[TUHS] A thing I did (AI Slop Alert)

[Luther Johnson via TUHS]

For those interested in knowing more about Forth (I think that's where 
this thread began), the FIG-Forth (Forth interest group) implementation 
of Forth has a very efficient, indirectly-threaded inner interpreter, 
which uses the PDP-11 instruction set well. A Forth implementation is 
not that hard, at its core, but you do want to write the most basic word 
implementations in assembly language. Once you have the model in mind, 
it's a matter of filling in the blanks. Claude and other AI tools don't 
really know the history or the techniques, they do deliver working code, 
but it's big and inefficient in comparison to well-crafted implementations.

My two cents, limiting myself to practically helpful advice if anyone 
wants to dive into how Forth implementations have been done in the past.

www.stackosaurus.com/figforth.html

On 08/11/2026 08:05 AM, Chet Ramey via TUHS wrote:
> On 8/11/26 10:54 AM, Katie via TUHS wrote:
>>> I think Alan believes that LLMs like Claude are going to be the best 
>>> tool
>>> for some jobs; one more tool in the bag. Not an obligation, but a
>>> prediction.
>>
>> If we're going to use LLMs, why not llama.cpp?
>
> If you think that will work better for you, why not?
>
>> I miss pagers.
>
> Ok. I do not.
>