[conspire] Discussion: Using LLMs the Right Way: 10/1/2025 7pm Eastern Daylight time

[Deirdre Saoirse Moen]

> On Oct 3, 2025, at 16:56, Don Marti <dmarti at zgp.org> wrote:
> 
>> That said, I did find it useful to be able to interactively ask questions that were specific to $FIRM’s processes and have a live answer that was usually at least close enough to correct I could find the answer from where it led me.
> Now I'm wondering about the categories of things for which an LLM would be useful.  Syntax for doing something that appears in publicly available scripts might be a good example -- I have been able to "vibe code" some correct Python code to extract material from an HTML file with Beautiful Soup, and get the jq(1) syntax right for getting a subset of the info from a big JSON file. (For both of those I could test the output, though)

I got a good idea from my friend Jack, who’s a sysadmin for a furniture company, and almost literally the only tech guy in town (he lives in an area of rural NZ in a town of about 5,000). He has need to write code, but he’s not a dev, and so he needs to have another set of eyes.

His most successful strategy has been to ask an LLM to explain his code to him.

I hear others (who don’t like writing tests) suggest it’d be great for writing tests, but that would be the absolute last place I’d use an LLM. A far safer strategy would be to write the tests and have the LLM write the code that makes the tests pass. (Assuming you were working on a well-enough trodden path that an LLM had sufficient code examples to do this, which is not the kind of work I do.)

Deirdre