Sharpee Paused

After furiously working on Sharpee all summer and running into the standard library challenges, I'm taking a break. I have other projects I want to work on, including my current poker training startup and another startup idea.

Even I'll admit I was overly optimistic about getting it completed by now. The upside still remains: the overall design and architecture have withstood a massive list of testing scenarios. The cons are that the standard library and world-model require magnification from my own development capabilities and zooming out on the GenAI side.

My original goal was to let GenAI do 95% of the work. That number is likely to drop to 75% or lower, which aligns with projects I see other people working on using OpenAI and Claude (for coding). It's not that GenAI can't construct the actions in some way. It's that I can construct the logic for each action faster than explaining all of the details to GenAI. I still use GenAI to discuss options, pros and cons, consequences, and scenarios; but I'm writing the code myself in many cases.

A fun example is pushing. GenAI will make the action mostly correct, but also add behavior for sounds (pushing buttons), which is fine but not necessary in the standard library. A lot of other examples of behaviors you'd see in a video game, but generally not in a text game. So I'm tackling each action, defining the base behavior, and implementing it. This way I can check off each action as "working as I expect it to work".

I'll get back to it later this fall, but looking to complete something else in the meantime.

As always, stay tuned.