Programming languages for the next 20 years

No Single Language Wins

People ask us which programming language to focus on for the next ten to twenty years. There is no single winner, but there is a strong pair: Python and TypeScript, with Rust as an optional third.

It is also the pair we use ourselves. Our backends run on Python and every frontend we ship is TypeScript.

Python For AI And Data

Python is the language of AI and data. Model training, fine-tuning, embeddings, retrieval pipelines and agent backends all run on it, and that is where most of the new work is landing.

TypeScript For Products

TypeScript is the default for the web, for React Native and Expo mobile apps, and for most MCP servers and AI-integrated apps. One codebase can cover iOS, Android and the web.

Rust Once The First Two Are Solid

Rust keeps growing in performance-critical and security-sensitive work: infrastructure, on-device AI runtimes and WebAssembly. It is worth learning after Python and TypeScript, not instead of them.

The Bigger Shift Is Not The Language

AI will write more and more of the routine code. The lasting value moves away from syntax and towards these kinds of work:

• AI-native apps and agents: products built around language models, such as copilots, agents that take actions, and MCP connectors to data and tools.
• Integration and data plumbing: connecting messy real-world systems like ERPs and payments to AI.
• On-device and edge AI: small models running privately on phones, which matters most where cost and connectivity are real limits.
• Verification, security and reliability: reviewing, testing and securing AI-generated code and the behaviour of agents.
• Domain expertise plus software: people who deeply understand a field and can ship software for it will beat generic coders.

What We Would Focus On

Go deep on Python for AI. Stay sharp in TypeScript for products. Invest heavily in system design, data modelling and directing AI coding tools well.

Languages will change. Those skills carry over.