TSE: AI: SAL: Rules

[knud van eeden]

Hello,

FYI

See attachment: tsesalrules.txt

1. Given is that those AI do not retain a history in general, so you have to repeat the below steps each time (in general, that might improve in time also).

2. Note: Some AI demand that you re-use your old chat to be able to have some history, I think Claude AI has this property.

3. These are currently the TSE SAL rules (see attachment) which I first paste into the first step

of each AI (e.g. DeepSeek, Claude, ChatGPT, ...) before letting it create my TSE programs.

4. After pasting those rules, the AI interprets that information

5. And after that I create my prompts of what exactly want in TSE SAL.

6. Then I get a TSE SAL program back from the AI

7. That SAL program I run in TSE

8. Any errors from running that TSE SAL program or my additional demands or changes I feed then back into the AI in the step. It is very good in interpreting that feedback text.

9. That works pretty good, though sometimes many iterations like these are necessary to get the working TSE SAL program that is wanted and or you hit the limits as the maximum output per iteration is about 8K.

10. By typing 'continue' in the next step you might be able to get more than 8K output after that.

11. I switch also between AI (DeepSeek, Claude, ChatGPT, Grok, Gemini, Perplexity, ...) as some AI give better

answers for some questions, but DeepSeek is pretty good for TSE SAL.

with friendly greetings

Knud van Eeden

[S.E. Mitchell]

Maybe I should dig up my EBNF grammar for SAL.
I saw an example where a guy gave the AI a grammar, and from that the
AI was able to do a good bit.
But not sure how one would specify semantics. Will have to leave that
to English I guess.

> --
>
> ---
> You received this message because you are subscribed to the Google Groups "SemWare TSE Pro text editor" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to semware+u...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/semware/210839018.1427414.1737152985494%40mail.yahoo.com.

[knud van eeden]

> Maybe I should dig up my EBNF grammar for SAL.

Yes, have a look for it, otherwise I might be able to dig up more,

I have a web page for it:

https://www.knudvaneeden.com/links/language/grammar/parser/backus/naur/form/syntax/library/bnf/type/ebnf/language/computer/tse/sal/syntax/file/source/ebnflanguagecomputertsesalsyntaxebnf/01/01.htm

https://www.knudvaneeden.com/links/language/grammar/parser/backus/naur/form/syntax/library/bnf/type/ebnf/language/computer/tse/sal/syntax/file/source/ebnflanguagecomputertsesalsyntaxebnf/01/SyntaxRailRoadDiagramTse.pdf

> I saw an example where a guy gave the AI a grammar, and from that the
AI was able to do a good bit.

Probably best to do it in smaller steps.

instead of the whole EBNF grammar at once take only 1 line of it, when that works

next line and so on.

---

> But not sure how one would specify semantics. Will have to leave that
to English I guess.

Yes, typically English.

Those AI are just incredible in interpreting the English you write, really phenomenal how they can extract

from your terse information exactly that what you want it to do.

To view this discussion visit https://groups.google.com/d/msgid/semware/CA%2BgEvka%3DV%2BsWHU_sNcB%2BXn_6a5J%3DTQPb18b2pU9XVYz%2BpTaGsw%40mail.gmail.com.