What are you building, and what does your stack look like?
by•
I am a Computer Science student doing research into how solopreneurs and small startups create new apps and what their stack looks like. Particularly, I'm interested in how you handle things like authentication, billing, and permissions/authorization in your apps.
Let me know what you're working on below and how you're going about it -- I'd love to connect for some quick calls to learn about your product and talk about your process in building it!
960 views
Replies
Hi! We’re building https://app.hirepilot.co/ an AI job search assistant. It’s a tool that helps people manage their job search in one place, automate repetitive parts of applying, and experiment with outreach instead of just submitting resumes into portals. Core stack is Python with Django on the backend, PostgreSQL as our main database, and TypeScript on the frontend to keep things maintainable. We use OpenAI for AI-driven features and Cloudflare for performance and security at the edge.
@vik_sh Nice! I noticed the application tracker board right away, I just talked about something like that. Love to see it!
Are you utilizing Django's built in management dashboards? How has the integration across your Django API to your TypeScript frontend been?
If you've got a moment to hop on a call, I would appreciate your time -- find a time that works for you here!
I'm building rankrankrank.com.
Backend: C# Web API - using JWT Authentication, a vertical slice architecture using Mediatr and driven by a SQL Database
Frontend: Next.js - first time I use this over regular React. Really enjoyed it and the SSR I really needed for SEO purposes.
Hosting: Azure - Frontend is a Static Web App and Backend a regular Web App.
So yeah, bit of a different story compared to all the new and hip Vercel and Supabase hostings. Also no AI involved here. I'm not a dinosaur, I swear.
@bruce_rankrankrank Ha ha, totally get doing it the "old fashioned" way. If you've got the skills to build and manage it yourself, it's something that's hard to perfectly replicate, both price-wise and fitting it in your existing workflow.
Do you have a lot of background in SEO? This is definitely a useful tool for preliminary research. Beautifully simple, and has the creature comforts with things like CSV export. One thing that may be the cherry on top would be a PDF export and/or public share link. It's awesome that it isn't a subscription and has quite a low barrier to entry with your $1.99 credit pack.
If you have some time in the coming days, I'd love to talk with you about your stack and process. You can find a time that works for you here, or let me know if none of these times work.
@ryan_hendrickson Thanks for the feedback! Booked a slot :)
PrivateACB — Desktop crypto tax calculator for Canada, US, Australia, and UK.
Stack:
- Tauri (Rust + React/TypeScript) — Desktop app, not SaaS. Rust handles all the financial math with
lossless decimal precision. React frontend, Vite bundler.
- SQLite + SQLCipher — Encrypted database, all data stays on the user's machine. Big selling point for
tax software.
- Typst — PDF generation for tax reports (CRA Schedule 3, IRS Form 8949, HMRC SA108, etc.)
- Cloudflare Worker — Tiny license key validator (~200 lines) that checks Stripe payment status. That's
the entire "backend."
- Astro website on Cloudflare Pages — Auto-deploys from GitHub. Stripe Checkout for payments.
For auth/billing/permissions specifically: Since it's a desktop app, there's no traditional auth. License
keys validated against Stripe via a Cloudflare Worker, stored locally with Windows DPAPI encryption. 8
license tiers control which jurisdictions you can access. Stripe handles all billing — zero custom
payment infrastructure.
Biggest decision was going desktop over SaaS. No hosting costs, no GDPR headaches, no database scaling —
"your data never leaves your machine" resonates with people trusting you with their financial data.
Tradeoff is update distribution, but Tauri handles that.
Solo dev, happy to chat if useful for your research!
@robert_vassov Love that you took the path of keeping it local and private over a SaaS model, it fits well with the idea of crypto and anonymity. Does it focus solely on that piece of your taxes? At least in the US, does it work with larger platforms such as TurboTax, or is it designed for those who do taxes by hand?
@ryan_hendrickson Yea, I specifically developed it because there limited tools to help those who wanted to do their own calculations. It fits between a spreadsheet and the SaaS model. To answer your questions — yes, it focuses specifically on the crypto capital gains / income piece. It's not a full tax return tool. For US it produces IRS Form 8949 and Schedule D which are exactly what TurboTax (or any tax software / your accountant) needs.
I do however, want to develop API fetch and populate tables (currently wired for CSV input only) - this is a bigger task, but the program is stubbed for this process.
