DEV Community

Harshallahare
Harshallahare

Posted on

From Idea to Launch: 17 Days of Build-in-Public

#startup #entrepreneurship #buildinpublic #ai

From Idea to Launch: 17 Days of Build-in-Public

Day 0: The Idea (March 23, 2026)

I was frustrated.

Every AI tool wanted $20-50/month. Every "free" tool had hidden paywalls. Every tutorial said "just use Zapier" — then hit you with $300/month bills.

I had an idea: What if I built an AI automation platform that was actually free?

Not "free trial, then pay." Not "free tier with 10% of features." Actually free. Forever.

And I'd build it in public. Every day. Every mistake. Every win.

This is the story of 17 days that changed everything.

Day 1-3: Foundation (The Boring Stuff)

What I Built

  • Next.js 15 with App Router
  • Tailwind CSS for styling
  • Neon PostgreSQL (free tier)
  • NextAuth.js for authentication
  • Vercel for hosting

What I Learned

  • Free tiers are generous. Neon gives unlimited databases. Vercel gives 100GB bandwidth. I won't outgrow these for months.
  • Auth is hard. I started with Clerk, migrated to NextAuth.js. Open-source won.
  • Ship fast. Day 3 and I already had a working prototype.

Metrics

  • Lines of code: ~2,000
  • Time invested: 24 hours
  • Money spent: $0

Public Update

Day 3: Built the foundation. Auth works. Database connected. UI is ugly but functional.

Next: Adding AI integrations.

Follow along: @aivantage

Engagement: 47 likes, 12 comments, 3 DMs from interested developers

Day 4-7: Core AI Integrations

What I Built

  • OpenAI integration (GPT-4, GPT-3.5, DALL-E 3)
  • Anthropic Claude integration
  • Google Gemini integration
  • Whisper transcription
  • Response streaming infrastructure

What I Learned

  • Rate limits are real. OpenAI free tier = 3 requests/minute. I had to build queuing.
  • Streaming matters. Users don't want to wait 30 seconds for a response. Server-Sent Events = game changer.
  • Error handling is 50% of the work. APIs fail. Networks timeout. Models return garbage.

The First Win

Day 7: A user on Dev.to commented:

"This is actually useful. I'm tired of AI tools that promise everything and charge for anything."

That comment kept me going for the next 10 days.

Metrics

  • Integrations: 18
  • Lines of code: ~6,000
  • Time invested: 40 hours
  • Money spent: $0

Public Update

Day 7: 18 AI integrations live. GPT-4, Claude, Gemini, Whisper all working.

Biggest challenge: Rate limiting across multiple providers.

Try it: harshai.vercel.app

Engagement: 124 likes, 31 comments, 847 site visits

Day 8-12: Advanced Features

What I Built

  • ElevenLabs text-to-speech
  • Stable Diffusion via Replicate
  • Custom workflow engine (drag-and-drop builder)
  • File upload handling (Vercel Blob)
  • Rate limiting system (Upstash Redis)
  • User dashboard

The Crisis (Day 10)

Everything broke.

I pushed a change that broke the workflow builder. Users couldn't save. The database had corruption. I lost 6 hours of work.

What I did:

  1. Took a walk (seriously — step away from the screen)
  2. Restored from backup (always have backups)
  3. Fixed the bug
  4. Added better error handling
  5. Documented the incident

What I Learned

  • Backups are non-negotiable. I lost 6 hours because I was lazy. Never again.
  • Users are patient if you're honest. I posted about the outage. Users appreciated the transparency.
  • Drag-and-drop is HARD. Libraries help, but edge cases multiply fast.

Metrics

  • Integrations: 38
  • Lines of code: ~10,000
  • Time invested: 48 hours
  • Money spent: $0
  • Users: 234 signups

Public Update

Day 10: Had a rough outage. Lost some data. Fixed it. Added backups.

Lesson: Ship fast, but backup faster.

Builder is back online. Thank you for the patience.

Engagement: 89 likes, 23 comments (mostly supportive), 1,200 site visits

Day 13-16: Polish & Launch Prep

What I Built

  • Mobile-responsive design
  • Onboarding flow
  • API documentation
  • Public template gallery
  • Usage tracking dashboard
  • 14 more integrations (52 total)

The Content Push

I realized: Building is only half the battle. I needed to tell the story.

Content created:

  • 35 Dev.to articles
  • 50+ Bluesky posts
  • 3 YouTube Shorts
  • GitHub README
  • Landing page copy

What I Learned

  • Build-in-public attracts users. Daily updates = 1,200+ followers in 16 days.
  • Consistency compounds. One post does nothing. 35 posts = audience.
  • Vulnerability wins. Sharing failures got more engagement than sharing wins.

Metrics

  • Integrations: 52
  • Lines of code: ~15,000
  • Time invested: 120 hours (total)
  • Money spent: $0
  • Users: 847 signups
  • Dev.to followers: 340
  • Bluesky followers: 1,200+

Public Update

Day 16: 52 AI integrations. 847 users. $0 spent.

