Hello Again

I've been a SWE for a few years now, and an adult for about as long. For most of that time, finding the energy to work on side projects and hobbies has been rough.

But over the last few weeks, I've been able to overcome my main blocker: I've effectively lowered my activation energy by working with modern LLMs and harnesses.

While my usage at work has steadily increased since the first GitHub Copilot, I resisted using these tools more for personal projects. It lacked clear evidence that they would reach a baseline level of intelligence at a reasonable cost.

I'm happy to change my position now. Models have become capable enough for my needs, with early signs appearing around Opus 4.5/4.6. The latest generation of models from OpenAI and Anthropic has also become affordable enough for me—at least as measured through GitHub AI credits across a series of commits over the last few months.

What does it mean to me?

Many of my personal computing tasks have shifted to an agentic workflow. Complex file-backup and management workflows are now skills that can evolve dynamically as I interact with my sessions.

It also means I've been able to make meaningful progress on projects before reaching the point of giving up on them. I expect to start sharing more of those things with the world.

What does it mean to us?

For us—the community of people who make software—it feels like a new Wild West. Engineers can iterate much faster on problems, platforms need to handle new kinds of loads, and anyone might be right about what the future will look like.

Welcome to my seventh or eighth attempt at having a blog.

Elisa