What are you building, and what does your stack look like?

We're building Copus — a curation platform where people save, organize, and share the best content they find online. Think of it as turning your bookmarks into a living knowledge network others can discover.

Stack: Next.js + TypeScript on the frontend, Node.js backend with PostgreSQL. Elasticsearch for search and discovery, Redis for caching. We also built a Chrome extension for the save-from-anywhere workflow.

For auth we went with NextAuth.js — easy to set up with multiple providers. The hardest part has honestly been building good content discovery without just becoming another algorithmic feed. We lean heavily on human curation, which is kind of our whole thesis — real people surfacing the best stuff beats algorithms.

Thanks Ryan! Love that you are taking the anti-algorithm route too. It is honestly the hardest design challenge — resisting the urge to just optimize for engagement. We have been experimenting with letting people subscribe to curators they trust instead of feeds, which has worked really well. Would definitely be down to chat — we are actually launching Copus on PH tonight so this week is a bit wild, but let me check out your calendar link after the dust settles!

Working on a privacy-first web tools app — keeping it simple with no auth for most features, and using basic Stripe billing + minimal backend only where truly needed.

building speakeasy — an ios app that turns any article url into audio. paste a url, get ai tts in ~30 seconds, saves to icloud

stack: expo + react native for the mobile side, fastapi backend, inworld tts (with openai as fallback), postgres. for auth its just device id stored in keychain — no accounts needed which keeps onboarding super minimal

biggest challenge has been text extraction — different sites block scrapers differently. revenucat for subscriptions has been great tbh, saves a ton of time vs rolling your own

been building a few ios apps under studio.gold 🔨 three live rn: speakeasy (turns any article url into audio), astrologica (ai personalized daily horoscope podcasts based on ur birth chart), and wordplay (daily cryptic crossword puzzles). stack is swift/swiftui native, firebase for auth + analytics, revenuecat for subs/billing which has been a lifesaver honestly. for the ai stuff in speakeasy and astrologica im using openai apis. biggest learning: revenuecat handles like 90% of the billing headaches so i didnt have to build any of that myself. happy to chat more if anyones curious about the indie ios stack 🙌

solo maker building 3 ios apps under the studio.gold umbrella. all native swift/swiftui, mostly vibecoded with claude honestly

1. speakeasy (speakeasy.studio.gold) — paste any article url, converts to audio. stack: swift, AVFoundation for playback, server-side tts pipeline. the core challenge was making voice quality good enough that you'd actually listen for 10+ minutes

2. astrologica (astrologica.app) — ai-generated daily horoscope podcast from your birth chart. stack: swift, openai api for generation, server-side audio rendering. the interesting bit is generating personalised content daily that doesn't feel repetitive

3. wordplay (wordplay.studio.gold) — one cryptic crossword clue per day. stack: swift, custom clue engine. the "only one per day" constraint is deliberate — forces me to make each clue really good rather than generating volume

for auth/billing: revenueCat for subscriptions, supabase for the backend. kept it simple on purpose — no fancy infra, just "does this work and is it cheap to run"

the common design principle: single-serving daily content. no infinite scroll, no engagement loops. each app does one thing then gets out of your way 🛠️

@ryan_hendrickson Great questions! The primary interface is actually both — we have an iOS app, but the real magic happens through WhatsApp and Telegram. Users manage their entire day (reminders, todos, calendar, meeting notes) just by chatting in WhatsApp without ever opening another app. The voice note feature you liked is a big part of that — you can dictate a task while driving and Aria just handles it.

For billing, we're using RevenueCat + App Store subscriptions on iOS. Kept it simple for now since we're iOS-first — it handles entitlements cleanly and has great analytics built in. Happy to connect if you want to dig deeper!@ryan_hendrickson Thanks so much! The primary interface is actually the messaging platforms — WhatsApp and Telegram are the core experience. The iOS app is more of a companion dashboard. The big insight was that people already have WhatsApp open all day, so we meet them where they are instead of asking them to open another app.

For billing, we're on Stripe — works great with RevenueCat on the iOS side for in-app purchases. Happy to share more if useful!

Building Social Intel (https://socialintel.dev) — Instagram influencer search API for AI agents. The twist: payment via x402 micropayments (USDC on Base). Agents find the API, call it, and pay automatically. No subscription, no API key for the caller. Stack: FastAPI + fastapi-x402 + MCP server. Anyone else building in the x402/agent payments space?

Building Genie - an AI analyst inside Databox. Users ask questions in plain English ("Why are leads down this week?") and get instant answers, charts, and dashboards from their connected data sources.

Stack for the AI layer: LangGraph for agent orchestration (evaluated several frameworks - LangGraph gave us the control we needed for stateful, multi-step tool-calling without abstracting too much away), LangSmith for tracing, Claude + OpenAI depending on task complexity, AWS Bedrock for RAG. We also just shipped Databox MCP so users can query their live data from Claude, ChatGPT, or any MCP-compatible tool.

On auth/billing - we're an established SaaS so we have our own systems, but the most interesting challenge on the AI side has been usage metering. Every Genie response chains multiple tool calls, and the cost per conversation varies wildly depending on what the user asks. We landed on a credit system where simpler tasks cost less and complex ones cost more - transparent to the user, visible right next to each response.

The hardest part wasn't the AI. It was making sure every answer traces back to real data. Confident-sounding wrong answers are worse than no answer when someone's making a business decision.

Currently building AI developer tools focused on improving AI workflows.

Lately I've been experimenting a lot with:

• Python / FastAPI

• LLM integrations

• automation for developer productivity

Curious what stacks other builders here are using lately.