Changelog Interviews – Episode #327

Untangle your GitHub notifications with Octobox

featuring Andrew Nesbitt and Ben Nickolls

All Episodes

Jerod is joined by Andrew Nesbitt and Ben Nickolls to talk Octobox, their open source web app that helps you manage your GitHub notifications. They discuss how Octobox came to be, why open source maintainers love it, the experiments they’re doing with pricing and business models, and how Octobox can continue to thrive despite GitHub’s renewed interest in improving notifications.



RollbarWe catch our errors before our users do because of Rollbar. Resolve errors in minutes, and deploy your code with confidence. Learn more at

LinodeOur cloud server of choice. Deploy a fast, efficient, native SSD cloud server for only $5/month. Get 4 months free using the code changelog2018. Start your server - head to

GoCD – GoCD is an on-premise open source continuous delivery server created by ThoughtWorks that lets you automate and streamline your build-test-release cycle for reliable, continuous delivery of your product.

Command Line Heroes – A new podcast about the epic true tales of the developers, hackers, and open source rebels revolutionizing the tech landscape from the command line up. Presented by Red Hat.

Notes & Links

📝 Edit Notes


📝 Edit Transcript


Play the audio to listen along while you enjoy the transcript. 🎧

Guys, we're here to talk about Octobox today, but let's talk about and where you guys have been over the last couple of years, just to catch everybody up. Andrew, we have had you on the show a couple of times, I think The Changelog once, talking about 24 Pull Requests and… Long time ago, I think episode #188. We'll link that up. Also, you were on Request for Commits back in episode #3, all about measuring success, with Arfon Smith, and of course, Libraries was/is all about measuring things… So we've had you on – Ben, we haven't had you on the show before, but happy to have you… Tell us, before we get into Octobox, what's up with you guys, what's up with Libraries… Give us just the recent history of what you all have been up to.

Oh yeah, I totally forgot about that.

Forgot about, forgot about being on the show before…? What.

No, sorry; I totally remember being on the show… It's great to be back. It's been a bit of a crazy year in terms of Libraries. I think the last time I spoke to you I had been working on Libraries on my own, kind of in my spare time, building it up from scratch. And I met Ben, actually I think during a 24 Pull Requests event… Was it 24 – or, definitely related to 24 Pull Requests.

Yeah, it was at a local Ignite night, a lightning talk session… I think you did a talk about 24 Pull Requests, and at the time I was working on another project that kind of became the Core Infrastructure Initiative, and it was like "Ah, that seems perfect!" and then we kind of started talking at that point.

Tell everyone real quick what 24 Pull Requests is… Since it is the season right now, go ahead and just give that.

It is the season. 24 Pull Requests is in its seventh year now. 24 Pull Requests is basically trying to encourage other developers and open source users to contribute back and give little gifts to the maintainers of those projects that they've been benefitting from all year round, and kind of trying to get that swarm of people working together to bring in the holiday spirit with software.

This year we've actually kind of changed [unintelligible 00:03:45.04] a little bit. In previous years it was literally "Try and send 24 pull requests during the 24 days in December, on the run-up to Christmas." This year we've opened it up to all kinds of contributions. It's still called 24 Pull Requests, but it's really 24 contributions to open source software in any way that you consider to be a contribution; that might be writing a blog post, or answering Stack Overflow questions, or running an event, or speaking at a conference, or even doing a podcast episode on a particular bit of open source would be considered a contribution. So you can record those alongside your pull request, as different ways of showing how you've contributed back to those open source projects that you have benefitted from all year.

[04:32] That's awesome. I'm happy to hear that you've made it more inclusive. Something we talk about often and stress is the importance of non-code contributions to the open source community; they're paramount, and they're so valuable, so that's pretty cool to see it moving beyond pull requests to more things. Very cool., check that out. You have a couple more weeks to get involved before the month is out, so check that out.

Ben, you were saying that you all met at one of these 24 Pull Requests events back in the day. Andrew, you were working on Libraries by yourself… Take us from there, guys. What happened next?

So at the time I was working at a civic tech company called My Society, and I'd got pulled by Ben Laurie into a group of workshops that became the Core Internet Infrastructure kind of workshop, which was built out of the kind of fall-out of the Heartbleed vulnerability. I was comparing, effectively, an event, and 24 Pull Requests was featured, and Andrew started talking about Libraries and started talking about how he was mapping the relationships between projects and their dependency tree, and at the same time I was trying to work out how to highlight projects that were like OpenSSL, that were part of this digital infrastructure concept that hadn't really had much thought contributed to it, so there's wasn't really a standard for that… And thought "Well, actually, that's pretty much the same thing." Andrew is coming at it from the exploring open source for projects to use and contribute to, and I was looking at it from the perspective of "What's the next thing that could potentially blow up and cause a massive problem for users of open source software, i.e. everyone?" And yeah, we just got talking after that, and it kind of built from there. Two aspects of the same technology, brought together.

I recall you guys got a grant or some sort of funding to work on libraries for maybe a year or 18 months, and you guys worked on it together… We met at the Sustain event; Sustain 1… Sadly, I didn't make it out to Sustain 2 this fall… But we met in San Francisco at GitHub headquarters, and at that time the grant was just about running out, or the time period for that was just about running out. You had been doing (both of you) for a while, and you were looking at what was next and it seemed like what happened next was this move into Tidelift, and working with them… And now you're working at Octobox. Give us just the 30-second version of that history, and… I just wanna highlight your guys' path together and through these different projects, landing us on what we're talking about today, which is Octobox, which seems like it's a different thing altogether than what you've been working on for the last couple years… Which is interesting to see how you got here.

[07:39] The long story short is that we were excited by Tidelift's high-level mission, but when it came down to actually trying to get work done and move towards achieving some of those goals, we kind of clashed with the founders quite heavily. There was a lot of frustration, and we ended up being pushed out of the company… Which has left us in a situation where we can't work on Libraries for, at this point, another six months. Libraries still remains open source. It's AGPL licensed for exactly that kind of reason, as a protection to ensure that it always remains open… And the Open Data release that was released just before we left, in – I can't remember the exact date, but we can always get that added to the show notes… It was also a Creative Commons license. So we kind of have that sat, waiting for the ability to pick that up and work with it again, possibly fork it off, so that we can continue to do that… But we kind of had our hands tied a little bit, of what can we do, to still help the developer community, and try and look at sustainability from a slightly different angle, rather than directly the financial sustainability, but also kind of thinking about developer burnout as an important part of that sustainability.

Octobox had been around for a year and a half, nicely ticking away, and it felt like a good place to jump to, where we can try out some of our ideas and get back to actually shipping software, and then try and also making Octobox itself a sustainable open source project.

So Octobox ( is a project "Untangle your GitHub notifications." So this is a separate client/tool/resource in order to do a better job of dealing specifically with the overwhelming amount of notifications that many maintainers get via this open source project that, Andrew, you began, like you said, maybe a year or two back. Was this just something that you were scratching your own itch, and doing on the side? Because libraries was your main thing, but Octobox - you put it out there and it seems like it's really been attractive to people.

Everything is always connected. Octobox was inspired by the need to do something to keep on top of 24 Pull Requests, which every December it comes around and it kind of punches me in the face with the amount of people that jump on it… Right now I think it's 22,000 developers active on the 24 Pull Requests website this year. It's always a flood of issues and pull requests to the project itself, as well as pull requests to my other open source projects… And I was just left feeling like I had no way of knowing what I supposed to be doing.

Also, the way that GitHub notifications has worked for the past few years is once you look at a notification, it's gone, and you can't get it back, unless you drive it entirely from email, which really doesn't work for me as someone who is a terrible email manager. I kind of wanted to separate those to pieces, but… You kind of get the fear then. If you know that when you look at an issue you might not be able to find or remember that you didn't solve that right away, then you kind of don't wanna look at it and you leave those things there for like "Oh, eventually I'll get to this important issue, but if I look at it, I'll forget that it's something I need to do."

[11:43] So Octobox started off as a simple idea to be like "Let's have a kind of archived state for notifications." It pulls in your notifications over the API, and then it basically says "Everything is unarchived" or is in an inbox, and then you choose when you are done with those issues, or pull requests, or release notifications, or all of the kind of things that you could get notified on GitHub… Then that instantly gives you back a level of control, of going like "Okay, now I know that none of my notifications are ever gonna disappear. Even after I've read them, I've still got a full list", which then we started to layer on different ways of slicing and dicing those notifications. Because now you're looking at a list that never goes away; you're like "Okay, I've got 1,000 different things here that are all incoming towards me. Let me filter out by pull request, let me then filter that by pull requests that have already been emerged, or have been closed, and I can throw those away; I'm fairly confident they're not needing any further action from me, especially if I haven't been mentioned on them since."

We get some information from the GitHub API that says the last reason that you got notified, which could be you were subscribed to this repo, or you got mentioned, or you were assigned to solve that particular issue… And so you basically end up with every possible different way that you can filter down those notifications, to really triage and drive through the list in an effective way, and actually leave you with the things that you haven't done yet, but still need to do… Almost like a to-do list, which then you can work from; new things come in at the top, or an existing issue that you've marked as Done, archived, will actually pop back up, in the same way that an email would in your email client. And the whole thing is heavily inspired by Gmail's interface.

Yeah, it looks very much like Gmail. It's funny that it happened because of 24 Pull Requests, because I just recently had such a scenario during Hacktoberfest… You know, I've maintained small open source libraries for years and I've always had kind of a trickle of things, or other people's issues that I'm involved in, or pull requests of my own on other people's projects, but it's never been an overwhelming thing with GitHub's notifications for me, and I am an Inbox (0) kind of person, so I've just always managed it via just another thing in my inbox. However, I've made the "mistake" (air quotes around mistake, because it was awesome), I accidentally promoted our transcripts repo as a really easy way of getting those Hacktoberfest pull requests opened up. I even said we had the fastest Merge button in the West… Just joking around, but setting myself up to do a lot of work in October.

So during the month of October - and just thank you to our audience, everybody who got involved, because I only kid, it was awesome how many people came out and helped us make our transcripts more awesome. I think I merged over 300 pull requests in that month alone, and my inbox was just completely overrun by notification emails.

So I can definitely commiserate with you with regard to 24 Pull Requests. Now, I'm not quite as industrious as you are; I didn't say "Okay, I'm gonna solve this problem." Of course, I did know Octobox was a thing… But I knew also that October only had 31 days, and after all those Hacktoberfest T-shirts got sent out, everybody would stop contributing to our transcripts repo… To the point where it's back to a trickle, which is kind of how I like it.

But it's hilarious - you have this problem; for you it's gonna happen once a year, and it inspires you to create a tool that now is helping out lots and lots of people, so that's pretty cool.

[15:47] Yeah, I mean, I've kind of put up with it for 2-3 years, and then as Libraries started to take off, Libraries spread across (I wanna say) something like 20 different repositories as well… So with a certain amount of automated actions to tell me that there were updated versions of dependencies as well. For lots of people who have Dependabot or similar services plugged in to their depositories, they're gonna get regular amounts of updates - either pull requests or issues - telling them that there's something to update…

So the number of projects you have kind of multiplies the amount of notifications you get, and it can quickly kind of – the thing I didn't want to do was to stop working on those projects. That would be the other way to solve my notification problems - it'd be like "I can't do this much work." But if felt like a solvable problem to actually enable me to effectively handle more things, rather than it being like "I'm overwhelmed by the amount of humans incoming." Actually, the tooling was failing me, so I kind of just tackled it in similar way that I tackled 24 Pull Requests or Libraries, which is "Let me see if I can spin up a basic Rails app that does just enough to get by, open source it", and encourage other people, if they have the same problem, to dive in and to add their take to it, or to drive it in a way that they feel would improve it.

We've had something like 80 different people contribute some very significant features and design work to Octobox over the past couple of years, that have made it into this kind of really nice, well-rounded, solid tool, that people actually really depend on to get work done now.

