This is the new classic with Holland's gene splicing. I let it run all night. Blue is lying and Pink is believing. I started believing and lying randomly at around 50%. The blip was interesting and deserved a blowup.

Blowing up the lying part after lying got to 1% gives us this. After an hour or two the liars find a con that works well for a while. A couple of strange plips occur later on for some reason.

When we look at the believers, we see that they do take note of the con and find a defense. The liars then go back to low level lying. The later blips are also in sync with the liars blips. It is the unbelievers that keep the liars in check. To check this hypothesis with Holland genetics, after the long night run, I hypnotized all the agents to always believe, getting the next graph.

They quickly start lying their little agent asses off. Let me show you a blowup of the last part.

Notice that it takes quite a few rounds, about 800 die before they catch on that they can get away with lying. Then it hovers at 50% for a while. What's happening here? I conjecture that they only learned that they could lie about one of the 2 matrices that is given to the potential liar at horizontal spot 80. It takes a while for them to do another mutation so that they can set off lying on the other matrix. After the mutation, it quickly becomes a winning plan.

I'm still working on the lying, truth telling, silence program, but it seems to have some weird bug I haven't figured out yet. Meanwhile, I have another computer working for quite a while on starting with all intelligent lying and no believing. I'm hoping that it ultimately converges on 1% lying and 99% believing. It's been running for about 14 hrs and up to 41% believing but it's been as high as68% believing. I'm going to let it go a while longer, at least 24 more hrs.