Earlier I was of a view that APIs are so powerful that we can just build UI around them and that they can be a standalone product
Like tape together 2-3 Apis and you would have a new application
For example Uber can just be a mix of Google Maps, stripe and some database that holds users and riders
Tradingview can just be Price data API from stock exchanges
We can abstract away application logic for most big software companies to just handful of APIs
But I realized just APIs keeps you constrained
They just give you some output and tell you to take it like a good boy
You cannot ideate by just looking at some API and think that I can turn this into a product
This is where libraries can come in
Out there are so many amazing libraries that are end to end features in themselves
APIs constrain you to their exact use case.
Stripe does payments. That’s it.
But a library like Recharts or D3 gives you composable primitives
You can build 1000 different visualizations from the same toolkit.
Libraries encode domain knowledge and flexibility that APIs can’t.
So we know that Recharts or D3 have thousands of stars on GitHub and people use Stripe to pay, Could there be a possibility that people would want to pay for good visualization softwares?
Yes! of course!
If you are able to somehow look a few libraries and some Apis and can conjure up some idea that fits a market
Then you’re golden
And this is something that can be built on top of just AI powered coding
With so much work Outsourced to external Apis and libraries, the AI is just left with orchestrating whatever is in the codebase
And that’s for the best
AI is overconfident as fuck and it will try to build things from ground up and waste hundreds of dollars when 50KB library that does it flawlessly already exists
In this paradigm you have to use AI as the doctor that creates the Frankenstein using all this external Apis and libraries and you are there to just collect the cash
AI flourishes the most rare binary results exist, The condition where stuff either works or fails
That’s why it is better at back end than at front end
Front end might look like shit but hey if it is not throwing any error it is fine by AI
So by heavily relying on external Apis and prebuilt libraries you are just connecting things using AI and either they are connected successfully or they are not
Thus you are forcing your environment to be such that it is really favorable for AI to succeed in
The existence of libraries or Apis can actually be a good signal for you as a founder
If the problem was big enough that even open source communities or some companies are creating a piece of software for it then that’s a good sign
If no code exists out there for it then maybe it’s not such a big problem that anyone bothered to work on it
Or maybe you just hit a fucking jackpot that 8 billion people haven’t thought about
I would rather not play any lotteries
So if code doesn’t exist for it either in form of accessible API or existing open source libraries, I would think that the problem is not big enough for people to pay attention to it and thus maybe not even pay for it
Plus existing code has survived the test of time and production
It has accounted for thousands of microinteractions and edge cases
Something you just cannot vibe code or even normally code unless you have years of context
A good library represents maybe 50,000 hours of collective debugging across thousands of production environments.
You’re not replicating that. Ever.
Use the existence of mature tooling as a proxy for market validation.
If you’re building a B2B SaaS and there’s zero existing infrastructure around your problem space
No open source projects, no API providers, no one even complaining about it on GitHub
That’s a massive red flag.
Claude’s Reaction :
The Framework:
APIs = Data & Processes (external resources you consume)
- Stripe: payment processing
- Google Maps: location data
- OpenAI: AI inference
- Twilio: SMS delivery
- Alpha Vantage: stock market data
Libraries = Functionality (capabilities you build your product around)
- Recharts/D3: visualization functionality
- Three.js: 3D rendering functionality
- PDF-lib: document manipulation functionality
- TensorFlow.js: ML functionality
- Tone.js: audio synthesis functionality
Product Building Pattern:
Library (functionality) + APIs (data/processes) = Product
Recharts + Stock API = Trading dashboard
PDF-lib + Stripe = Document signing platform
Leaflet + Location APIs = Fleet tracking app
Chart.js + Analytics APIs = Business intelligence tool
The Validation Signal:
Popular library = proof that functionality is hard/valuable enough that people built sophisticated tooling for it
Popular API = proof that data/process is commodity infrastructure you can rely on
Your Job:
- Find library that represents valuable functionality
- Identify who needs that functionality in a specific workflow
- Build UI/UX around the library
- Use APIs to handle data and processes
- AI orchestrates the glue code between them
The library does the heavy lifting (visualization, 3D rendering, PDF parsing). The APIs provide the inputs/outputs (data, payments, storage). You provide the workflow and UX. AI writes the integration code.
This is the recipe. Simple as that.