GitHub Discover
github.com
github.com
To me it feels like they sacrificed functionality for aesthetics. It's not the right trade-off to make for a professional tool. I don't want my Github feed to be Facebook-like. I'd rather have it be compact and useful. A feature that'd let me browse by "topic" (e.g systems programming) regardless of language would be welcome as well.
In this update, the three culprits are the font-size, the spacing (so much space wasted) and the (slow) auto-load. It wasn't great before but this is most definitely worse.
Dear Github, please make technology discovery efficient rather than pretty. I spend hundreds of hours on your platform. You could save me a lot of time focusing on the right stuff.
It's a shame because there's so much potential there.
What you're talking about already exists. In fact, they're called topics, e.g.: https://github.com/topics/systems-engineering
No one or few people self-reference their repo with something as broad as "systems engineering" or "systems programming".
The closest thing that exist on Github is the explore tab which lets you browse random repos by categories (e.g open government). Such feature would require Github to do the appropriate groundwork to become great at technology-discovery, namely: interest maps, automatically infer a repo's category based on its content, description and the profiles that star or fork it, for starters. They can even make their search engine relevant and pleasant to use if they feel like treating us.
And I'd want that to be integrated in my feed. Right now, I follow around 150 active users and it is still pretty sparse: it's not uncommon for me to see "X started Y 2 days ago". It would be good to add more (good) signal.
I want to see more of what better programmers are starring.
> A feature that'd let me browse by "topic" (e.g systems programming) regardless of language would be welcome as well.
Topics are exactly what you initially described. Perhaps you didn't do a very good job describing what you'd like.
Also, they're called topics. Not tags. It's in the documentation.
> Such feature would require Github to do the appropriate groundwork to become great at technology-discovery, namely
They did a bunch of work to automatically add topics for many repositories to solve for this: https://githubengineering.com/topics/
I tell you it only does so nominally. I didn't fully develop its specs but you could have inferred that they weren't matching the existing one's. The topic feature is half-baked, poorly thought-out and integrated.
What team have you worked on where that is literally ever the case?
Look at how information dense Excel can be with a typical spreadsheet, and the vast majority of users have no problem with it - yet when it comes to a basic page listing a few items, there's 50px of padding around each item in 12px font. Suggestions: remove the padding, group actions by the same person or same repo to remove repeated clutter, add descriptions of the repos in the front page so it doesnt require a click to see more, etc...
I'm not disagreeing with you, but is this perhaps the reason why this is happening? If we have so much more space, we've got room for extra big bars. If we get higher resolutions, smaller fonts are (possibly) harder to read, so we make them bigger.
Maybe I'm just making stuff up, though.
Am I just a curmudgeon, or does every step that GitHub takes these days seem like wrong one? Everything, from small design changes, to the results of their discovery engines, to the pricing and offering of there paid services, just seems to get worse over time.
As somebody who visits GitHub more than any other page, it makes me a bit said. I do hope they can turn it around or GitLab will come up on them fast.
I feel like discovery engines making obvious recommendations is actually a non-trivial problem to solve, I remember Spotify used to recommend me a lot of 'well duh' stuff in Discover Weekly. When just going off raw data, you can't really just 'not recommend' popular things. I routinely tell fairly aware co-workers about JS projects that have thousands of stars on Github that they've never heard of.
(Also I'm kinda cynical of anyone 'not liking pricing', of course you don't, pricing models are designed to make you need to pay).
Unfortunately some of the slots are wasted on ever-popular projects which are always on the list, particularly in some of the more esoteric languages. It is also interesting to see how few stars are required to hit the language-specific top 25, especially on weekends/holidays.
No matter how awesome GitHub Discover's predictions someday grow to be, it's always useful to monitor the collective wisdom of the crowds.
csurfer/gitsuggest
A tool to suggest github repositories based on the
repositories you have shown interest in.
It looks like it requires a local copy of the original Roller Coaster Tycoon 2 to work. Why is that?
https://openrct2.website/getting-started/index.html
"OpenRCT2 needs the object files (containing graphics, sounds and models) from the original RollerCoaster Tycoon 2 to work. You'll need to have a copy of RCT2 before beginning."
There is a logged-out feature called 'Explore' [1] whose prominence on the GitHub frontpage has fluctuated with time; now it's prominent again. Both of these features showcase projects; one is curated by the company while the new 'Discover' is curated by an algorithm based on your activity. They appear to exist side-by-side without any linkage or references to the other.
I have a lot of JS blogs/twitters/subscriptions that I use to keep up-to-date, and generally most of the articles I read will appear in the first couple of pages of the trending repositories.
- The feed design is poor, I don't like the alway black bold font for every repo name, but they keep blue font for branch name.
Reason: I know 80-90% of recommended projects.
I know Github has some sort of /explore link when you were logged out, but it's so out of sight that I only remember it when I accidentally visit Github on my incognito tab every few months.