Brian Lopez

Professionals don't use FFmpeg and FFprobe. They use extremely expensive software that runs FFmpeg inside (and that vendor contributes nothing back)

@vmg Excited to see you inspired again!

Cursor 1.2 is out! This is the first release with major contributions by yours truly (you can't tell they're there -- they're all on the server side, hehe). Do check it out, it's chock-full of new features!

@svscnne @holman And what do you think would happen if you gave 100 million people simultaneous access to the same Google Doc?

@holman @brianmario Google Docs has more users (and versioning features) than GitHub at this point

On Friday last week, GitHub had a significant, multi-hour outage. This sucks, but it happens. I can be sympathetic to the struggle of shipping at that scale. Today, GitHub is having ANOTHER multi hour outage. I’m out of sympathy. It’s time to distrust GitHub.

Public facing? No, not even close. Google could line those bits up to the end of the universe lol. I was referring to the *internal* traffic for scaling their source control stuff. They don't have anything public facing that's in any way similar to GitHub, so I thought that's what you meant.

@brianmario @holman You don’t honestly believe GitHub deals with more scale than Amazon and Google right?

@svscnne @holman We're talking about differences in scale in multiple orders of magnitude here. I think the reason we're not changing each other's minds is because we're not talking about the same thing.

so are lines of text… I don’t think either of us changing our minds here. But just consider that even if it’s not “public”, Google and Amazon have xx,xxx engineers that all require authentication and different level of segmented access control. Amazon is not exactly known for being a homogeneous org. Their use cases aren’t magically different than if its users were external customers instead of employees.

I'm not sure how I let myself nerd-sniped into this, but here we are :) I think you may have misunderstood what I meant by scale - it's not the size of the codebase that runs the system. It's the scale of the system itself in order to provide the experience nearly every developer on earth relies on. Including Google, Amazon, Meta, Apple, etc. The size of Google's codebase has nothing to do with what GitHub the product and system needs scale. Which also has nothing to the size of GitHub's own codebase. I'd bet Google and Amazon's monthly use of their own internal source control systems pale in comparison to what goes through GitHub in even one day. And even at that, the requirements are completely different. Besides, most companies their size use GitHub Enterprise internally. Back in July of 2024 Amazon effectively deprecated CodeCommit. Google shut down Google Code (which we all loved) back in March of 2015. They even helped users migrate to GitHub. And aren't Sapling and Glean client-side apps? Also, the code for Sapling and Glean are hosted on GitHub. Bitbucket and GitLab are probably the closest comparison in terms of user base and what needs to be scaled from a technological perspective. Even as of Jan 2023, GitHub had over twice as many users as both of them combined today. GitLab even uses a ton of GitHub code (open source and some eerily similar to internal code). It would appear that they needed to copy GitHub in features *and* scaling techniques. Scaling the internal use of source control is not the same thing as scaling source control operations that happen with nearly every request (api, browser and git itself) on the #37 website in the world by traffic. That all said, git itself doesn't require GitHub at all. You or anyone else are welcome to use a different hosted solution, or even make your own! You'd find out for pretty quickly how easy it would be to scale to GitHub's size. For fun, try recreating the Pull Request page. Clone a fairly popular repo like rails/rails and use that for the git data in the UI. Experiment to find out how much diff content you should load and how much you should defer. Don't forget about comments that need to be rendered inline, with markdown support, automatic links to other issues/pulls, etc. Once you've done that, think about how you'd do it with random diffs (which could be any size, up to gigs) from 420 million repos and then hitting it with 2.8 billion requests in 24 hours. All while 1/4 of those repos are also receiving constant updates via git push. If that sounds easy, then maybe you should be applying for a job at GitHub ;)

@svscnne @brianmario insanely cacheable.

I didn't say it's so overly complex overall, just parts of it arguably were (when I was there anyway). The scale is mainly the point I'm making. It's almost unimaginable how much structured data is moving through the system *per-second*. Most of the hard part is with git itself. Short of building a reverse index of every single object in every single repo - effectively doubling (if not more) your storage requirements, you just have to do the work to traverse a repo to find the content you need. It takes time (CPU+disk) and there's almost no way around it that makes sense cost-wise. In terms of raw scale, I'm talking about even a single allocation in the wrong place having a huge impact. It's the stuff nightmares are made of (I can still hear the Pager Duty alarm sound in my head at night). The tooling for rolling out changes was best in class though. We were able to confidently make substantial changes to multiple systems with no user impact whatsoever. Thousands of times. Of course, nobody hears about those because things went as planned. So when you see downtime, just know that there were countless other changes rolled out that you didn't notice because, well, frankly they're doing an amazing job at scaling one of the largest websites on the internet. What product(s) do Google, Amazon and/or Facebook offer with regard to distributed source control that works like GitHub? My understanding is that essentially 99.xx% of all developers *in the world* use GitHub as their source of truth.

