Let's Open Source the World — Brandon Thomas, ArtCraft — Transcript
Full transcript
- 0:01[music]
- 0:12>> Hey everybody. Thanks for attending my
- 0:14talk. I had to put this together late
- 0:16last night. I've been in a whirlwind
- 0:18after this project went viral of
- 0:20unblocking thousands of open source
- 0:22developers. We've had VC's approaching
- 0:25us. I've been in constant meetings. Like
- 0:28this has been
- 0:29a sleepless month of activity since this
- 0:32happened. So if you're not familiar,
- 0:34I kind of cloned Adobe Photoshop and a
- 0:37bunch of other apps and that went super
- 0:39viral.
- 0:40My phone, my inbox, everything has just
- 0:43been absolutely lit up and so I I I
- 0:46apologize for the structure of the the
- 0:48the the talk. This was thrown together
- 0:50as quickly as I can, but I've got a lot
- 0:51of really exciting stuff to talk about
- 0:54and I think it talks about the future of
- 0:57not only the industry, but just
- 0:59technology as a whole. So
- 1:02we're going to open source the entire
- 1:03world. Like why not? It's it's something
- 1:06that I've always wanted to do as a
- 1:08developer. That's something that
- 1:09a lot of us in the open source community
- 1:11are interested in. Open source powers
- 1:15a lot of the you know, technology that
- 1:18we're familiar with, but it's never been
- 1:19at the forefront. Like Linux has always
- 1:21been like a second class citizen. We've
- 1:23been pulled into a whole bunch of
- 1:26platforms where we don't really have
- 1:27ownership and AI kind of gives us a
- 1:30chance to take the web back, if you
- 1:33will. So
- 1:36I might have to do this manually.
- 1:39Cool. So a little bit about myself. I've
- 1:41been an engineer for about 15 years. I
- 1:43spent eight years at Square doing heavy
- 1:47hitting infra. So online payments,
- 1:50the the authentication stack, basically
- 1:52anything that could take down the
- 1:54company my team was kind of in charge of
- 1:56and it was you know, eight years of
- 1:57carrying the pager, a lot of really
- 1:58heavy hitting infrastructure
- 2:00Um and then I've spent seven years doing
- 2:02startups, and that's a whole different
- 2:03world. It's a lot of risk, a lot of
- 2:05learning like what the market needs, a
- 2:08lot of correcting your own assumptions.
- 2:11And I've also been um I'm based in
- 2:13Atlanta. It's a big film film state.
- 2:15I've been very big in the arts. So out
- 2:17of college, made lots of films, built
- 2:19laser projectors that play video games
- 2:21on the side of skyscrapers, that kind of
- 2:22thing. And those are the sort of
- 2:25customers that our startups or the
- 2:27startups that I built have have catered
- 2:28to, creatives, artists, filmmakers, and
- 2:30so forth.
- 2:32Um
- 2:33I want to transport everybody back in
- 2:34time, back to the very beginnings of the
- 2:37web and the internet, the '90s, the the
- 2:40the early 2000s. So for us millennials,
- 2:42we're going to feel old. And for
- 2:46um everybody that's that's younger, this
- 2:48is just let me kind of set the stage for
- 2:49what the internet was like.
- 2:51Um we had a bunch of chat applications
- 2:53that were completely open. So IRC is
- 2:56like Discord, but you could spin up your
- 2:57own servers that you actually own. You
- 2:59can use whatever programs to automate
- 3:01them and interrogate them without rules.
- 3:03All of the commercial
- 3:05messaging apps were basically they
- 3:07didn't care if you used third-party
- 3:09tools to connect to them.
- 3:11Every website had data exports. They had
- 3:13APIs. You could literally build really
- 3:15cool mashups with Reddit and Flickr and
- 3:18Twitter. And these were like trending on
- 3:21slashdot before Hacker News was a thing.
- 3:23And then, you know, in in the more
- 3:25modern times, these kind of trickle down
- 3:27to
- 3:29a much slower pace because the APIs got
- 3:31got removed.