Were you surprised by how many people shared in this problem space in terms of needing Octobox the way that you did, considering how prolific you are with the open source work, and then 24 Pull Requests just getting so much attention, as well as Libraries being spread [unintelligible 00:18:01.01] but that's another reason why it hasn't been too much of an issue for me - we have very few repos in terms of number count… But when you have a single project that has maybe 20 repos, it's very hard to track everything.

One thing that surprised me - maybe it doesn't surprise me, but… Well, it does, but it shouldn't. I'm live-thinking this through… It's just the amount of people that have been like "This is THE thing that I've been waiting for. This solves a huge problem for me", which goes to show how many maintainers out there are really drowning in our inboxes, or drowning in our notifications.

The thing that surprised me the most is that the amount of notifications I get is tiny compared to people who are maintaining projects as big as VS Code, or Electron. They get absolutely drowned in notifications, and have already had a number of systems in place. Mike McQuaid, who is the maintainer of Homebrew - Homebrew is one of the most active repositories on GitHub, or at least the Formula repository, with updates coming in for new versions of a different Formula multiple times a day… And then actually having the codebase move forward and have to keep up with all the macOS updates…

He's had to build up sets of clever Gmail filters to be able just to compartmentalize all of those different things, so that it doesn't just overwhelm him. That felt like such a – it's a clever hack, but it's such a hack to have to use Gmail to augment the features in GitHub; especially as a kind of allergic to email, that was not particularly useful for me, and it felt like having something that was specific to the developer problem was kind of what (I guess) kept my interest in the project past that simple "Oh, I've got a basic thing working here."

[20:13] 4.74 million notifications managed, and counting. That's quite a few. Ben, you were gonna say something? Go ahead…

I was gonna say as well - in my opinion, Mike has also got a lot of procedure and process around how he deals with people that sometimes he gets some flack for, but it also kind of works as the maintainer of one of the most active projects on GitHub. So it's not just the tooling, but there's also an attitude and a way of thinking about a project that means that he can get effectively quite a lot done.

I think, to be fair, he gets a lot of flack for that, but it's part and parcel of the problem, and this is one of the extra things that people don't necessarily think about… It's not just the tools - the tools are there to support - but also it's how you deal with the project and how yo deal with the people in that project, as well.

I thought I'd say that just because I know Mike gets a lot of flack for that sometimes, but it's difficult for him, and you have to–

He does use [unintelligible 00:21:14.00]

Yeah, exactly. You just have to deal with it.

He's got a number of ways of almost shielding himself from the onslaught.

Yeah. I mean, the thing is, if you wanna take the conversation back just a minute - you started off at the top by saying that compared to Libraries, Octobox seems like a project that's quite separate in terms of what it's trying to achieve, but the threat [unintelligible 00:21:41.27] is for a month your life was full of pull requests with Hacktoberfest, right? And imagine if that was your every day. The thread that pulls through this whole thing is we wanna try and help solve the problem that exists today for maintainers of popular open source packages… And Octobox is part of that. Octobox is one of the tools that helps solve people's problems today… So I would say there's like a pretty strong thread that links from libraries through Tidelift into Octobox, in that respect.

Yeah, definitely in the spirit of what y'all are up to, for sure. I agree with that. I think I was referring to it in terms of functionality it just seems like a different thing altogether, but in terms of what you and Andrew are doing with your life's work these days, I absolutely see those ties, for sure.

Speaking of people who do this every day, as opposed to just during October, we went to Microsoft Build last spring, and we were speaking with the VS Code team. Andrew, you mentioned how they have to deal with a lot of issues… And I'm not sure if this conversation made the final show or not, or if we just had it after we finished recording with them, but they were sharing with us some of the lengths that they go through just to triage their issues… Not even to deal with them necessarily, but to be the incoming person who labels, and assigns for code review, or closes things that are off-topic and whatnot… They have a full-time employee there. They transfer ownership of this triage position on the VS Code issues, similar to how you'd be on like PagerDuty. This is a full-time thing that they're adding to the other work that they're doing, just to maintain the status quo. That doesn't mean to get down to zero issues, it means just not to let it explode into thousands and thousands, so… Yeah, there's lots of people with these problems.

[23:41] Yeah, the Microsoft team - especially the Microsoft Open team - are one of the biggest users of the hosted version of The amount of stuff they get coming in is just overwhelming just to look at it, and I'm not even involved.

Talking about that triage and the way that they manage that - we were actually talking about potentially a feature to enable that within Octobox, to kind of have the ability for a team who have a lot of incoming support requests, or activity from external people, rather than, say, an internal team that are mostly doing progress and their issues are managed within, say, more like a project management tool. Actually, you get kind of an interesting shared inbox, despite Octobox looking like it's literally just a tool for the individual developer and it's entirely based on their context and their view of all of the work that they're involved in. You could actually get to the point where a team could almost triage and work through a certain amount of the other team members' inboxes for them.

You can imagine, like, "Oh, I've seen these five new issues come in, so I label them up, I close the ones that didn't make any sense, and I commented to ask for more details on these ones." Actually then, for the other team members, they could filter their inbox in Octobox to go "Well, lower the priority of everything that has had someone else on the team go through and touch these things", thereby leaving me with things that haven't been touched yet, or things that I'm involved in the conversation. Effectively then, the team has the ability to share the work and pre-filter for each other, so especially if you're distributed across timezones, you're able to essentially treat it almost like a help desk, without needing to literally use a help desk piece of software, because everyone still reports their issues via GitHub issues.

[25:59] to [27:11]

So let's talk about how Octobox works, and that will lead us into what y'all are doing with the Octobox app, and the GitHub marketplace, and trying to make a real go at this. Andrew, you said you started up a Rails application… I'm on, I can sign in right here; I assume you can also run this on your own server, maybe you've got a "Deploy to Heroku" button… Tell us about the way it works and the way people use it, and then we'll go into where it's going from there.

