buddybuild

» see buddybuild.com as it lived, on the wayback machine

~ the obit ~

buddybuild started in Vancouver in 2015 around a complaint every mobile team had. Getting an app from a git commit to a tester's phone involved certificates, provisioning profiles, a build server somebody's colleague had configured and then left the company, and a separate tool for collecting the crash that happened afterward. buddybuild did the whole chain. You connected a repository, it built every push, it shipped the result to your testers, and when the app fell over it handed you the stack trace with the tester's screenshot and comments attached.

Apple confirmed the acquisition on 2 January 2018, its first of the year, and the team went into the Xcode engineering group. The announcement on buddybuild's own site was friendly and short, and it carried the part that actually mattered to customers: no new signups, and both the free Starter plans and all Android support would be discontinued on 1 March 2018. Existing iOS customers could keep using the service.

They kept using it for three years. On 1 March 2021 an email went out saying support would end on 31 March, with one month to migrate, and that the operation would close entirely by the fall. The marketing page stayed up through September 2021 still advertising a service nobody could sign up for, and then that came down too. Today buddybuild.com redirects to Apple's page for Xcode Cloud, which is the same idea built again inside Apple and announced at WWDC in June 2021.

~ the rap sheet ~

~ the android half ~

The Android cut is the part developers remember with the most feeling, because it was announced in the same paragraph as the good news. buddybuild had supported both platforms and plenty of teams used it precisely because they shipped both. Apple bought a cross-platform product and gave its Android users eight weeks.

Nobody was surprised. Apple does not run build infrastructure for Google's operating system, and anyone who had watched an acquisition before could do the arithmetic the moment the headline appeared. It was still a bad eight weeks for a lot of small teams.

~ what three years of maintenance looks like ~

The odd thing about buddybuild's middle period is how little changed. The homepage captured in February 2018, a month after the deal, is the same marketing page as the one captured in December 2019 and again in March 2021: the same four feature headings, the same promise to get you up and running today, no banner, no notice, no sign anywhere on the public site that the company had been bought. A service in long-term maintenance under a new owner can look identical to a live one from outside, and this one did for over a thousand days.

~ the engineer's view ~

I have never shipped an iOS app, so buddybuild was never going to be my tool. The part I recognize is the build box. Every team I have worked on has had one, usually a machine under somebody's desk or a Jenkins instance nobody wanted to own, and the promise of handing that whole category of chore to a service you pay monthly is a promise I understand from the inside. The models I work on now get trained and packaged by pipelines with exactly the same problem and exactly the same bus factor.

What buddybuild sold was not really speed. It was the end of being the person who has to fix the build server on a Friday.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »