My second most frequent bit of advice to people, after “never push on a string” is “never try to have both sides of a conversation”

But now with a bot frontrunning my consulting dealflow at 1/100th my consulting rate (if I charged for this at cost-plus billing and say reasonable % margin), I’m not even having my own side of my conversations. Hell, I’m not even seeing about 93% of these chats. Eating my own dog food and then some. Plausible deniability, thy name is parasocialized AI. This is not a trivial thought btw. It was publicly reported that that zizian cult leader apparently read some of my old writings. When that shitshow unfolded some friends were asking me how I felt being part of a psycho killer’s autodidact education syllabus. I was initially wary of even acknowledging that footnote to notoriety publicly. But with a live bot emulating my thinking to readers, it’s much harder to stay arm’s length. Building defenses against unwittingly becoming an enabler of psychopathy is becoming a non-trivial part of evolving this bot. A non-trivial chunk of this bot’s system prompt is devoted to such defenses. Also important to note that no matter how realistic I’m able to make this bot, it’s not going to be better than low-fi simulation at best, so there’s a need for treating it as a stylized interpretive rendering. Like an actor playing me.

People are suggesting I charge for this, but I have a cunning plan to make a true consulting-sparring partner bot. This vgr_zirp bot is just the first piece of a mixture_of_vgrs bot with enclaved NDAed conversational capability. I’m considering offering that at 1/10th my human bill rate. Not kidding. Genuinely interesting challenge. Once I have 2-3vmore bits like this built and tested, you’ll see me shilling a pilot program.