In England, after the Norman Conquest, most of the defeated English became the property of the Norman Lord who was granted the land by the King. They were called serfs, and were bound to the land. If the land was sold, they went with it. You could also become a serf if you owed money… Continue reading Sixteen Tons
A Really Old Tip
I just hit an error message that just said "You did <something>" and the process halted. That was not helpful Back in the v6/v7 days, there was a widely known but unofficial way to write error messages: The name of the program (ie, $0) ":" What specifically went wrong "," What to do about it… Continue reading A Really Old Tip
How to make LLMs Lie
I knew that, in Canada and the US, the wealthy pay less taxes than the poor. So I asked Claude if that was so. I asked In the US, what percentage of the population make 50% of the income? Do both pay 50% of the taxes? It replied: I'll search for current information on US… Continue reading How to make LLMs Lie
Hormuz and Tariffs
This is a little experiment in using an LLM to validate or invalidate motivations. It is widely reported that Mr Trump and his associates made money in the stock exchange, through advance knowledge of the war winding down or heating up. That's unproven: only one person has been charged, and he hasn't come to trial.… Continue reading Hormuz and Tariffs
Stop Trying so Hard!
LLMs used more than once to hunt for problems will try really really hard to find problems, even if they're gone. On the second or third iterations they will start hallucinating problems. What I Did I gave Claude instructions about how to do a developmental edit Claude did so, but each time I finished fixing… Continue reading Stop Trying so Hard!
Recursion
In computer science, Morvin’s Meta-theorem applies: “Any problem can be solved with one more level of recursion.” Quite by accident, I did this for LLM prompting. I asked Claude to tell me how to do a “development edit”, because I wanted to know how to look at one of my essays, to see if it… Continue reading Recursion
Hedgewitch Part 8: Coda
AI advancement has had roughly four eras. They’re about 20 years apart, last a few years, but have an “AI Winter” in between If we want credible attempts to create AGI … then we need at least the next, due around 2065. Maybe 2095. Ignore the Hype Start learning to use it now. Start with… Continue reading Hedgewitch Part 8: Coda
Hedgewitch Part 7: Oh, and it’s in a Bubble
If they’re big enough, every fad turns into a bubble. Actually every fad does, but the little ones are hard to see. Currently the Fad is “Use LLMs for Everything” It starts with a discovery or invention, and zooms to a peak that it can’t possibly deliver on. Then it fails to do something it… Continue reading Hedgewitch Part 7: Oh, and it’s in a Bubble
Hedgewitch Part 6: What would Dave like me to Say?
LLMs constantly make errors. Why? Answering the wrong question is the first underlying problem. Phrased in polite Canadian, an LLM always answers the question “what would a reply to this look like?” In less polite language, it asks “What would Dave like to hear?” The US and business term? Suckup. If all your subordinates suck… Continue reading Hedgewitch Part 6: What would Dave like me to Say?
Hedgewitch Part 5: Upsy-Daisy!
Using it for what it’s good at was an obvious approach. Turning it upside-down was good. What About a Little Kid? Treating it like a little kid and walking it through something, like a math problem, also looks promising. I wanted to draw a particular mathematical figure. So I told the three-year old to 1.… Continue reading Hedgewitch Part 5: Upsy-Daisy!
This Blog has No Leaves 
