After 12+ months of using Claude Code & Ollama daily, I’m still not certain they are as big of a productivity boost for coding as when I added a second monitor back in the day.

There absolutely is a productivity gain for me, especially when it comes to doing research in the code base, and being more ambitious about starting big changes and refactors. I wouldn’t want to go without Claude Code in my workflow

I’ll definitely try any new models and tools, but there’s a loss of flow state with agentic coding tools, which offsets much of the gains. It’s definitely a 2 steps forward, 1 step back situation.

I also primarily work in a code base that I wrote ~50k lines of code and set up the whole architecture by hand before adding AI coding tools, which from what I’ve read is the best possible situation (versus letting them run wild from the beginning of a new project).

The biggest advantage of coding agents is that I can let them run while I’m in a meeting and feel like I still get stuff done. But that’s also an edge case many developers won’t care about 🤷‍♂️

TBD how much more the models and tools will produce. Maybe there are bigger breakthroughs ahead, or things will get worse because AI labs won’t be able to find profitability without charging 10x+ more for today’s service. We’ll see. I’m still using them in the meantime