How I started programming differently over the last year. What about you?

How I started programming differently over the last year. An interesting observation: I’ve stopped using the LLM-powered autocomplete in my IDE. At…

By AI Maestro May 16, 2026 2 min read
How I started programming differently over the last year. What about you?

How I started programming differently over the last year.

An interesting observation: I’ve stopped using the LLM-powered autocomplete in my IDE. At first, it was one of the key features for me; it felt extremely convenient to have a language model complete functions based on common sense or context from open tabs. However, after six months, I developed a script that could go through source files and generate appropriate context for an LLM chat, which would then suggest what to add or fix.

Nowadays, it’s easier to use a CLI interface with a coding agent without even launching the IDE. You describe your needs, point it to the relevant files, and everything is streamlined. The only things I still rely on my IDE for are nice Git diff visualization, step-by-step debugging, and jumping into function implementations via code navigation. This functionality is used in about 5-10% of my work.

It’s fascinating to think about what will come next. For instance, with an all-products subscription from JetBrains covering multiple languages (Java, Scala, Python, TypeScript, Rust), why should I keep paying?

I’ve already tried another approach: using a coding agent that can read the logs of program behavior and ask for additional context when needed. When debugging becomes complex, I instruct the LLM to write tests based on these logs.

One recent technique involves creating a plan.md file where I outline tasks. The LLM then completes one task at a time from this plan. This method helps break down large problems into manageable steps.

In summary, the industry is evolving rapidly. How has your approach to programming changed? I’m curious to hear how things have evolved for others who have been programming for longer than me (since 1990).

Key Takeaways

  • I no longer use LLM-powered autocomplete in my IDE.
  • I’ve developed a script that generates context for LLMs, which I previously relied on the autocomplete for.
  • The industry is changing rapidly with new tools like coding agents and logging interfaces becoming more prevalent.

Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.

Name
Scroll to Top