- 3:33There was a bunch of file sharing. So
- 3:35not not just file sharing, you could
- 3:37actually buy MP3s and feel a sense of
- 3:39ownership in in everything that you were
- 3:41doing on the web. And now it's more
- 3:43licensing.
- 3:45Um
- 3:46there was so much open creativity
- 3:48because largely the people on the
- 3:49internet were technical. They were early
- 3:51adopters. So Flash kids were making all
- 3:53kinds of animations. There were diffuse
- 3:56communities with PHP
- 3:58PHPBB and Envision and just just a ton
- 4:01of creativity that was happening at that
- 4:02time.
- 4:04Wikipedia got started sort of in this
- 4:06ethos, and we don't really see this kind
- 4:09of stuff anymore in the current era and
- 4:12incarnation of tech.
- 4:14Websites were highly customizable, like
- 4:16MySpace at the time. You could actually
- 4:18change the HTML and CSS and make it look
- 4:20however you want. So this was a totally
- 4:23different alien time compared to what we
- 4:26have today.
- 4:30When Web 2.0 happened, broadband,
- 4:31smartphones, it brought everybody into
- 4:33the fold. So the technical people got
- 4:35sort of pushed to the sidelines. And
- 4:37it's not a bad thing. Like we needed
- 4:38our, you know, our parents and our
- 4:39grandparents to be able to use these
- 4:40products. But we sort of we rounded the
- 4:43edges on things. We took away all the
- 4:44technical pieces. We made everything
- 4:46easy for people to adapt because they
- 4:49couldn't contribute from a technical
- 4:51footing. They didn't want to unpackaged
- 4:53things. They didn't know what APIs were.
- 4:55And so the platforms and the the the
- 4:57protocol like protocols got replaced by
- 4:59platforms. And they sort of grabbed as
- 5:03many users as they could. And sort of
- 5:04the diffuse, experimental early web
- 5:06started to disappear.
- 5:08The APIs got taken away from us.
- 5:11And then we started seeing distribution
- 5:13being gatekept by the big providers.
- 5:16Like we never once had a phone ecosystem
- 5:19where we could freely distribute to like
- 5:21like we could on the desktop.
- 5:25Things started to cost more and more as
- 5:27as startups began to like win market
- 5:30share. They realized that they had
- 5:31pricing power and
- 5:33suddenly your Netflix actually became
- 5:34cable once you started subscribing to
- 5:36all the streaming services.
- 5:38And then they started to get very
- 5:40extractive. For companies that maybe
- 5:43maybe hit the limits of their growth,
- 5:45they started coming up with crazy crazy
- 5:47gotchas. So the one that got me is Adobe
- 5:50has this
- 5:51you can subscribe to a monthly plan and
- 5:54they don't tell you this. It's in like
- 5:55the very small print. If you cancel your
- 5:58monthly plan before the 12 months or 18
- 6:00months, they actually charge you up to
- 6:02like $200 to unsubscribe, which is
- 6:05really inconvenient If you're going to a
- 6:06film festival like once a year, making a
- 6:09short film, and then suddenly you've got
- 6:11an extra $200 bill for an application
- 6:13that you were just using over a weekend.
- 6:14And that's that's hit so many people and
- 6:16it's really frustrating
- 6:18that the web that was once really really
- 6:21promising and open and free
- 6:24has turned into this
- 6:27extractive
- 6:28owned rented rent-seeking world.
- 6:33So it it used to be a very special place
- 6:35in the '90s in the 2000s
- 6:37and
- 6:39more and more has been extracted.
- 6:42This is a term that's commonly used in
- 6:45shitification. I think that aptly
- 6:47describes everything that happened.
