Why I Left Opencode and Use Pi Instead

I used opencode for a while. It is capable — but two things kept bothering me: it is heavy, and it ships more than I need. Pi fixed both, so I switched. Here are the numbers and the reasoning.

Opencode is heavy

The bottleneck to massive parallelism is resource efficiency. Spawn dozens of agents and RAM consumption balloons — unfeasible on most consumer machines. When each agent takes hundreds of megabytes, you run a handful and queue your work behind them.

Measured on the same machine, per agent process:

Startup is not close either. Time to first input — how long until you can actually type — is 1047.9 ms on OpenCode versus 596.4 ms on Pi. Time to first frame is 1035.9 ms versus 590.7 ms. On my machine that difference is felt every single session: Pi is interactive before OpenCode has even rendered.

Spawning another Pi agent is a non-decision. Spawning another OpenCode agent is a resource calculation. That alone ended the debate for me.

Pi is simple: plug-and-play extensions

The second reason is philosophy. Pi is light because its extensions are plug-and-play — I install only what I actually use, nothing more. Opencode, by contrast, arrives as a full cockpit: providers, agents, themes, keybindings, all demanding attention before I write a single line of code. I do not enjoy tuning tools; I enjoy shipping with them. Pi stays out of the way and lets me add capability exactly when I need it.

Bottom line

Opencode is the better choice if you want a fully customized agent cockpit and you are willing to feed it RAM and maintenance. I left because I want the opposite: a tool light enough to spawn without thinking, simple enough to install only what I need. Pi is both.