Maybe someone out there interested in collaboration on building a robust API engine for crypto data :)
Velocity: AI User testing
Building @Velocity: AI User testing mostly with Figma Make, Figma, Claude Code, Node, sitting on GCP. We just do loads and loads of interviews, I believe interviews are the best roadmap. Off the back of them we prioritise and ship!
@kevin_mcdonagh1 Nice! Yes, interviews really get you the insight, instead of just off the building what the user may not even want!
Love the concept of Velocity, testing it as much as you want with AI before asking/paying people for their time to test it. If I'm understanding correctly, it works on your Figma prototypes so you can iterate before you start building the actual product off the designs? Seems like your website isn't rendering correctly in Firefox, so I'm having a harder time fully checking it out.
Would love to chat if you or a member of your team has time in the next few weeks! You can pick a time that works for you here.
Building Aitinery (aitinery.com) — an AI travel planner that actually reasons about your trip instead of just generating a list. You say "10 days in Japan, solo, quiet temples and good coffee" and it builds a day-by-day itinerary with real logic: neighborhood flow, timing, what's worth skipping.
Stack: Next.js + Vercel on the frontend, Supabase for auth and DB, Claude (Anthropic) as the core reasoning layer. Auth has been the smoothest part — Supabase handles it without much friction. The hard part has been making AI output feel structured and actually editable, not just a wall of text the user has to manually rework.
Pre-launch now, aiming for PH in a few weeks. Happy to connect if you're building in AI / consumer space.
@giammbo Love how you've nitched the product down to specifically Italy so you can really serve that particular market, instead of spending time trying to cater to a large population of people. Agree with what you said about Supabase handling it without much friction. Have you had a chance to set up billing for Aitinery, or not yet?
@ryan_hendrickson Ha — Aitinery is global, not Italy-specific. The name is 'AI' + 'itinerary' so I can see the read. The philosophy is closer to what you described: specificity of context over breadth. Instead of building a generic tool for everywhere, the focus is making the reasoning actually work for whatever trip you describe — Japan, Portugal, wherever.
On billing — not live yet, pre-launch. Stripe is the obvious next integration with Supabase. Have you seen interesting patterns from other founders on when to set it up — before vs. after first users?
Hey Ryan — solo founder here building AnimalID (animalid.app), an AI-powered pet health records app. Upload a vet document, AI extracts medications and vaccinations automatically. Also does lost pet QR codes, weight tracking, photo galleries.
Stack:
- Frontend: Next.js 16 (React), deployed on Vercel
- Backend/DB: Supabase (Postgres + Row Level Security for auth/permissions)
- Auth: Supabase Auth (email/password, handles sessions and RLS enforcement)
- Billing: Stripe Checkout with webhooks syncing subscription state back to Supabase
- AI: Claude API for document extraction
- Storage: Vercel Blob for photos/documents
- Email: Resend
The interesting part: I built the entire app using an agentic AI workflow — I architect and orchestrate while AI writes the code. Codex for general code, Gemini for UI code reviews, Opus for high-level architecture discussions. Solo founder, no team, went from zero to production in one week. Still tweaking daily.
Happy to chat if you want to dig into any of it.
@jeffla Love this; does it have a mobile app to go along with it for quick access at the vet or elsewhere, or is it solely a web app? On your AI workflow, how did you land on the set of agents that work best for your use case, was it just a lot of experimentation?
You mentioned that you architect alongside Opus it looks like, how do those conversations go? Is it you suggesting an architecture and asking it to "shoot holes" in it, or describing your idea and asking it to design an architecture around that?
Thanks@ryan_hendrickson Right now it's a PWA — works on any phone, tablet, desktop, whatever. On mobile you can install it to your home screen and it feels pretty close to native. I'll build actual native apps once I know there's real demand, but as a solo founder the PWA lets me ship everywhere at once while I figure out if this thing has legs. Main use case is vet visits — pull it up, show the doc your records, done.
The agent workflow was a lot of trial and error. Started with one model doing everything and hit a wall fast. The big shift was realizing different models are just good at different things. Opus thinks like an architect, Codex cranks out code, Gemini's surprisingly solid on UI stuff. Once I split those responsibilities it all kinda clicked. Built the whole MVP in about a week. I'm using Openclaw as the orchestrator, 'cause - fun! 🤩
With Opus it's less "here's my design, tear it apart" and more like a working session. I'll come in with a problem — "here's what I need, here's what we have so far, here's the guardrails" — and it proposes an approach. Then we go back and forth. I push on edge cases, throw constraints at it, poke holes. The trick is giving it enough context about what already exists so it's not designing in a vacuum. Honestly? Not that different from whiteboarding with a senior engineer.