- 6:51Our company started with the ethos of
- 6:53let's build everything open source. Like
- 6:55the applications that we built in the AI
- 6:56space were
- 6:58100% open, not only the desktop apps but
- 7:01the actual infrastructure as well. So
- 7:03Artcraft originally started as sort of
- 7:05like a professionally minded
- 7:07Higuysfield. So all of the the
- 7:09filmmakers that I know and that I've
- 7:10I've I've grown up with, we were making
- 7:12a product for them to use rather than a
- 7:14UGC centered approach like like
- 7:17Higuysfield or or Open Art or so forth
- 7:19have. And that enabled us to give them a
- 7:22desktop tool that lived on their machine
- 7:25where all the files that they were
- 7:27generating lived right next to their
- 7:28editor.
- 7:29And one of the things that we observed
- 7:32and my advisors thought I was absolutely
- 7:34crazy that artists, you know, they don't
- 7:35know what open source is. They don't
- 7:36they don't need it. They don't talk
- 7:38about it. You know, isn't really
- 7:40used by artists. We actually observed
- 7:42artists vibe coding their own workflows
- 7:45on top of the software that we gave
- 7:46them. And that is something that like
- 7:48was brand new to all of us. Like
- 7:50there were micro drama studios that were
- 7:52doing order fulfillment for some of the
- 7:54top micro drama startups. They would do
- 7:5730 to 50 grand of fulfillment every
- 7:59single month and they had big teams
- 8:01and no product out there fits their
- 8:03needs. Like nobody's going to go and
- 8:04workshop and build that specifically for
- 8:07the type of delivery that they need. Yet
- 8:09they in house have people that know
- 8:11cloud code and they can take the APIs
- 8:13and they can generate an interface
- 8:15that's like that. So we started
- 8:16experimenting with completely open
- 8:18source, giving it to our customers, and
- 8:20they really responded to that. And then
- 8:23we thought,
- 8:24what if we could turn back time on
- 8:26everything? Like all of these things
- 8:28that they're consuming and building
- 8:29would be much more friendly and and and
- 8:32give them way more capacity to create if
- 8:35all of the infrastructure was open
- 8:37source, if all of the application shapes
- 8:40that they wanted to leverage were open
- 8:41source. Like they use Premiere, they use
- 8:44Adobe. Like what could what could we do
- 8:46to give them something that they could
- 8:47actually take away all of the surface
- 8:49area that they don't use and constrain
- 8:51the apps to their specific workflow?
- 8:54What would it look like if they could do
- 8:55that, you know, five or 10 times for
- 8:58each different artist on their team
- 8:59that's doing a particular part of that
- 9:01workflow? And so we wanted to kind of
- 9:04undo what had happened during Web 2.0
- 9:07and everything became owned and and
- 9:09consolidated and and
- 9:11gate kept and just give people back
- 9:13their applications. And AI really is the
- 9:17means for us to do that. So we took a
- 9:20stab at the Adobe Creative Suite.
- 9:24We looked at Photoshop first. We
- 9:25basically prompted, you know,
- 9:27this is the basic shape of a photo
- 9:29editing app. You you've consumed and
- 9:31trained on all of the tutorials about
- 9:33what Photoshop is. That's in your
- 9:35training data.
- 9:36We we laid out a large large prompt to
- 9:39say let's let's build out the the
- 9:42application. Let's separate the UI UX
- 9:44from that. Let's make everything MCP and
- 9:46CLI compatible. Let's build out plans
- 9:49for every single tool, how all of those
- 9:51work. Let's talk about the data
- 9:52structures, the file format importers.
- 9:54We basically built these large prompts
- 9:56and within 24 hours we were looking at a
- 10:00pretty good approximation of what
- 10:02Photoshop was. And it it kind of blew
- 10:04our minds away.
