Minor release of my uploader, with some essential upgrades, unit tests, and a complete redesign.
Still just a minor release because I will be adding wikis and long-form article processing, as well as folder-based 30040 structures, and some other stuff, and that'll be the 0.2 release. Everything I'll need to produce test data for #Alexandria's Gutenberg edition.
Inching my way there.
GN
Don't forget that Alexandria is aiming to cover: wikis, long-form, publications, highlights, and kind 1111. We still have a very long way to go!
We will not be using kind 01, except perhaps as part of the communities integration in the Defoe edition.
Tried something new. It didn't make me much faster, but the end result is still better, so I'll keep doing it.
i've tried AI things several times and been completely underwhelmed by its value compared to me just spending time on something
As someone more specialized on test and UX, it saves me a lot of stackoverflow searching, which is nice, but I really have to sort of inch forward and constantly test because it occasionally breaks stuff or generates wrong solutions. The more time I spend writing the scenarios and tests, beforehand, the better the coding part is. The resulting code is better than what I usually write, but there's lots of room for patterns, tighter algorithms, and other architectural improvements.
Someone specialized on coding, probably feels that way about tests.
TL;DR -- It can leverage your current talent stack and even out your weaknesses, but a specialist would get better results. Since nobody is a specialist in everything, it's good that we at #GitCitadel can cross-pollinate the same codebase. Leverage upon leverage.
I'm now more convinced than ever, that this is going to have a bigger impact in making programmers faster and better, than in making them obsolete.
Also, it shows that diverse-talent teams are a total flex. Definitely will get better results, for enterprise-level systems, than just one dev and an AI. 😎
But, we knew that.
yeah, technology is always about efficiency
but honestly, the LLM leaves a lot to be desired for programming, it does not have enough depth to really write correct code
probably the clowns at microsoft and google think if they give it more space and feed it more data and throw more GPUs at it, that it will somehow fix this but honestly i doubt they don't already have the alexandria of code already, so it's really a case of diminishing returns
my bet is that eventually the consensus will be that for less experienced coders it makes them 10% more efficient, which is a LOT LESS than the hype makes out
My code base is already beyond the reach of anyone inexperienced, I think. And I'm not a dev, just a programmer.
Large, elegant, systems wouldn't be possible, but those are the ones that are the most-fun to use.
I'm paying for a special coding AI, as well. The quality differences in the offerings is gigantic.
it's my opinion that the main thing that AI will do for more serious programmers is allow them to test and iterate faster, which means they will advance a little faster
refactoring work is the main reason why i have such a fancy keyboard, it takes a LOTTA typing
if i could trust an AI to do this work for me it would speed up my testing/refactorings by enormous amounts, they are the most time consuming parts of the work
Yeah, I'd been procrastinating the refactor, but it just being "supervised typing" took a lot of the tedium out of it.
Do you have it review your code and write tests? It found some copy-paste errors I made, that were ded code (so, running tests couldn't find them, but they made the codebase slightly confusing).
And the tests let you freeze the state, refactor, and then check to make sure you didn't alter the state.
