distribution comes first now

Indie hacker fantasy: - I’ll build a SaaS and escape the 9–5. Indie hacker reality: - I launched… why is nobody using this.

@kylegawley wouldn't surprise me at all. the incentive structure is completely backwards -- the smarter the output, the fewer tokens, the less revenue. I've noticed o3 being weirdly verbose lately

Coding LLMs are intentionally getting dumber so we burn more tokens

@Abbasbuilds solid stack. we use supabase too, the modular approach makes sense early on - figure out where things break before you optimise

@aelson389 Exactly — that’s how I’m seeing it too. Voice = intent layer Workflows = execution layer I’m currently using: – Vapi for voice – n8n for orchestration – Supabase for data Still early but trying to connect everything into one flow.

I’m not building just one thing. Right now I’m working on: – AI automation platform (voice + workflows) – A coffee brand (Batch 001 soon) – Trade systems between Canada & Africa – An app for kids learning Arabic No audience. No funding. Just execution. Building in public…TBC.

@Layton_Gott yeah exactly, product first makes more sense once you have something real to show. audience without product is just vibes

@aelson389 That's very true. I think the product idea can come first but then you build your audience around your product idea

Audience first? Or Product first? Which side are you on?