- 10:06It was functional. It was very familiar
- 10:08to us. It started 10 times
- 10:11than the original Photoshop. Now we
- 10:12didn't like actually cover the entire
- 10:14surface area of Photoshop. There's a lot
- 10:16of work. We're still not at 100% parity,
- 10:18but we're measuring ourselves against
- 10:20what has what needs to be done and we're
- 10:22getting closer and closer. We were at
- 10:2450%. We've just passed 60% and in the
- 10:27fullness of time like Pareto principle,
- 10:29we could reach 99.99%
- 10:32parity. Um
- 10:34I'm going to take a little bit of a sort
- 10:35of a diatribe. Our team has talked about
- 10:37whether we take this further now or
- 10:39whether we just go for parity. And right
- 10:41now we're just we want to have we want
- 10:43to fulfill the muscle memory of all the
- 10:45artists that use Adobe. So we are going
- 10:47for Adobe like their sort of interface
- 10:49and look and feel directly. In the
- 10:51fullness of time we will build net new
- 10:53apps that have new interfaces that are
- 10:54more agentic. Um but we wanted to do
- 10:57that also as sort of like a proving
- 10:58function that we have actually
- 10:59accomplished something real completely
- 11:01agentically.
- 11:03Uh so we did Photoshop. We did um
- 11:06basically all of their apps. We did
- 11:09InDesign. We did Premiere. Um and uh
- 11:13this
- 11:14I I did not expect this to go viral. I
- 11:16was withholding posting it on Hacker
- 11:18News. It got out on Reddit and then like
- 11:20two days later like the like New York
- 11:22Times was asking me for an interview and
- 11:23our Discord went from 3,000 people to
- 11:26It's now got more people in our Discord
- 11:27than Jev, Suno, Higgfield. It
- 11:31I I'm still struggling to bear with what
- 11:33has happened. This was a It struck a
- 11:34nerve with people. Uh they've felt sort
- 11:37of maligned by Adobe and other tech
- 11:38companies and uh what we've built has
- 11:41has really resonated with them and it's
- 11:43sort of
- 11:44built up sort of this mission to build
- 11:45more. Uh so all of the anti-AI artists
- 11:49who have been pushing back against AI
- 11:50all of this time have you know they've
- 11:52sort of turned around and they're like
- 11:54we actually love what you're doing.
- 11:56They've started using these apps.
- 11:57They've started giving us like feature
- 11:59requests and bug reports and we're
- 12:01engaging them and I think we're going to
- 12:03be able to teach them this is what an
- 12:05MCP is. This is how you agentically
- 12:07create things and maybe we'll start
- 12:08pulling away some of these bloated
- 12:10interfaces and giving them things that
- 12:12are faster and more fluid, and let them
- 12:13do their work faster. Um, so we didn't
- 12:16stop with those apps. We've done all
- 12:18kinds of stuff. We've done CAD, we've
- 12:20done um uh uh uh uh Flash, we've done uh
- 12:25a bunch of digital audio workstations.
- 12:27It's not just us, it's the community
- 12:29doing these things as well. And there's
- 12:31really no limit in terms of what can be
- 12:33done. Uh so we've done the entire
- 12:35Microsoft well the core apps within the
- 12:36Microsoft Office suite. There's been
- 12:38Google Docs. We have uh companies and
- 12:41universities and all sorts of people
- 12:43approaching us wanting to use these
- 12:45things. And we're we're basically trying
- 12:47to fulfill all of these different asks.
- 12:50Um they're being used on Linux where
- 12:51they've never been used before. Um PDF
- 12:54Craft is probably the best app. You
- 12:55should all go download it right now.
- 12:56You'll never have to use Acrobat again.
- 12:58This is like really fast,
- 13:01really smooth.
- 13:03[laughter]
- 13:06It works on Linux, it works on NetBSD.
- 13:08It's it's brilliant. Like this one's
- 13:10ready to go. So go ahead and get this
- 13:11one. Um we did Figma. We've actually
- 13:14done the server component for this too
- 13:17and you can basically have your team
- 13:19collaborate on on documents. It's got
- 13:21like the raft paxos like multi-user
- 13:24setup. So we're starting to go into
- 13:25infrastructure too. And you know I the
- 13:27entire infrastructure and data model all
- 13:29of it is open source. We're not keeping
- 13:30any of it proprietary.
