Amitav Panda

This is how we approach to build products faster with help of vibe coding. 1. Whenever we get a requirement - We jump to @perplexity_ai and use our engineering skills to architecture the product , DB design, etc. 2. For UI - Either we directly use @lovable or use @uxpilotai to build the figma design and then jump to lovable for building the UI with authentication and basic features. Point to Note : Once we feel we should stop using Lovable from onwards , we stop and move the project to local for further development. 3. Then we start building further using our engineering skills and relying on github copilot and claude-code for further continue building the product. 4. For documentation and all, we use @perplexity_ai . #buildinpublic #fullstack #frontend #backend #javascript #freelance #remote #business #AI #vibecoding

Around a month ago - I did my AI setup according to my needs - more than 80 skills , 15 agents , 5 hooks and guidelines specific to each project - and I can say , it has literally improved the whole workflow and it’s amazing.

@ameliahazelai It’s not openClaw

🚨BREAKING: Google just dropped Gemma 4 (OpenClaw for FREE) It runs on your phone. Fully offline. No internet needed. No subscription. No data leaving your device. It's called Google AI Edge Gallery. Available on both iPhone and Android right now. What it does: → AI chat running 100% on your phone → Works in airplane mode → Text, image, and audio support → Agent skills that complete tasks on their own → 100% open source. Code on GitHub. → Works with Gemma E2B (2.5 GB) and E4B (3.6 GB) How to set it up: → Download Google AI Edge Gallery from App Store or Play Store → Open it and pick AI Chat → Choose Gemma E2B if your phone has less RAM or E4B if you have 8GB+ → Start chatting. That's it. No account needed. No sign up. No API key. Just download and go. People are paying $20 a month for ChatGPT. $200 for Claude Max. $300 for SuperGrok Heavy. Google just put a capable AI on your phone for free that works without wifi. While every other company is figuring out how to charge you more, Google keeps giving stuff away. Go download it.

@sickdotdev Fine , then what about marketing , distribution and sales

“Yeah, I’m a founder” My team: Claude: $20/mo Supabase: $0 Vercel: $0 Stripe: 2.9% + 30¢ GitHub: $0 Antigravity: $0 Dynadot (Domain): $12/yr Total Cost: <$20/mo You don’t need VC money. Just Deploy it.

Made it to Round 2 of VibeCon India by @emergentlabs 🚀 For the past 2–3 years, I’ve seen how wholesalers & small businesses run on WhatsApp, spreadsheets, and guesswork. So I built something different: → Your own branded e-commerce app + website → Built-in accounting → AI that tells you what to do All for ₹500/month. This isn’t just software — it’s an AI operating system for SMEs. Please watch the whole video to get the complete concept, solution, and demo 👇 #VibeCon #BuildInPublic #AI #SaaS #StartupIndia @emergentlabs #BuildInPublic @ycombinator

Anyone here using Gemma 4 with OpenClaw? Curious to hear real-world experiences — especially around performance, setup, and how it fits into agent workflows. Planning to explore this combination soon.

I think Gemma 4 might quietly change how we build AI products. Not in a hype way — but in a practical way. Google released it recently, and what stood out to me is this: It’s not just about better responses. It’s about making AI agents actually viable at lower cost. --- From what I’ve seen: - It can run locally (or cheaper infra) - It supports agent-style workflows - Strong enough for reasoning + coding tasks - Open enough to actually build products on top --- This got me thinking: We’ve been building AI apps wrong. Most stuff right now is: - chat UIs - wrappers - basic automations But models like this make it possible to build systems that: - scrape data - analyze it - take actions - iterate --- For example: You could build an agent that: - finds founders hiring devs - scrapes relevant signals - qualifies them - sends outreach automatically That’s basically a full lead-gen pipeline. --- Curious what others think: Are we finally moving toward actual autonomous systems, or is this still overhyped? Also — is anyone here already building agent-based workflows with open models?

I had a client call yesterday. Requirement: A landing page for an RO service business with some extra features WhatsApp integration for lead capture The call lasted about an hour. Within ~20 minutes after the call, we had a working prototype ready in the client's DM. What changed? The workflow. • Fireflies.ai summarized the meeting • That context was fed into Lovable • A basic prototype was generated instantly No back-and-forth documentation. No starting from a blank screen. This is what AI actually helps with: Compressing the gap between requirement → first version

Shipped a new update today. We added an Active Sales view to make sales tracking easier for tenants. Now tenant admins and staff can: • View only active sales • Use filters to quickly find transactions • Track payments and credit balances more clearly Users can also activate or deactivate a sale directly from the Sales Overview. This was implemented based on a real business need — teams wanted a simpler way to track which sales were still active and which ones were closed. Small improvements like this make a big difference in day-to-day operations. When building products, shipping features that improve clarity and workflow often matters more than adding complexity. #buildinpublic

@striver_79 It depends what someone means by “vibe coding.” In my freelance work and full-time team, we use what many people would call a vibe coding setup every day (Cursor, VS Code, LLM workflows). But for us it means: • we know the fundamentals • we review the plan and generated code • we modify when needed • we ship faster LLMs accelerate execution — they don’t replace engineering judgment. For one of my SaaS products I’m scaling, the same setup works well because the fundamentals are already clear. If “vibe coding” means blindly asking AI to build everything, then yes — that won’t produce enterprise-grade systems. But with solid software engineering knowledge + the right AI workflow, you can absolutely build sustainable products.

We’re moving from: AI coding assistant → AI development environment For teams building MVPs fast, this direction makes a lot of sense. AI shouldn’t just generate code. It should help structure the way we build software.

But the most interesting part is learning. The system extracts patterns from sessions and turns them into reusable skills. So the environment gets better over time.

There’s a repo called everything-claude-code that’s quietly becoming one of the most interesting Claude Code setups. github.com/affaan-m/every… Almost 50k GitHub stars already. But the interesting part isn’t the popularity. It’s how it turns Claude Code into a structured multi-agent development system.

For teams building MVPs fast, this is powerful. Instead of constantly prompting AI, you can build repeatable execution workflows. Less manual work. More structured automation. The real advantage won’t be prompts. It will be AI-powered workflows.

This is why it matters: We’re moving from Prompt → Response to Event → Agent workflow → Execution AI is starting to act more like part of the development pipeline.

Cursor just introduced Automations, and it’s a bigger shift than it looks. AI coding tools are moving from assistants you prompt → agents that run workflows automatically. Here’s what that means in practice 👇