Yeah, so… Interesting how you said that it's slightly different from the other projects that I've worked on in the past, like and 24 Pull Requests… Octobox has that challenge of where and 24 Pull Requests are kind of the one instance that's running online. You can run it yourself, but it's not designed – it works best when there's multiple people all using the same thing, the network effect, and all of the data is in that one place.

[28:06] With Octobox, it was kind of designed from the start for anyone to be able to spin up their own version, and that's mostly from a privacy point of view, that you just might not want to give me access to your notifications, or to have them stored on Heroku - you might just wanna keep those things to yourself - as well as enabling people to use it for their GitHub Enterprise installation. So you can actually point your version of Octobox at your own company's internal GitHub enterprise, or, and then suck down all of your notifications from there.

Shopify are big users of their own hosted instance. I think they had something like seven million notifications across their internal team on their GitHub Enterprise installation, which was like "Whoa…"

That's actually bigger than's installation.

Yeah. That's a big company.

And another big user is GitHub itself. They run their own internal Octobox instance, which gets used a lot… Which is really surprising, because it's kind of an admittance that notifications isn't as good as it could be.

Yeah… We'll definitely get into that when we get to the business side, because I've got questions there… But yeah, continue with this instance thing that you were telling us about.

So the main way that most people would deploy Octobox is using Docker. It has a docker-compose file that will basically group everything up - Postgres, Redis, the Rails app, and stand it up in basically one command, in a similar way to deploying to Heroku, and that's gonna be configured as a GitHub app, OR… So this is where it gets – it's kind of perhaps why no one else has built this before, because the GitHub permissions gets really, really weird around notifications. A lot of the GitHub permissions APIs are designed around individual repositories; the GitHub app, kind of the new – it's not the GraphQL in particular, but the new GitHub app style setup specifically for the marketplace is designed entirely around "You install this integration into a single app or multiple apps within an org."

The notifications API is different to that, because it's based entirely on the user that enabled it… So it spans across every repository that that user has access to. So to be able to download a user's notifications and then also be able to pull in extra information, so the status of an issue, or the labels on the pull request, or if your CI is passing or failing on the pull requests - we actually then need to go and hit the individual endpoints for each of those bits of data. The notifications API is not available via the GraphQL API, so we can't do a nice N+1 set of queries in one go.

And you also then kind of have to work out, like, "For each of these notifications, do I have the ability to pull in the extra data for each one of these subject types?" So it gets slightly complicated in the different ways that you can configure it. And actually, if you run it yourself, the simple way is to plug in your own personal access token, and that enables everything, because the personal access–

That's always the easiest way, isn't it?

[31:54] Yeah, it gives you full permissions and doesn't require permission from the owners of the organizations that would be the gatekeepers to install in the GitHub app. But the GitHub app does come with the nice benefit of the webhooks, so we actually can then listen for any changes to issues or pull requests and instantly react by syncing that data in and updating your notifications, as well as using that as a hint to update other people's notifications that may have heard about that before… So it kind of speculatively syncs people's notifications as it hears from one of the webhook events, and that has made everything seem like it all just happens magically.

I was gonna say, in lieu of that webhooks, are you pulling on an interval then, if you don't have the webhook capabilities?

We basically have a sync button, which is very much like Gmail or Apple Mail's kind of check for email.

You don't wanna exacerbate the problem with people who feel like they're overwhelmed by notifications already, right? You don't want that…

Just keep piling them in…

…new notifications popping in at the top when you're in the middle of triaging them, so…

I see, so this is by design so that you – you have to go check your notifications by your own agency, and you have to say "Okay, sync my notifications, because I'm ready for the flood", as opposed to them just popping in, "You have a new email, you have a new email…" all day long.

Yeah. And that was pushed, again, by 24 Pull Requests, because I was refreshing the page and seeing more things come in; I was like "I can't keep up with this. I need to chunk this up into at least something that can fit in my brain until I can [unintelligible 00:33:40.29] done with all these, and now I can check to see if there's some more.

So I'm just sitting here, staring at the UI as we talk, and I think – and I think this about Gmail all the time, too… I'm like, "This seems like it's better as a desktop app", not as a hosted thing with either your own instance, or the shared instance… Was that anything you ever considered, or you just reached for your must trusty tool, which is Rails, and built a web app right away?

That was definitely – kind of the start was "Oh, I can solve this with Rails. It's using a lot of the same libraries that and 24 Pull Requests are using to interact with the GitHub API", but it also works to building a web app that is mobile-friendly means keeping data in sync across multiple devices. Octobox works really nicely on your phone, so that you can actually kind of – I can look at a notification, and then when I get back to my laptop, it's still there, and I can continue to work on it… Which then means you don't need to build different clients. But you can wrap those up so there's a nice selection of tools to put a website inside of an Electron desktop wrapper. There's one that – I forget what it's called; I'm sure we can find one and stick one in the show notes, but you command line with a URL to a website, and it will just hook up everything, generate a Mac app or a Windows app, and then it'll pick up the nice icon from – I guess it's the Apple Touch icon… So everything kind of happens automatically, and then you have it sat when you're ready.

But personally, I'm kind of against the always-on, constant stream of notifications. I like to actually choose and be thoughtful about when I'm gonna check my email, or when I'm gonna get notified about new pieces of work, because it can just be distracting if you're in the middle of some fairly complicated bit of code… You don't wanna have someone reporting a bug on one of your other projects pop up and kind of distract you. So it wasn't kind of a "I'm imagining this as being something that would always be on, and always be able to tell you when there's new things."

[36:12] Well, I can get that sense from the Sync button, and I also am somewhat intentional with the way I use specific communications. I make rules for myself… For instance, on my phone I will not set up my mail client to pull in new emails and show me the Unread count, because I'm a completionist, and I can't deal with an Unread count; I have to go read it. So I always check my mail. My old saying was "Don't let your email check you. You check your email." And then I can only check Twitter on my phone. I actually break this all the time, but I had this rule; it was hard set for a while - I would not check Twitter from my laptop, it had to be phone only… Just to kind of silo things and feel like I'm in control and the devices aren't in control. Ben, do you have any sort of things like that? You seem like you're keen on that Sync button being intentional… Is this something that you work through?

