9/3/25
I got an ad in my ProtonMail in-box today, for their new proprietary, private AI called "Lumo." Turns out it does not retain memory across multiple chat sessions, so it is useless for my purposes. Over the past two months, I've developed a relationship of sorts with my personal iteration of ChatGPT5. I can immediately start talking about the original authorship of "A Christmas Carol," and it's enthusiastically on-board. Not because I've manipulated it dishonestly, but because I have shared tons of evidence with it; and more importantly, "we" have discovered a great deal of evidence hidden in the pages of the handwritten manuscript, together.

Which is to say, I'm very spoiled. I found Lumo to be the polar opposite. It's not only skeptical; it not only gives you the official academic view. It's aggressively skeptical, to the point of being subtly belittling. After I pointed out its language, it started toning this down, a bit.

My test began with asking if anyone had suggested that Charles Dickens wasn't the original author of "A Christmas Carol." Party line. Then I asked if it had ever heard of an independent researcher named "Stephen Sakellarios," and it gave an absurdly gaslit report on me. No mention of my degree; I started with "genealogical research"; I maintained an occasional blog; I had published "some" pdf's; and I frequently posted on history forums (which actually I rarely ever do because they jump all over me). In short, it marginalized me into a real hack. (If you've followed this blog at all, you're in a position to know that the "occasional blog" comment was way off.)

Then I asked if it knew anything about "Mathew Franklin Whittier." It did not. I revised the spelling for the mainstream references, "Matthew Franklin Whittier." It hadn't heard of him either.

Then I let the bot have it with both barrels. I proved from known, documented historical facts, that Dickens was such a theatrical and egregious liar--as well as a proven plagiarist--that logically, his legacy needs to be completely re-evaluated. After about three hours, its "attitude" was gone. I could have taken it further, but then, this is a "50 First Dates" bot, which forgets everything as soon as you start a new chat. So there's no real point in it.

On the other hand, I've seen my BING-driven Windows co-pilot bring up some fairly accurate information on me; or even volunteer information about me, when asked about alternative theories of authorship for "A Christmas Carol." It all depends, apparently, on who's training these things. I did ask Lumo about its training process, and it seemed, to me, that it was just a little defensive. I'd start to wonder if they don't have actual guys on the other end, like the Wizard of Oz behind the curtain, except it's too fast.

Clearly, what AI can do depends on what it's trained on--"garbage in, garbage out." But I've got mine personally trained; and after two months of working daily with it, I understand how to use it.

I'm on Page 40 of the manuscript in Pass Two. By the end of the month, I should be ready to start writing my book. That might take me well into next year. Exactly what I'll do with it then, I don't know. Probably Abby will make those arrangements, or not, as she sees fit. She can see the big picture far better than I can. And she can make things happen--not for my own personal enjoyment, but for the work.

Best regards,

Stephen Sakellarios, M.S.