Harshita Soni

Day 7 (Extended) You are watching a non-tech founder slowly become technical... ✅ Finally fixed the app. ✅ Better onboarding. 📚 Learnt: config.js, manifest.json, localhost, OAuth & frontend ↔ backend communication. Turns out the biggest win wasn't shipping. It was debugging

Day 7 Built: A smarter product. Broke: My assumptions. Turns out building software isn't writing code. It's getting every moving piece to agree with each other. Tomorrow: The app gets a memory.

@jubayers_r @antonosika @amasad @MaorShlomo This was genuinely helpful. Had to go learn what state boundaries and API contracts actually meant. Thanks for pointing me in the right direction !!! adding it to tomorrow's fixes.

@harshitabuilds @antonosika @amasad @MaorShlomo The agent broke it because it lacks context on state boundaries. Don't let it touch config files directly. Explicitly isolate your backend logic behind strict API contracts (JSON schemas) so client changes cannot structurally break the server.

Day 6 (Extended) Vibe coding as a non-tech founder:Prompting is the only easy part btw ✅ Shipped a better product ❌ Broke the app 📚 Learnt config.js, manifest.json, localhost & frontend ↔ backend. Is this the normal learning curve? @antonosika @amasad @MaorShlomo

Day 6 Built: A much better founder experience. Broke: The app. We’ll pretend that’s progress. Tomorrow we fix it.

Day 5- Extended Shipped: ✅ My first AI-powered Chrome extension ✅ Backend talking to OpenAI ✅ First end-to-end build Learnt: ChatGPT Plus ≠ OpenAI API API keys shouldn't be leaked. Twice. A product can work... and still be bad. Broke: My.env My confidence Tomorrow we rebuild.

Day 5 of trying to build an AI startup with Claude Code + ChatGPT Achievement unlocked: ✅ Built my first Chrome extension ✅ Connected it to AI ✅ Got my first real response Then I used it. Found 17 things I hate. Tomorrow we fix them.

Day 4.3 Today’s build log: ✓ Pivoted. ✓ Built a Chrome extension. ✓ Installed Git. ✓ Learned HTML isn’t an app. Spent more time learning how to work with AI… than asking AI to work for me. See you tomorrow.

Day 4.2 Installed Git. Apparently that’s step one. Also found out an HTML prototype isn’t actually an app. Today was incredibly humbling.

Day 4.1 Started the day building one startup. Spent 3 hours with Claude. Every idea: “Already exists.” “Someone built it.” “Why would anyone use this?” Fair enough. Pivoted. Apparently arguing with AI is now part of product discovery.

I spent today looking at SaaS products making $30K–100K MRR. Here’s what surprised me: None of them started with AI. They started with an incredibly boring problem Today’s rabbit hole: CAs spend tax season chasing documents on WhatsApp. Maybe that’s the product. Still validating.

Day 3. Today’s achievements: • Didn’t build anything. • Questioned every idea I had. • Somehow convinced myself chasing people for tax documents might be a startup. Zero engineering background. Zero coding skills. Maximum delusion Going great so far #VibeCoding

Day 2. Spent yesterday trying to buy a business instead of building one. Almost got fooled twice — once by fake growth, once by fake revenue history. Walked away with nothing. Correct call. Still don’t have the idea. Sitting with that instead of forcing it.

33+ startups scaled with my marketing. Zero of them were mine. Today that changes ! building my own product, non-technical, fully in public. Day 1.