I don't mind having an inbox… I use the Apple Mail client, and my process has always been for the last decade like I'll read something and flag it if I need to do it later, and I might panic if I have too many flagged emails. So I would in Gmail use the equivalent of a star, in Octobox we have stars as well, and we go by that.

I'm not the Inbox (0) kind of person, but I completely use the same batching process, and I think maybe the three of us are the same type of person, and it turns out there are a fair few other people like us as well out there, who prefer to batch those kinds of tasks up. That's what Octobox is - Octobox is a shift in paradigm from what is effectively like an activity feed style notifications experience to something that's more inbox.

You also don't wanna carry around that kind of mental baggage of knowing that you've got things… I don't know if you have this - if you copy something and you're going to paste it, can you kind of feel that it's on your hand? You've pressed Cmd+C and you can almost feel it until you put it down with Cmd+V. It's there, like, mentally… [laughter]

You know it's still there, right?

…and I have that with some important messages, or issues, or notifications, that I'm like "I can't put this down until I know it's somewhere that I'll be able to make sure it's there…" And yeah, a clipboard manager changed my life. Once I – I was like, "Oh, I don't have to worry about this thing going away or not", and then I became a lot less scared of having something I've copied that I haven't yet pasted…

Do you have a good clipboard manager you'd recommend for me? Because I've always looked at them and I've never found one that I actually liked.

I use Alfred, which is a combined search, and it has a clipboard manager built-in, with kind of a nice ability to search through those clipboard things, and I use that hundreds of times a day… It's wonderful. As well as just being a much better Spotlight replacement.

Well, lots of Alfred fans out there… I'm [unintelligible 00:39:09.16] with my software. If the operating system provides it, I'll tend to use that, so I just use Spotlight, and it's just like good enough, because I don't wanna have to install yet another thing and manage it… But that's because I'm particular in my ways, as we all find out we are.

That's funny that you think that when you have the copy, but not the paste, you can almost sense it on your finger… It's like part of you until you can put it down. I definitely have that in my head, but I don't know about how strong the sensation is as it is for you, Andrew.

I think that's possibly the greatest compliment you could give to the UI developers of desktop software in general, is that you actually feel that it's in your hand and have to drop it. [laughter]

Literally walking around with…

"I can't get it off…!"

[39:52] It's similar to like if I've not put my kids somewhere down, that I know I've got this – "I need to go put my keys on the hook, because otherwise I'll never find them again."

Well, that's my problem - I can't find my keys, but they're my hand the entire time, and I'm walking around the house looking for them, holding them in my hand, like a fool…

I think you've basically got PTSD for skeuomorphism, right? That's basically what it is.

[laughs] Well, I like that. That needs to be like a Tumblr, or something… Is Tumblr still a website…? Anyways. Alright, back to Octobox - so that's a little bit about how it works. I guess this Rails-by-default in your brain, Andrew, worked out pretty well, because it allowed you to create a centralized service for people who don't wanna manage their own instances and to do all the heavy lifting for them, so hence the GitHub app.

Let's talk about the move from a side project really to something that you guys are trying to give a go as a sustainable open source business kind of thing, with the GitHub Marketplace. What are your plans with Octobox in terms of generating revenue?

Do you wanna go, Ben?

Yeah, sure. Octobox is two things for us - it's a tool for ourselves and for maintainers to solve one of the problems that we feel maintainers have, which is they have lots and lots of notifications, whether it's from many repos, or a single monolithic repo with lots and lots of activity… But for us it's also a trial by fire experience of putting ourselves into this situation that a lot of open source maintainers are in, where they have a project which is popular, which is used, and they would like to work on it more, and they need to find solutions to be able to enable them to do that, whether it's gonna be sponsorship, whether it's gonna be donations, whether it's gonna be paid support…

These are some of the models that have started gaining a little bit of head room in terms of popularism, and have had kind of dribs and drabs of experience from people who have tried [unintelligible 00:41:50.22] One of the things that we wanted to do is just expose ourselves to that as a microcosm of the types of people that we're trying to help; we want to throw ourselves in it.

The main goal for us is to make Octobox sustainable for ourselves and our community, which we're kind of at a bit of advantage - we can play both sides; we happen to be the maintainers of, and there are certain things that we can do with, and we are also members of the community. We're members of the 80+ developers who have contributed to Octobox as a service, that people are running themselves, that people are running in Docker containers, and so on… And we want also to experiment with some of the questions that we've been unable to answer in the past around sustainability of open source software.

Questions like "Do people care more about supporting a commercial entity that provides for a project, or do they care more about supporting the community directly, and being able to ask and answer those questions with data?" We want to just prove it to ourselves and to other people who are in a similar situation to us that it is possible, how to do it, and that it is repeatable. And that extends for the community, as well.

We as the maintainers of Octobox and the operators of are in a certain position when it comes to having that shared instance that people go to, and having a point of focus for users, that some open source maintainers do not have… The guys who work on the key core components that you just include as a library; the people who write API interfaces for things like Redis, and so on. It's very difficult to do some of the things that we can do with Octobox to make money, effectively. So what we wanna do is also demonstrate some of the things that we can do as a collective community to support one another, to support those maintainers that don't have the opportunities to expose the service effectively and make money from that kind of service, so that we can support one another. At this point it's a big experiment in terms of what a new kind of small, but through a certain measure successful open source project can do.

[44:33] We're also trying not to be too restrictive. There's been a spat of open source projects recently that have kind of swung very far the other way in the ways that they try to monetize their projects by some literally forgetting where their roots are in the open source movement, in the free software movement, and kind of rescinding on those basic freedoms of free software.

It feels like there are ways to get around that without needing to literally put up big roadblocks, and that kind of works out really nicely with the way that Octobox works, because because it's open source and because it's AGPL licensed, we don't ask for the copyright of people who are contributing, so we can't change the license without literally reaching out to everyone, or without ripping out all of their contributions, which we're definitely not going to do.

