[theturtletalks]

The LLM model battle might be lost, but the harness battle is just beginning.

If you can build a good harness around a weaker LLM and get good at prompting, you will still be able to out perform people using CC. CC has context bloat and even more checks to make sure it’s not being distilled, doing anything shady, or building a competing LLM. Those things add context bloat.

Hope is not lost, we just need to open a new front.