@brianmario @theo Uhm. Distributed source control is practically a solved problem at this point. Amazon does it fine. Google does it fine. Facebook does it fine. If GitHub system is overly complex to the point where it’s unscalable, then challenge unnecessary complexity.

@ImpulseLabs_ Hi! How much does the cooktop weigh? I can't seem to find that anywhere...

@GRAndroid22 Hey! That's awesome, wish I could have been there! He tells me you're a programmer?

@brianmario Hey, having a drink in Tallahassee with Hunter Criazzo. Looking to network and get more information. Thanks!

In case anyone was wondering why I haven't been keeping up with my open source projects ;) winebusiness.com/news/people/ar…

I'm stoked to showcase our implementation of vector search for PlanetScale. This is not a HNSW plug-in for MySQL (like pgvector, etc), it's a full fledged index managed by InnoDB that scales into terabytes of vector data. planetscale.com/blog/announcin…

@mitchellh I'm sure he's busy, but just in case, @jasonlong is one of the best in the biz.

I'm looking for someone to help me design and build the Ghostty website. If you know a frontend person or team who understands and loves the terminal experience, please email me. Ideally a full stack thinker who can help guide the automated doc process. Paid, of course.

To this day I have two Apple IDs because of the shit show this created. Back in the day there was only the iTunes Store (media and apps purchases). But when MobileMe launched you had to create a new ID for it. Then iCloud launched and they "merged" the functionality of both account types without letting users merge their two accounts they now had. And consistently since, I've felt nothing but punishment for being a long time Apple user. When iTunes Match came out, it trashed most of the irreplaceable music from my library. Stuff from my various bands I had in there for ages, are now silent tracks of the same length. Similar things happened with iCloud Photo (or was it Photo Stream or whatever?). Almost all of my old videos won't play (think pre-H.264 stuff) and are seemingly somehow corrupted. After that I learned my lesson and try to wait to opt in to newly shipped features like that - if it's even a choice. Sometimes it's not. I could go on and on about long-standing annoying bugs that I *know* Apple staff hit too. Honestly, I feel like they need a "Steve" in there to yell at folks again or something. Someone who holds quality above all else (though tbf their hardware is pretty incredible). If for nothing else, to match what they're selling as a brand. All that said, unfortunately there is no alternative for me at the moment. Especially given how deep I am into their walled garden. Which was no accident by them.

Good night sweet prince. Retired 12 years ago when iCloud launched. I still can’t believe Apple named their web service ‘mobileme’.

@sdamico @ffensivebias @ImpulseLabs_ No idea if he'd be available (or down) but chef @therealweissman would be amazing to hire for trying it out. Pretty sure he uses a Duxtop for some things currently...

@ffensivebias @ImpulseLabs_ We did that — next step is get a pro in

Iteration ~three (next time trying with an even smaller pan). Incredibly easy on @ImpulseLabs_

@janaboruta @gthumb7 @t3dotgg @charles_irl @lizziepika @leahculver @klaralosert @mies @Advait_Ruia @CompuIves @fouadmatin @diptanu @amanda_robs Done.

@gthumb7 @t3dotgg @charles_irl @lizziepika @leahculver @klaralosert @mies @Advait_Ruia @CompuIves @fouadmatin @diptanu @amanda_robs @brianmario Tasting Lopai Wine on the deck while the sun is setting sounds like heaven to me. This feels like a Saturday afternoon in April type activity to me.

I made a site for Founders on the Deck. Wish I had captured everyone that’s stopped by Café Jana but now I know. Thanks for coming by @t3dotgg @charles_irl @lizziepika @leahculver @klaralosert @mies @Advait_Ruia @CompuIves @fouadmatin @diptanu @amanda_robs foundersonthedeck.com

@sdw @halidecamera This photo is insanely good

San Francisco. The most beautiful place to live.

@chacon @gitbutler I can hear @SethMacFarlane’s voice coming out of this

At @gitbutler we have a running “dad joke” culture and we found an emoji for it that makes me lol every time.

@jonrohan I love Caddy

Looking to play with some new code. What’s your favorite dead simple light weight web server, preferably with hot reloading? I’m not deploying this anywhere just want to run a server locally. Any coding language.