NoKode: A Web Server Powered Entirely by an LLM

Software engineering usually means translating requirements into deterministic algorithms. But with LLMs, an intriguing question arises: where is the line between writing procedural logic and delegating decision-making to a model?

One engineer took this concept to its extreme and implemented a web server powered entirely by an LLM called NoKode. Instead of routing handlers or database queries, it infers user intent directly from the HTTP request path, method, and payload, generating appropriate responses dynamically.

While obviously impractical for production, it’s a fascinating thought experiment in how software architecture could look in a model-centric world.