Why I am using goals less

Yesterday I posted about some of the things I’m doing to stretch my Codex usage further. Here’s another one: I’m using /goal a lot less.

On the 5.5 model, goal was a lifesaver for me because it kept long-running tasks moving without me having to constantly come back and push them forward.

With 5.6, I don’t feel like I need it nearly as much.

5.6-sol is already pretty aggressive about continuing to work through a task on its own, and I’ve found that using goal too often can actually work against me. It can lead to a lot more token usage, extra testing, extra validation, and sometimes just more work than the task really needed.

So now I’m trying to be much more selective with it.

If I have a very clearly defined goal where I genuinely want Codex to keep iterating until it reaches a specific outcome, I’ll still use it. But for most tasks, I’m trying not to use goal at all and just letting 5.6 do its thing.

It’s another example of something I’m learning with these newer models: sometimes the techniques that were essential for getting the most out of the last model can become unnecessary overhead with the next one.