This is a quick follow-up to my previous post about agentic coding from November 2025. Not much has changed in the overall approach, but there are two notable shifts in my tooling worth mentioning.

Open Code for Local Development

While I still use VS Code with GitHub Copilot for my day job, I’ve been increasingly reaching for Open Code for local agentic development on personal projects. The terminal-based workflow feels snappier, and I find myself more productive with it for focused coding sessions. My AGENTS.md files and general workflow patterns remain mostly the same.

I really like the flexibility of Open Code and since I’m a Terminal guy anyway, it fits well into my existing habits. If you haven’t tried it yet, I recommend giving it a shot. Currently, I can’t use it for work because it needs an API key instead of integrating directly with GitHub like Copilot where I’ve got an enterprise license.

GitHub Copilot for Async Work via Issues

The biggest shift is how I handle smaller personal projects. I’ve moved almost entirely to assigning GitHub Issues directly to Copilot and letting it work asynchronously. The workflow is simple:

  1. Create an issue describing the task
  2. Assign it to Copilot
  3. Wait for the PR to be created
  4. Use Copilot again to review the PR
  5. Do a quick manual test without diving into every detail
  6. Merge

This mostly just works. It’s a great way to kick off work when I have a few minutes or a new idea strikes me. I can jot down the issue quickly on my phone and leave Copilot alone to do the heavy lifting. When the PR is ready, I can review it when I feel like it. Even if the first attempt isn’t perfect (or isn’t working at all), I don’t care about throwing it away and fine-tuning the issue to get what I want. Sometimes the PR shows that the initial idea wasn’t quite right. I don’t care, because I spent minimal time upfront and can quickly iterate. AI really allows me to try out ideas without larger time commitments, which leads to way more experimentation and learning. With every new model update, I find myself trying out new ideas more often and realizing way more personal projects and solutions to one-off problems than I would’ve done otherwise. I strongly believe that agentic software development is a game-changer for the way we build software and that it will enable a lot of innovation in the space, especially allowing for more personalized, tailored solutions to niche problems.

Final Thoughts

The core principles from my previous post still apply — good instructions, focused toolsets, and clear plans make the difference. The main evolution is leaning more into async workflows and trying out Open Code as an alternative to Copilot’s agent mode. I’ll keep experimenting and share updates when there’s more to report.