- 13:32Um
- 13:33there's a community that's working on
- 13:35Solid Craft that's led by engineers.
- 13:37They've actually got uh an open
- 13:40uh engineering kernel that manages all
- 13:42of this. Uh and it's it's really solid.
- 13:44Like the MCP you can actually do 3D
- 13:46printing projects really quickly with
- 13:48Solid Craft.
- 13:49Um
- 13:51we've had so so much happen. We've got
- 13:53over 3,000 300,000 uh GitHub stars.
- 13:564,000 different people have been
- 13:58contributing to this project and it's
- 14:00picking up. It's it's it's actually
- 14:02become sort of a movement if you will.
- 14:04And we're we're trying to adapt to how
- 14:06much influx of of people trying to
- 14:08contribute to this. I'm going to pick up
- 14:09the pace cuz I've been kind of lagging
- 14:11on this. But um uh we're We're about 1.7
- 14:14million downloads a day and this is
- 14:16accelerating. We've got a launcher app
- 14:18that puts us on the desktop, and like
- 14:19you can you can basically manage
- 14:21everything here. It works on all
- 14:22operating systems.
- 14:24A quarter of our users are on Linux,
- 14:26which is really fascinating. Like Linux
- 14:28has been historically really, really
- 14:30underserved, and now they have super
- 14:33high quality professional apps for
- 14:36creative and and office work, which is
- 14:39it's really nice. So
- 14:41we've you know, we we love and respect
- 14:43all of the work that's been done there
- 14:45in the past, but people are really
- 14:46excited to use the apps that they didn't
- 14:48have access to.
- 14:51All of these things are MCP and CLI like
- 14:53first class. You can totally command
- 14:55these, and you can watch them build and
- 14:57assemble creative projects right in
- 14:58front of you.
- 15:00They've been localized into languages
- 15:02that were never like officially
- 15:04supported by Adobe or Microsoft. Like we
- 15:06were told that Arabic was not supported
- 15:08in InDesign. We've also supported these
- 15:11apps past their support lifecycle. Like
- 15:13Adobe Animate was being
- 15:16end of life or it was being like put on
- 15:18life support. We've we've picked that
- 15:19up. We've had basically animation
- 15:20studios come to us saying, we use this
- 15:23for our jobs. Like this is important.
- 15:25Please please support this. So a lot of
- 15:28a lot of really fascinating things are
- 15:29happening where we're getting this into
- 15:31more places than they've ever been
- 15:32before. It's it's definitely not me.
- 15:34Like I I started this and now we've got
- 15:36a community of like 4,000 developers.
- 15:38This is taken on a life of its own. They
- 15:40are doing everything from CI/CD. It's
- 15:43it's incredible. This this is way bigger
- 15:45than just me.
- 15:46I really got to pick up the pace here.
- 15:47I've made way too much content. So
- 15:51there are 300 apps being developed by
- 15:52the community. We've developed an app
- 15:54where non-technical people can start
- 15:56contributing and use the basis of the
- 15:58apps that we've created like the UI/UX
- 16:00libraries. We're decomposing this into
- 16:02all of the the components and primitives
- 16:03so that you can reuse them. So you'll
- 16:05have nice
- 16:073D libraries, nice 2D libraries. It's
- 16:09all great. I'll get into the technical
- 16:11stuff. So
- 16:13Rust was probably essential in this
- 16:15product's product success. We'd seen
- 16:16people doing it in Electron before, but
- 16:18for high-quality creative apps, you
- 16:20really need something that's native. Um,
- 16:23there's just a lot of really, really
- 16:24good quality things about Rust.
- 16:26Google had papers about like the low
- 16:28defect rate. It starts really fast. It's
- 16:31it's the best language for this.
- 16:33Uh,
- 16:35yeah. Um,
- 16:37it supports all all different platforms
- 16:40and operating systems. We've got this on
- 16:42NetBSD. We've got this on tablets. It
- 16:44natively deploys to the web via Wasm, so
- 16:47it can get anywhere and everywhere.
