I'm curious whether anyone has run into hallucinations with this kind of use of an LLM.
They are pretty great at converting data between formats, but I always worry there's a small chance it changes the actual data in the output in some small but misleading way.
I've used OpenSCAD for a few small projects. The main draw for me is avoiding all the "fiddly mouse stuff". I figure it's a trade-off between simple stuff being more complicated, and complicated stuff being easier to bulk edit and organize.
I don't know enough about the underlying proof-of-work stuff to comment on how effective this could be, but I think it's pretty funny that the UI examples say "I'm a human".
I guess "there's only a few of me at most" or "I could allocate enough computation to this that I'm probably not up to no good" don't read as clearly.
I'm probably just not thinking through the problem fully, but:
Wouldn't this be better solved inside the coding assistants? They're the parent process and should be able to tell when a sub command has hung.
This is incredible, love how much further you've taken it than just the proof of concept as well.
This is probably just my opinion, but I kept thinking that a better word for "drive" given the domain would be "clock". It's basically a binary clock signal driving these, right? Maybe "clock drive"?
I opened this on a mobile device, and it only shows the top and bottom topic buttons. I was pretty confused until I tried turning on "desktop site" in Chrome and saw the left/right buttons.
Having only a linear track of topics kind of undermines the overall idea!
> here are the empirical distribution functions (ECDFs) with 30ms added to each response time
> The added constant seems artificial, but it's just viewing the results from the point of view of a client with 30ms ping time. Otherwise the log scaled x-axis would overemphasize the importance of a few milliseconds at the low end.
I thought this was interesting - maybe it's a standard practice I was just unaware of but it seems like a smart trick.
The tone is entertaining, but some of the snarkiness around the code is bit frustrating at times. e.g.
> # do i really need to say I'm not serious about this?
Can you just come out and say what you mean here? Like, presumably it's bad, insecure code, but can't you just spell it out for the benefit of your audience?
I'm probably being really picky here - just find this kind of developer in-crowd signalling to be really irritating and counterproductive.
I once inherited a codebase that did this, and it was a nightmare to learn. All the classes were named after types of alcohol for some reason.
Naming things clearly is really hard, but this approach is just giving up completely.
Interesting that most of the top comments here seem to be about the day to day process of estimation itself, when the linked article is about how poorly the research on estimation has been done to date. Did any of you actually read the article?