So actually, if people don't like the service that we offer via, they can always go and run their own version. That keeps us honest to a certain degree, and kind of avoids us poisoning our own well, or kind of trying to take everything for ourselves. It forces us to act as good actors within the community, whilst providing a service which people see as valuable enough to pay for.

What's pretty cool here - you talk about the experiment that you are doing, Ben, and you guys have Octobox free for open source projects, with basic notifications for private projects. Of course, it's also open source, you run your own instance, so if you wanna do it that way, that's all good. We're talking about And then it says "To add enhanced notifications for private repositories to your organization's account, there's two ways to pay", and I think this speaks to what you were talking about, Ben, where they'd rather provide funding to a community versus to a commercial enterprise that's really supporting that community, but is somehow distinct from it in terms of limited participation… So you have an Open Collective donation, and then you also have the GitHub Marketplace option.

Yeah. And then we have another page underneath our Pricing page that explains "Wait… What?", which is kind of like… [laughter] It's kind of like, "You probably need a little bit of a back-story in order to understand why we're telling you that you can pay for this exact same thing two different ways." It just explains what we're trying to do, and it's just proving to ourselves - if other people are gonna do something like Octobox, should they create a company, should they have that company set up as a shell that basically provides for themselves and their community, or should they just go whole hog for the community, and which one is more successful?

I mean, sample size of one, but this isn't the only experiment that we're gonna do… But I do think it needs some explanation. And I'd also shout out for the whole "What you're paying for enhanced notifications" - a lot of that is to do with the dance around the notifications API versus the way in which GitHub apps work as well, so…

Yeah… I think we're gonna try and simplify that a little bit in the new year…

I think we have to.

[47:52] It often confuses people, and they think either "Oh, is this gonna be like having to pay $100/user/month?" Or sometimes they think "Oh, I don't need to pay for anything", because everything continues to work with the basic notifications… So I think we've exposed the levels of permissions a little bit too much, with kind of my developer hat on, compared to – someone who comes to it without understanding the complications involved often won't appreciate that, and we should do a better job of explaining it, or hiding the complexity.

Initially it was actually just a lie. It's easier to explain to people that the standard [unintelligible 00:48:44.09] "Free for open source, paid for private projects", but then with the notifications API you get some small amount of private repo notification information as well, and people are like "Hang on, there's something wrong… I've got some private data here." Like, "No, no… That's the way it works. We were lying to you effectively, because we were trying to help you understand it easier." So it's an experiment, and it's still very early on.

I think – when did we go live on the GitHub Marketplace, Andrew? It was probably about four, five, six weeks ago maybe?

Yeah, six weeks maybe.

Six weeks, yeah… So we're still kind of finding our feet on how to talk about the service and what you're paying for, but we'll keep experimenting with that. Going back to the business model though - one of the other things that we're trying to find our feet on and experiment with, and kind of demonstrate to ourselves and others that this might be a responsible way of running an open source company is doing things like saying "We're gonna share 15% of our revenue as a commercial company with the community as well", because we don't want – if the experiment about "Hey, do you wanna pay like a commercial entity, or do you wanna pay the community?" to end up with "People wanna pay the commercial company", which means Ben and Andrew are gonna win. We want to make sure that the company is tied to the community in such a way that it has to provide for its community, as well.

So there are various things that we're doing as an experiment that aren't just overtly user-facing in terms of pricing, but they're also kind of behind the scenes in terms of how we set the company up, how we commit the company to the community… And in the future, the goal is if the commercial company holds more revenue than people have donated to the community, to pull people from the community into the commercial company as contractors, employees and so on, and work out how that relationship would work most effectively, so that people have the protection of a commercial company that owns in this particular instance, and they also have the freedom to come in and do paid pieces of work on not only Octobox as an open source project, which has benefits for the whole community, but maybe even specifically for if they see something that needs to be done on that particular instance.

So it's all an experiment, it's all gonna be as well documented and publicized as we possibly can when we have that evidence. Andrew and I - we both live and die by the evidence of the data that we collect, that comes back from working with things like libraries, where we have this vast wealth of data that we can pull from.

Talking of data - we've literally turned off Google Analytics on last week… To give ourselves just a little bit more challenge.

I was gonna say, because you don't want data…? I thought you guys were evidence-based.

Yeah, but also, my background is in computer security and I'm a massive privacy wonk. [laughs]

[51:42] It turns out that quite a lot of developers actually block Google Analytics, as well… Comparing the data that we were seeing from usage on from Google Analytics, compared to coming through Cloudflare and compared to our server logs - it was wildly different, so we could really rely on it to make many good decisions anyway… So we figured we'll just cut Google out of it, make the page more secure, make it faster, and stop exposing people's data out there, while still being able to have a good indication of how people are using it via Cloudflare and our own server logs.

The other area we don't have much visibility on is how much it gets used on individual instances, so… It's been downloaded almost 600,000 times from Docker Hub. We have little visibility, but I get the feeling that there is a lot more going on with Octobox outside of our control. People want Octobox to continue to work and be developed, as well as kind of integrating with what Microsoft are pushing forward with the changes they've recently came in and started to encourage… So you just see the bookmarks have shown up in notifications, so you can bookmark an individual notification - we'd like to be able to feed that back and kind of sync it up with the stars in Octobox, so that you can have your data the same in both places.

So it's not like Octobox is gonna be a done project any time soon and we won't need ongoing maintenance. There's a lot of work to keep it moving forward, with this moving target that is the GitHub universe.

[53:47] to [56:25]

Andrew, one thing that you mentioned is how GitHub has recently put some efforts into notifications with the bookmark feature. We know GitHub under new management, new CEO, Nat Friedman - he seems to be very focused on small polish and improvements to areas that maybe have been neglected over the last few years, that power users such as maintainers care about… And of course, Octobox is all about the power user.

Any concerns with GitHub basically building what you guys have built internally, as a first-party thing? Of course, anytime you build on a platform you don't wanna get sherlocked, as the Mac community well knows… Apple is known to sherlock their platform vendors. And here you are, you're on the GitHub Marketplace… What are your concerns about GitHub and replicating some of these features and making Octobox not quite so intriguing to folks?