- 16:50Um, do you need to know Rust to start
- 16:52using Rust today as an engineer? Uh,
- 16:55definitely not. Uh, I've been using Rust
- 16:57since pre-1.0 times, but everybody on
- 16:59the team uh it's a mix of people that
- 17:01have never experienced the language
- 17:02before. Your agents will drive it just
- 17:04fine. You're not going to have any
- 17:05trouble. It is a fantastic language. The
- 17:07one big thing you're going to run into
- 17:08is compile times. Um,
- 17:11is it GUI yet? There's this famous
- 17:12website or series of websites in the
- 17:14Rust Rust world. Does it support like
- 17:16modern features? Yes. You can build GUI
- 17:18rich GUI apps in in Rust, UI UX. Uh,
- 17:21we've run into a lot of gotchas like
- 17:22permissions, GitHub, scaling, CI. A lot
- 17:24of people have talked about that. I'm
- 17:25going to just speed through this cuz I'm
- 17:26kind of running out of time. Uh, CI is
- 17:29like falling over.
- 17:30[laughter] We've been spinning up our
- 17:32own. We found a good place to get RAM on
- 17:34co-located servers. We're made of that.
- 17:35Um, this is a good one. So like agentic
- 17:38develop versus human uh review. If
- 17:40you're doing backend work or things that
- 17:41require schema migrations, API design,
- 17:44let your humans do that, but for like
- 17:45desktop apps and games and anything
- 17:47that's basically immutable, throw your
- 17:49agents at it and give them like full
- 17:50confidence. You don't need to be human
- 17:52in the loop fully. Like the upstream
- 17:54like directing where things go, just
- 17:55just let it run. Tell it to build
- 17:57things. Somebody in our community is is
- 17:58building Uh, they've got 40 machines
- 18:00rebuilding Unreal Engine in Rust. So you
- 18:04you don't even have to mind these
- 18:05things. Um, Autocraft is a completely
- 18:09autonomous software factory that we're
- 18:11going to open source. It ingests issues,
- 18:13it triages them, it does agentic
- 18:14development, it post-trains any vowels
- 18:16on all of the decisions. This will be
- 18:17open source soon. Um, basically we've
- 18:20got an entire app factory that's
- 18:22designed in code. You can express this.
- 18:24We'll open source this soon. I'm just
- 18:25going to have to hurry cuz I'm overtime.
- 18:27Uh
- 18:28injustice issues.
- 18:31This is open source code rabbit open
- 18:33source all the things. We're we're doing
- 18:34a lot here and
- 18:36I I thought I would save more time for
- 18:37the technical, but this is really good
- 18:39and please please look out for this. Um
- 18:41it will support all of your agentic
- 18:43development. It supports all of the sort
- 18:45of crafting and curating how you uh
- 18:47approach a project. It keeps your
- 18:48engineers on task. Um again fully open
- 18:51source. All of the decisions made by it
- 18:53are training data that we're going to
- 18:55use for evals and post training on the
- 18:56decision trees. Um
- 18:59let's take back the entire internet.
- 19:00Let's rebuild everything. All software,
- 19:03all platforms.
- 19:10We can we can turn all of these things
- 19:12into proper protocols. Um we can open
- 19:15source the entire world. Let's own our
- 19:17apps, our infra, our protocols, our
- 19:19models, all of it.
- 19:20Um
- 19:22we are hiring. So if anybody is
- 19:24interested and wants to work on 100%
- 19:26open source, none of our code is closed
- 19:27source. All of it is in the open. Um uh
- 19:30you get time to work on your own
- 19:31projects. Just reach out to me at hello
- 19:33at storyteller.ai and thank you so much
- 19:36everybody.
About this transcript
This page contains the full transcript of Let's Open Source the World — Brandon Thomas, ArtCraft by AI Engineer, generated from the public captions YouTube serves with the video. The transcript has 3,619 words across 562 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.