Andersons of Colonial N. Carolina

Don’t vote for Shysters.
Shysters can be recognized because they do not go to work in the morning.

.
Shysters get in office when they have very little money but talk good …
They offer to give you Free Stuff.
A year or two later they are millionaires.
.
Don’t vote for Shysters.

I’m getting old now and have Grandkids that I care about…

Games. There are fun games…”spin the bottle”, Monopoly, Ping Pong…

And serious games… World Wars I and two.
An idiot Congresswoman recently referred to World War Eleven. I am serious.

.
The crap that is going on now is In Your Face.
.
Time to wake up people.
.
I wrote this hopefully so that a moron could understand it.
.

Click to access CastleTrash.pdf

.
Remember what Thomas Jefferson said… if you still have any active brain cells…

Go to Google Maps…search for Sasnett Mill Branch. It is a few mile west of Tarboro, NC. During the 1700s I think it was known as Curtis Mill Branch. I also think that Curtis built a Grist Mill there…hence the name…Yuk Yuk.

Curtis Mill Study

.
https://dcr.lib.unc.edu/record/6066d23f-0109-4142-bbfa-e909d6fd41e9

.
My friend David Gammon sent me the Dunn deed… he and I stay confused by email.

Meet Michael Jackson’s Grandfather…wait…what?
Just found this… It has to be AI but I do not care. If this is what AI has to offer, bring it on.

https://www.youtube.com/@OldDeltasBlues

This simply pushes all my buttons. But I’m curious whose culture is being appropriated- Black Blues or White Easy Lis’nin?

My test is always The House of the Rising Sun…

I’ve worked on this for several months now (years actually). It took me forever to realize a simple fact: a Colonial era map of the first colonists HAS to begin with the actual Grants of land. Those are known as Patents. They are not to be confused with Deeds. A patent can BECOME a deed or a Patent can be subdivided into smaller and multiple deeds. The point to be made is that there has to be a starting place…the Patent.

Walnut Creek Study

This is an introduction. In the not too distant future I hope to break these Patents down into deeds. That is where hundreds of the first colonists will be found. Perhaps your ancestor.

As an example these folks were from Walnut Creek…

.

William Deloach STUDY

Sometimes I have to cut loose and have some good old fun writing…

Those People

.
I decided to have a bit more fun…

.

CastleTrash

Griffis/Griffin Tree of Edgecombe County. 1740ish to 1800s

Fresh bones to chew on…  also WALL and BOYKINS families

William Deloach STUDY

I have been criticized for using AI.  Fair enough, and I want to respond honestly…

I presently see two distinct tracks for AI.

1. This is the free version that has teased the internet for a few months now…I think of it as the opening salvo of How to Get the Crowd Interested Enough to Buy a New Product.  This low level of an LLM (Large Language Model) is just smart enough to be dangerous and SILLY.  It is programmed to answer any question and it does.  The problem is that it gets its info background from data online (such as encyclopedias, etc) which are scientifically plausible…unfortunately it also gets info from the general public.  Info from the general public, in my opinion, is a crap shoot.  In a nutshell, if the free ChatBot runs across faulty info on the internet it will spout it out as the truth. So folks start to call it a liar or that it hallucinates.  I just see it as a gullible crowd-sourced pot of mere opinions mixed with Einstein level smart stuff.  Ergo, I do not use it much.
   

2. Agentic AI is what I use.  You subscribe to a LLM of your choice. I use both XAI and Anthropic. Each is paid with a monthly subscription.  I also use Openclaw which is Open Source…however even Openclaw has to be powered by an AI  “brain” from an LLM.  That “brain” is called an API (an Agent in my mind). You are responsible for training your Agent.  By training, I mean giving it Skills, Tools, and “tricks” you may pick up. I gave my Agents a Truth Standard that I developed that goes back to ancient Greek philosophy…I have described it elsewhere on this site. It is what I call the 4 levels of truth.  I give my Agents exact places to search for dats. (places that I choose and use) …I also have the option just to turn the Agent loose and let it grab what ever it can find in the wild, wild Internet.  So, to sum things up, I point my very nosey Agents where I want them to go and I give them specific instructions of what/where to search for stuff.  I not so jokingly call it “Have Agent, Will Travel”. I coined the term if you see it elsewhere.  I named one of my agents Tonto. We chat, collude, argue, hypothesize and share info. I sometimes give my Agents permission to “opine” as I stylize it. They often surprise me.