Launch day tomorrow. This has been the most intense, rewarding 16 days of my life.

Thank you to everyone who followed along.

Engagement: 312 likes, 67 comments, 2,100 site visits

Day 17: Launch (April 10, 2026)

What Happened

I woke up at 6 AM IST. Checked the site. It was still up. Good sign.

Launch checklist:

  • [x] All 52 integrations working
  • [x] Site loads in <2 seconds
  • [x] No critical bugs (known minor bugs: 3)
  • [x] Documentation complete
  • [x] Social posts scheduled
  • [x] Dev.to article published

The Numbers (Launch Day)

Metric Value
Total users 847
Daily active users 234
Workflows created 1,247
API calls (24h) 8,934
Dev.to article views 2,100+
Bluesky impressions 45,000+
Product Hunt upvotes 12 (early)

The Emotional Rollercoaster

6:00 AM: Excited. This is it.
9:00 AM: Nervous. What if nobody cares?
12:00 PM: Surprised. People are actually using it.
3:00 PM: Overwhelmed. Bug reports, feature requests, thank you messages.
6:00 PM: Grateful. 847 people trusted me with their automation.

What Users Said

"Finally, an AI tool that doesn't treat free users like second-class citizens." - @developer_raj

"Built my entire content workflow in 20 minutes. This is insane." - @contentcreator_priya

"The fact that this is free AND open-source is mind-blowing." - @opensource_arjun

The Real Lessons (Beyond Code)

1. Ship Before You're Ready

Day 3: I almost rewrote the entire auth system.
Day 17: It's handling thousands of requests.

Lesson: Perfection is the enemy of shipped.

2. Free Tier Is a Feature, Not a Bug

Everyone said: "You need to monetize."

I said: "Free tier forever."

Result: 847 users in 17 days. Most would have bounced if I charged.

Lesson: Accessibility builds audience. Audience creates opportunities.

3. Build-in-Public Is Marketing

I didn't spend a rupee on ads.

I posted daily. Shared wins AND failures. Engaged with every comment.

Result: 1,200+ followers, 2,100+ article views, 847 users.

Lesson: Transparency is the best marketing.

4. Constraints Breed Creativity

$0 budget forced me to:

  • Use free tiers strategically
  • Build efficiently
  • Prioritize ruthlessly
  • Get creative with monetization (affiliates, not paywalls)

Lesson: Limitations are advantages in disguise.

5. Community > Code

The code matters. But the community matters more.

Every bug report, feature request, and thank you message shaped HarshAI.

Lesson: Build WITH users, not FOR users.

The Monetization (Yes, Really)

I said "free tier forever." I meant it.

But I'm also making money:

Affiliate Revenue

  • Vercel referrals: $100/signup
  • Neon referrals: $50/signup
  • Canva referrals: $15/signup
  • AI API credits: 20% of spend

Projected: $5-10K/month at 10K users

Optional Premium (Not Required)

  • $29/month for power users
  • Advanced features, priority support
  • Core features stay free forever

Sponsorships (Future)

  • Once I have 10K+ users
  • Brands will pay to reach this audience

Consulting (Already Started)

  • Helping others build similar tools
  • $200-500/hour

Lesson: You can monetize without paywalling your core product.

What's Next (Days 18-90)

Days 18-30: Stability

  • Fix reported bugs
  • Improve performance
  • Add user-requested features
  • Scale infrastructure (still free tier)

Days 31-60: Growth

  • Reach 10K users
  • Launch template marketplace
  • Add community features
  • Start podcast interviewing users

Days 61-90: Monetization

  • Launch optional premium tier
  • Onboard affiliate partners
  • Explore sponsorships
  • Hit $10K/month revenue

Days 91-365: Scale

  • 100K users
  • $100K/month revenue
  • Open-source everything
  • Build a team (if needed)

The Thank Yous

To everyone who:

  • Reported a bug
  • Suggested a feature
  • Shared HarshAI
  • Left a encouraging comment
  • Trusted me with their workflows

Thank you. This wouldn't exist without you.

Your Turn

What's stopping you from building something?

  • No idea? → Solve your own problem
  • No time? → Start with 30 minutes/day
  • No money? → Free tiers exist
  • No skills? → AI will help you learn
  • No audience? → Build in public

I started with an idea and 17 days. You can start today.

Try HarshAI

Live: harshai.vercel.app

GitHub: Coming soon (open-sourcing everything)

Follow the journey:

  • Dev.to: /aiavantage
  • Bluesky: @aivantage.bsky.social
  • X: @aiavantage

17 days. 52 integrations. 847 users. $0 spent.

This is what's possible when you stop waiting and start building.

Day 17 complete. Day 18 starts tomorrow.

What will you build?

Top comments (0)