- The nature of software engineer is changing fast
- AI is much better at coding than most junior/middle engineers
- AI is not just a tool, it's a new platform/OS
The Death of Product Development as We Know it
- Deeply understand the AI market: needs, limits, trends.
- Study failed and successful products.
- Build, Iterate faster
- Streamline app production "not" to stifle creativity, but to reduce time sinks and focus on what we do best.
=> We must ship quickly to get real feedback.
- Keep it simple: Monolith > microservices (Facebook ran a monolith to 1B users).
- Optimize for the next 6 months, not 3 years.
- Cloud-first: Self-host only for tech limitations or 20x cost savings.
- Choose what LLMs are good at: Popular, well-documented stacks
- Mobile apps: react native with expo
- cross platform
- good tooling(dev, build, share, debugging)
- it's js so LLMs are familiar with it, also easier to hire devs
- Firebase
- all in one: authentication, database, media storage, server-side functions
- Payment: revenuecat
- AI pipeline:
- recommended to use provider's APIs to start with if possible
- pipeline builder: comfy UI
- AI agents:
- claude code
- cursor
- Use claude.md or cursor.md
- Ask it to asks question for clarification and generate a plan before implementation
- Context is everything, think of it as you're working with a middle engineer that was just onboarded today.
- Apps should include privacy and terms of use
- Subscription based apps should include sandbox account when submitting
- Should have a restore purchase flow
- Use real screenshots for the first submission
- Generate a random UUID on start and store it in keychain(ios)/keychain(android) so it's synced across devices of the same accountreact-native-keychain
- Use this ID for
- revenuecat user id
- firebase user id (use signInWithCustomToken, generate the token by using firebase function)
- We can link user accounts(google, facebook, apple etc) in the future to this ID if we want to