So my project is a lot more humble than most here. 3D Figurine Generator is a MicroSaaS I threw together for my 16 year old son. He likes 3D printing and is just getting into Dungeons and Dragons (Thanks Stranger Things!) so I thought it would be fun for him to be able to quickly and easily generate print ready STL model files of his and his friends characters based on some simple choices and a description. I've been a software engineer for almost 30 years so it was also an excuse to really try out vibe coding to the extreme. My goal was to not write a single line of code myself. Can't say I got that extreme, but pretty close.
Stack (picked mostly because it's what I work in the most for work right now):
Spring Boot for back end service
Angular for front end
Auth:
Rolled my own, but I'll throw in a recommendation for FusionAuth for small startups. They are great. I should know I used to work there a few years ago. I didn't use them just because this was so simple.
Payment/Billing:
Just went with PayPal because I already had the account laying around. Might add Stripe later. I just went with a pay as you go credit based system and I just track the credits in my own DB.
Tools/Process:
Fired up VS Code, installed Claude Code, used it to author everything for the front end and back end.
Used Amazon Q to help me setup all the cloud infrastructure. Went with Elastic Beanstalk for the back end and just a S3 bucket with static hosting forthe front end. Simple Postgresql RDS.
@robdavis967 This is awesome -- super cool that you made this for your son. How do you generate the STLs themselves, is there an external model that you use for that?
Couple other questions here; How was the process of integrating PayPal into your app, compared to something like Stripe? How long did this take you to make, and was there any speed bumps in the process that took you longer than you hoped to resolve?
Building Suitegenie, a social media automation suite for agencies and founders who want to grow on LinkedIn and Twitter/X and threads without spending hours every day.
The core idea: one dashboard to schedule posts, auto-engage with comments, and generate AI-powered content tailored to your brand voice, across both platforms.
Stack:
- Frontend: React
- Backend: Node.js + Express
- Queue: BullMQ + Redis (for handling scheduled posts and automation jobs)
- DB: Supabase (Postgres)
- Auth: Custom-built auth system
- AI: OpenAI for content generation
- Billing: Stripe
The hardest part honestly has been dealing with Twitter API rate limits and LinkedIn's strict OAuth scopes. Happy to chat about that if you're researching API-heavy products!
We're launching on Product Hunt soon, would love any feedback or support!
@kanishk_saraswat I can just imagine how much of a headache dealing with the Twitter and LinkedIn APIs were, especially nowadays.
I'm curious about your custom auth; is there a particular reason that you didn't utilize Supabase's auth features?
Would love to chat if you've got a moment in the next few weeks; pick a time that works for you here!
I'm currently building my first indie app! Still in the early stages, but I'm using Swift/SwiftUI for the frontend and Firebase for the backend. Trying to keep the stack as simple as possible so I can ship faster!
@thanh_cong_tr_n_van Sweet! Wishing you luck on your app!
Love this thread! Building HelloAria — an AI productivity assistant that works through WhatsApp, Telegram, Email, and has a native iOS app + web dashboard.
Stack:
• Backend: Node.js + Express
• Database: MongoDB + Redis for caching
• AI: OpenAI GPT-4 for conversational intelligence
• Integrations: Google Calendar, Gmail, Microsoft OneDrive
• Messaging: WhatsApp Business API, Telegram Bot API
• iOS: Swift/SwiftUI
• Auth: Custom JWT + OAuth2 for Google/Microsoft
• Hosting: AWS
The biggest challenge was making the AI feel conversational and not robotic across different messaging platforms. Each platform has its own quirks with formatting, media, and message limits.
For billing, we went with a simple freemium model — free tier covers basic reminders and todos, paid tiers unlock calendar sync, team features (Circles), and priority AI responses. Kept it simple because overcomplicating pricing early on kills conversion.
Happy to chat more about any of this!
@sai_tharun_kakirala Super slick website, love the design and storytelling. When you think of the primary way to interface with the AI, is it through your web/mobile app, or is it via a messaging platform like Telegram or WhatsApp? I like that you can send it a voice note instead of just text, that's super helpful and a big time saver.
I see your custom auth setup, where the assistant issues you a one-time code to login -- very interesting setup! How are you managing billing at the moment, is it through Stripe or another platform?