I kind of swing backwards and forwards on this, but I have spoken to a few people internally at GitHub since the acquisition, and I'm fairly confident that Octobox isn't in their line of sight right now. Octobox really works well for a particular kind of user - the users who are not wanting to use email, but are getting a load of notifications and are driving across a number of different repositories, which is actually quite a particular set of users… And trying to solve those problems while still enabling all of the other kinds of users that are on GitHub actually becomes incredibly difficult.

We can kind of lean on the fact that we're only solving problems for a particular kind of power user, that we can be like "Well, actually, Octobox might not be for you if you're okay using email." Or "If you only get a few notifications or you're not working with many other people on a private repository, that's okay. Octobox isn't for you. But you can use it if you like." But we're really gonna focus on making those people who are doing huge amounts of work on managing a lot of communication to really make those people seem like they've got super-powers, because they've kind of been struggling for a long time… And a lot of the testimonials we get are people kind of suddenly feeling like they're back in control or they're able to actually then start to take on more things. You're kind of like, "Oh, I could never possibly watch this repo, because I already get too many notifications." And then suddenly, you're like, "Actually, maybe I can."

I personally started watching the Ruby on Rails repository again, after I guess five years of not doing it.

Oh, wow.

I used to do it back when I was a junior, learning Ruby on Rails and kind of wanting to see what the masters were doing. I would watch for what was happening on the pull requests, and of course, then as work happens, you're kind of like "Oh, this is too much…" But Octobox has actually let me compartmentalize that enough to be like "Oh, let's see what's going on over here" and then put it away again, so you can context-switch out to just the Octobox stuff, but when December comes around, I just wanna look at the 24 Pull Requests stuff. Then once I've gone through that, then let's go see what else is still there, if I have the bandwidth for it. Otherwise, it will be there tomorrow.

[01:00:05.11] That's pretty cool. I used to follow that repository as well, and I think I lasted maybe two or three weeks, and I had to just – I just couldn't. Even just like "Meh, I'm not that interested." Do you ever do that? You star a repository or you subscribe, and then a few emails come in and you're like "I'm just not gonna do this."

Oh, yeah… You wouldn't believe the amount of – I must have starred over five thousand repositories or more, and my actual activity feed on GitHub is the most useless thing. The homepage tells me nothing…

Oh, it's never been useful, yeah…

…recommendations are like "Oh, here's EVERYTHING."


So I can't get much out of that page, because I just broke it from starring too many things.

I mean, the question is what does a star mean, right? It's something different to everyone. We could definitely [unintelligible 01:00:50.06] on that, so… Maybe not.

I think we literally covered that in the last time I was on Request for Commits.

That's right… That's right. And the answer is a star doesn't really mean very much at all. Because it means something different to so many different people, it makes it very difficult to mean anything at all that's useful.

But it's an interesting point that you say about Nat and GitHub… Notifications was one of the three things that he said in his opening gambit as the new CEO that he wanted to focus on… And we've seen a lot of those improvements, but as Andrew says, I think personally that there are so many more casual users of GitHub than there are the power users that we're catering for. It would be difficult, or at least it wouldn't be a problem that I personally would wanna solve, to bridge between the two in one interface.

Having something on the marketplace, even from GitHub's point of view - I think could be very positive for them, because it can allow them to refocus their efforts on what might be their core group of users, which are the maybe more casual, medium-level, while they still have something like Octobox to offer people.

Yeah. And as the platform, that's what you want - you wanna provide the 80% solution, and then you want that marketplace, which you're still getting cuts out of; it has this cottage industry around your platform, filling in all those gaps that you don't wanna fill in yourself, or aren't worth it for you, but are worth it for somebody who's smaller than you. So hopefully that symbiotic relationship will just continue on forward.

So you're focused on the power users… Octobox is in a place now where it has a nice core set of functionality; it seems like you've got a good 1.0, or I don't know if you consider this 2.0 or what version it's at… But it's there, it's available, it does what it's supposed to do, but for power users, we always want more, better, faster, deeper, more power, so what are you guys thinking about Octobox moving forward? Some things that maybe current users can look down the road and maybe hop in and help out with, or even give a thumbs to "Yes, I want this feature. Where are you gonna take Octobox in the next 6-12 months for the power users?

I love this. There's so many different ways. As you say, we're in a nice place where we can kind of take stock, listen to users, feel their pain… Because we solved a lot of our own pains that now it's like "Okay, now let's go out into the world a little bit more, interview some people and see how they use it."

I regularly watch Suz Hinton when she's streaming on Twitch, and she starts every Twitch coding session by looking at her Octobox and going "Okay, what do I need to work on today?" That's a great way to see – it's interesting; she's using it like this – it looks like maybe she could do something to help her shun a few things away that like "Oh, this isn't ready to work on until next week", so perhaps features that are maybe a little bit more like a to-do list. I can imagine having the ability to snooze notifications, to say "Put this away until next week, because I need to go check on it again" or "I'm totally not ready to deal with that right now."

[01:04:06.29] Or you're waiting on, say, an API change somewhere else, and similarly, maybe having a due date or some other way of highlighting the importance of particular kinds of notifications… That would allow you to really focus down on the things you need to do today.

The other really interesting area is trying to get into some more automation, or – I don't wanna say intelligence, because I really don't wanna add an unknowing machine learning, with no real clear boundary to why it did something, or having to train it up, because what we don't wanna do is have to share behaviors across different users. It's very much like "This is your data that's only used for you." So I can imagine allowing users to say "Well, if a project comes in and it's been labeled as, say, a bug, then can we automatically assign that to this person, if it's on this repo?" Or "If it is a notification from a bot user, then I just wanna automatically archive that… Potentially even for people who have usernames like me."

I'm "andrew" on GitHub, and I get a lot of actual spam through my GitHub notifications where people have mentioned "and Andrew", and it comes up in my Octobox, in my GitHub notifications, and it's completely irrelevant to me. So being able to look at it and go "Well, you've never interacted with this repository before. You got mentioned, but there's no indication that you would ever want to do that." Maybe that is actual spam, and that could just be automatically moved away.

