Posts Tagged "llm"

3 posts ยท See All Tags

A successful first run for Slartibartfast

Last night's first overnight Slartibartfast run was a success. I marked 4 issues as acceptable for it to work on, and it produced a draft pull request for each.

Tonight I whipped up a quick status monitor as an admin page on this website, so I can peek at what it's doing.

The costs shown below are not what I'm actually paying. To my knowledge, there's no straightforward way to understand how many tokens you have access to within each 5-hour Claude subscription window. Slartibartfast maintains statistics on how much work it can perform in 5 hours, then estimates both that window size and the cost per worker process, so I (and it) can eyeball how much work it can do in the remaining time. The actual stats are measured in tokens, but since those are large numbers, everything's converted to an approximate dollar cost for the end user.

Continue reading

Slartibartfast: hands-off coding in the wee hours

I've spent the last few nights rolling myself a "software factory", a system for using an unattended coding agent to autonomously pick and work on issues from my GitHub repositories.

The system's name is Slartibartfast, after the hapless designer of fjords in Hitchhiker's Guide to the Galaxy. For short, I call it "Bart".

The primary Bart process is a good old-fashioned program, not an AI. It runs on the same server that hosts my 3 live websites (Zingor; Unsphere; and this blog).

On startup, Bart builds a list of all open issues from my GitHub repositories which satisfy the following criteria:

Continue reading

In reply to https://gist.github.com/MatthewDailey/118ecbf098d4a1f9b31829dbf43691c4

Hi, MatthewDailey. Couldn't find your email so I'm contacting you here.

I don't recall how I found you and your projects -- probably HackerNews -- but I was intrigued by your GPT DND. You aren't the first to do "AI as Dungeon Master", but you are the first I've seen to add voice to the equation, so kudos.

I thought you might be interested in my gm-trainer project, which comes at things from the other direction: using LLMs as pseudo-players, rather than the GM, to provide GMs with a sort of training sandbox. Still early days, and not as configurable as I'd like it to be for other people. But to my knowledge I'm the first to flip things around like that. (Which makes sense if we consider that most people working in the nascent "AI + RPG" space seem to want to replace the "scarce resource" that is GMs/DMs).

Continue reading