Programming with LLMs

Time to Pair up

Alex van Vorstenbosch

2025-02-01

Pair-programming with LLM-Chats

What can LLMs mean for you?

  • Writing code:
    • Great for code with extensive syntax, but simple logic:
      • Regex patterns
      • SQL queries
    • Getting better at complex logic which reasoning models such as:
  • Translation between code-languages:
    • i.e. Python to R
    • i.e. R to Python
  • Writing boilerplate-code
  • Writing documentation
  • Writing unittests
  • Explainig code
  • Debugging code
  • Reviewing code
    • Optimising code
    • Improving code

What can LLMs mean for you?

  • How does using LLM chat differ from github copilot?
    • General purpose versus only code specific
    • More interactive
    • Longer conversations
    • Ability to guide more in the Context
    • But also more removed from IDE
    • Cannot see code context from your project

Demo: GitHub Copilot Chat in VSCode

Demo

Writing code with LLMs

“What I love is that it will come out of left field with methods I didn’t even know existed. Of course in some cases those methods actually don’t exist…”

  • AdventOfCode
  • Example of a task where there reasoning models feels like a big leap forwards:
    • Can handle more complex code involving more distinct reasoning steps to solve

Writing code - SQL queries

Writing code - Regex patterns

Writing code - Pong in Python

Translation between code-languages

  • Remember this function from the exercises :)

Documenting code

Explaining code

Improving/Optimising/Reviewing code

Pair-programming with LLMs