But rather than try and do it in a one-size-fits-all, I feel like doing Gmail-style filters and automated actions would probably be the best way to allow developers to build the "if this, then that" that they need, rather than try and make a set of simple actions that would be very easy to do, because the power users are gonna be like, "Well, I've got these very specific sets of things I wanna do when these things happen", and Octobox already has a really powerful search that can let you filter down by every different kind of state to get exactly what you need; every time that there's a state change, fire off, see if there's any search results that match that, and run that particular action on them.

I think there's also some of the stuff that we might see coming in the more medium-term, as well. At the moment it's just in beta, but we have the thread view, which we'll show you the content from the thread of that notification within Octobox, rolling that out to users and putting more time into that, so that people don't have to jump out of Octobox as readily as they may already be… And then that can kind of build into some of the potential team discussion stuff in the future, as well.

I've been using that feature quite a lot. You can enable it from the settings, if you're in, and basically it will give you a three-pane view on a regular laptop size screen… So you can jump through all the different conversations and catch up with them without needing to open many different tabs to GitHub. Then potentially that also opens the door to being able to comment directly from Octobox, or even label and close issues without needing to jump backwards and forwards between the different tabs, that is the current behavior if you're gonna have a lot of things to work through.

[01:08:06.13] There's a balance for us there, between – you know, we talked about GitHub sherlocking Octobox, but also, we don't wanna rebuild GitHub, so it's finding that fine line between where do people want Octobox to be a part of their workflow, and where do they want GitHub or their other existing tools; talking about contribution to open source in the greater, more whole sense… We need to find that line, and one of the things, Andrew, I think you did recently was just kind of reach out on Twitter and say "Hey, we're really interested in talking to people about their current workflow in their tools, whether they use Octobox or not", because we're getting to that point now where, you're right, we do have a reasonable 1.0, and now it's kind of finding your way as a built product, to add and take away things that are gonna make the users that we're building for as productive as possible.

That's the key, really - we wanna solve for people like us, who have the same problems as us… We wanna kind of do that, between Octobox and the other tools that people use.

Yeah. For example, the one thing that Ben and I actually do quite a lot is we have a backchannel for Octobox. We don't set it up as a separate, private repo, but instead we're – depending on where we're chatting at the time, it might be text message, it might be in Slack DM, but often it's about Octobox. Potentially, with this thread view, we actually have the ability to then allow Octobox users to message each other directly. And I don't know if you were on GitHub back nine years ago, but actually GitHub used to have the ability to send messages to other users, and they yanked it, and it has never come back since…

Right. I can remember that.

But the potential to have that kind of data, that is not just a mirror of GitHub data, but other data in Octobox, or potentially even having an API that allows users to push data from other platforms in… So you can imagine, take your notifications from Stack Overflow and feed them into Octobox, so that you can actually drive multiple different kinds of developer-focused events that maybe act as to-do's, or things that I will need to check on and confirm that I have done something with them - in one nice, unified UI, that doesn't just fall down to the lowest common denominator of email.

Very cool. I like that idea quite a bit. Actually, a lot of these are good ideas, so you guys have a lot of work ahead of you. So is, of course, the website… How do people get involved from a community perspective? Maybe they love that messages idea; they wanna let you know that you should build that, and they will come… Or they wanna get involved and sling some code - what are the waypoints for people to get into the Octobox community and maybe become a user, but also, hopefully a contributor?

We drive most of the development from an issue tracker, encouraging people to propose new features or report bugs. We also have a roadmap document within the repository, so actually proposals to add things to that roadmap are very cool.

The other thing that we use is Gitter, which is similar to Slack, I guess, but focused on GitHub repositories, or GitLab as well, now that they – I think they got purchased last year… So that's kind of the more real-time chat area, and it's completely open. You can just drop in and someone will probably be hanging out there.

[01:11:58.27] And then we also try to make the project really friendly for new people to get in, so one of the design or architecture decisions was to try and stick with Rails conventions as much as possible. So if you have any experience with Ruby on Rails, this will feel right at home. You'll be able to find exactly where you would expect the logic to be, because we don't try and do custom bits of code that stick outside of it; it's literally like models, controllers, views, Turbolinks, and Postgres with Sidekiq as the queue… So it's very easy to get involved.

We've had people build whole features, because it was so easy for them to dip in and go like "Oh, I recognize this… This is a Unix system." [laughter]

Jurassic Park reference? Nice…

…and literally be able to fix bugs. I'm probably pretty bad because I'm checking Octobox so often that I will often merge a pull request and roll it out within minutes of seeing them sometimes. Then other times they'll get merged, but maybe not rolled out straight away. We don't have continuous deployment mostly because we are the only people who are responsible, so kind of 24-hour ops makes me slightly more cautious… But absolutely open to different ways of contributing, as well as potentially even as people can spin up their own forks. We have a lot of people that fork the project, making changes, seeing how they work for their own instance, and then seeing if they can suggest them as features after they've kind of kicked the tires on them a little bit.

Very cool. Well, guys, thanks so much for coming on the show. Thanks for all the work you've done on Libraries, 24 Pull Requests, which was super-cool… Octobox - I hope you guys have great success with this. Hey, come back after the experiment has some facts that we can get some evidence back, find out what people are up to with regards to do they wanna support a community, do they wanna support a commercial enterprise? Do they wanna do both, or maybe do they wanna do neither? Maybe that's what we'll find out… Or hopefully we won't find that one out.

And then also, one thing we didn't talk about and maybe we'll just tease it now and have you back for another show later - all about how you also have this desire to divvy out revenue to your downstream (or is it upstream? I don't know), to the stream dependencies that help Octobox be what it is, because that is something that I definitely wanna talk about. But we're out of time for now, so we'll have you back on later on to talk about that and get a follow-up to find out how you all are doing.

For now, that's our show. Thanks so much for joining us. Ben, Andrew, it's been a joy.

Thank you.

Thanks very much.


Our transcripts are open source on GitHub. Improvements are welcome. 💚

Player art
  0:00 / 0:00