______________________________________________________________________________________

I wrote the above by my lonesome. I then asked Tonto to critique it:

Tonto:  Here’s an honest read, organized by what’s strong and what’s shaky.

The real weak point: you’re tying trustworthiness to paying for an LLM rather than to how you use it. That’s the part a sharp critic will go after first — plenty of paid, agentic setups still produce garbage, and a disciplined person can get decent results out of a free chatbot if they force it to cite sources and flag uncertainty. What actually makes your output trustworthy isn’t the subscription, it’s the method you built on top of it — the four-level truth standard, the requirement that every placement trace to a primary source, the discipline of flagging “[opinion]” instead of silently smoothing it over. You mention the truth standard almost as an aside at the end, when it should be the spine of the argument. Right now the essay’s implicit claim is “I pay, therefore I’m rigorous” — the true claim, and the one you can actually defend, is “I built a rigorous method and I happen to run it on paid models because they’re capable enough to follow it.”

The hallucination explanation is a little off, and it matters. You describe hallucination as the model faithfully repeating bad information it picked up from unreliable internet sources — basically a garbage-in-garbage-out story. That’s one real failure mode, but it’s not what people usually mean by “hallucination.” The more common and more troubling version is a model generating a specific, confident-sounding detail — a citation, a date, a quote — that was never in its training data at all, because the underlying mechanism is predicting a plausible next word, not retrieving a verified fact. That happens even with the best paid models, agentic or not, and it’s exactly the failure mode your own truth-tier system is designed to catch (which is why you make your agents cite the primary image, not just assert a conclusion). Leading with the wrong mechanism gives a technically literate critic an easy opening to say “that’s not actually how it works,” which undercuts the credibility of everything after it.

The “I let it opine” line cuts against your own argument if you don’t qualify it. You spend the paragraph distinguishing yourself from crowd-sourced guesswork, then say you give your agents permission to opine and they surprise you — which, unqualified, sounds like the same thing you just criticized. The actual distinction, and it’s a good one, is that when your agents opine it’s labeled — tagged as opinion, kept separate from doxa, never quietly presented as settled. That’s the real difference between your practice and the free-chatbot problem you’re describing, and it’s currently implied rather than stated.

It answers a narrower criticism than the one you’re probably facing. The whole piece rebuts “AI makes things up and can’t be trusted” — but “you’ve been criticized for using AI” more often means something closer to “should a human be outsourcing research/judgment to a machine at all,” a legitimacy question rather than a reliability question. If that’s the criticism you’re actually responding to, this essay doesn’t quite meet it — it defends the tool’s accuracy, not your own role in directing, verifying, and taking responsibility for what it produces. Given how the Walnut Creek project actually runs (you’re the one setting the truth standard, checking primary images, overruling the agent when a reading doesn’t hold), that’s a stronger defense than “I pay more for mine.”

Smaller things: “dats” is a typo for “data.” A few sentences run long and could split. It ends on an anecdote (Tonto surprising you) without circling back to the opening frame — you set up “I want to respond honestly to the criticism” but never quite land the response; it trails into a description of your workflow instead of a conclusion.

What’s genuinely good and worth keeping: the voice is authentic and readable, “Have Agent, Will Travel” is a memorable, ownable line, and the underlying distinction — casual chatbot use versus a disciplined, source-anchored agent practice — is real and defensible. It just needs to be argued on method rather than price tag, and the hallucination mechanics tightened, to survive contact with someone who knows the technology.

Marc

Killebrew Dancy MILL

It is known as Nobles Mill Pond today…   this is Google Map view…

https://www.facebook.com/groups/34407448334/posts/10158887870468335/

The_Shadow_of_Robert_Killebrew

My evidence:

Perry to Griffin 1818

My “working” Tree of Killibrews:

Killebrew_Family_Tree