YouTube2Text

Automate tasks with the Codex app — Transcript

by OpenAI · 907 words · 132 segments · language en · Watch on YouTube

Full transcript

  1. 0:00So there is a lot about the job
  2. 0:02isn't actually that fun that I have
  3. 0:05now automated away almost entirely.
  4. 0:09And so I'll show you a few of those
  5. 0:10here in our automations tab. And we've
  6. 0:13got a bunch running now, but we can
  7. 0:15run down a few of these because I kind
  8. 0:16of have set up a few different types
  9. 0:18of automation. And the first type is
  10. 0:19just sort of information. I'm asking
  11. 0:21Codex to look at the last day of
  12. 0:23commits into this section of our
  13. 0:26monorepo. And basically group them,
  14. 0:28tell me who worked on what, what got
  15. 0:31done, what do I need to know about.
  16. 0:32And so I wake up in the morning with
  17. 0:34almost like a, if you've used JotGPT
  18. 0:36Pulse, it's like a personalized pulse
  19. 0:38that I walk, I open the app in the
  20. 0:40morning and sitting there is like,
  21. 0:41here's what happened yesterday in the
  22. 0:43code base, which is super,
  23. 0:46super useful. So I have one called
  24. 0:47Upskill. And what Upskill does is
  25. 0:50it looks at the last day of skill
  26. 0:52usage. So it says, hey, skills are
  27. 0:55really useful, but sometimes they take
  28. 0:57a little bit of tinkering to get
  29. 0:58right. And so it looks at Codex usage
  30. 1:00throughout the past day and detects if
  31. 1:03any skills, like if Codex had any
  32. 1:05trouble with skills or if any of the
  33. 1:07scripts didn't work correctly or if
  34. 1:10like something could be sped up. And
  35. 1:11it makes improvements to the skills.
  36. 1:13So it's upskilling like overnight.
  37. 1:15Like I'm going to sleep, I wake up,
  38. 1:16Codex is smarter in the morning based
  39. 1:18on what we're trying to do here.
  40. 1:20Similarly, update AgentsMD, like
  41. 1:22this I just have run every, I don't
  42. 1:25know, six hours. And it just looks at
  43. 1:27what we've been doing. And if there's
  44. 1:28any like disagreements or
  45. 1:30misunderstandings that I've had with
  46. 1:32Codex or things where things took
  47. 1:33longer to communicate than, or I used
  48. 1:36some shorthands that Codex wasn't
  49. 1:37familiar with, it will just add that
  50. 1:39to the personalization. So that next
  51. 1:41time things can just be quicker.
  52. 1:44I've got a few others that are
  53. 1:45integrated deeply into our other
  54. 1:47tools. And so one of them is Sentry.
  55. 1:50This is kind of a different type of
  56. 1:52automation where it's going to go
  57. 1:54through on whatever schedule, whatever
  58. 1:56cadence, and it picks off one of the
  59. 1:58top Sentry issues. So that could be
  60. 2:00like a performance regression or
  61. 2:01a crash or an error that's being
  62. 2:04thrown in the app. And it like goes
  63. 2:07through all of the information that we
  64. 2:09have about it, the logs, the maps,
  65. 2:12everything that Sentry has, it looks
  66. 2:13at the code base and it picks
  67. 2:16something to fix basically. And
  68. 2:18automations have memory. So it
  69. 2:20remembers what it tried to solve last
  70. 2:22time. And so you're not going to get a
  71. 2:23PR like solving the number one issue
  72. 2:25every hour, right? But it's really
  73. 2:28good because what this, you know, what
  74. 2:30this type of automation does is it
  75. 2:32allows us to focus on what is going to
  76. 2:35make the Codex app great to use and
  77. 2:37useful and joyful and all of the
  78. 2:40things that like, you know,
  79. 2:43bugs and things that fall through
  80. 2:45cracks that like aren't quite as
  81. 2:46fun to figure out, at least for me.
  82. 2:48I know some people love it.
  83. 2:51That stuff can happen kind of behind
  84. 2:52the scenes, right? And so it's really
  85. 2:54common that you run into merge
  86. 2:57conflicts when you try to merge your
  87. 2:58thing. And like you find, you know,
  88. 2:59you iterate on your approach and
  89. 3:01you've got like different types of
  90. 3:03people giving reviews and then you're
  91. 3:05finally ready to merge something and
  92. 3:06there's a conflict. Your CI is failing
  93. 3:08for, you know, some small reason.
  94. 3:10You got to kind of dive into that. And
  95. 3:12so I don't like dealing with any of
  96. 3:14that. Maybe, you know. Most people
  97. 3:17probably don't. Some people love doing
  98. 3:19that and they work on CI tools.
  99. 3:22But I have a green PRs automation here
  100. 3:24that uses some of our skills. So
  101. 3:26we use BuildKai, we use GitHub. And so
  102. 3:28what this does is it checks if I have
  103. 3:31any open PRs and I usually have like
  104. 3:34at any given time between 10 and
  105. 3:3520 PRs these days open at a time
  106. 3:38because of the Codex app. Separate
  107. 3:41problem. And this is going to go
  108. 3:43through and it's going to say like, do
  109. 3:45any have CI failures because of like,
  110. 3:48you know, something that like linted
  111. 3:50wrong or like didn't get caught in the
  112. 3:53pre -push? Or maybe it's going to
  113. 3:55merge conflict because someone is like
  114. 3:57also changing it because we're rushing
  115. 3:59to, you know, finish up a project or
  116. 4:01something like that. And this is going
  117. 4:03to go through and it's going to make
  118. 4:04sure that all my PRs are green. It's
  119. 4:06going to update the base branch.
  120. 4:07It's going to intelligently resolve
  121. 4:09merge conflicts by looking at like
  122. 4:10what have people been trying to do. So
  123. 4:12it's not just a factual like, you
  124. 4:15know, I'm going to clean up the like
  125. 4:17conflict markers. It can look at,
  126. 4:20oh, like Joey was putting in this
  127. 4:21feature. So we changed this block for
  128. 4:23this reason. I'm trying to do this.
  129. 4:25I'm changing it and there's a conflict
  130. 4:26and like kind of knows how to do that.
  131. 4:28And so that's a very long way of
  132. 4:30saying my PRs are always green.

About this transcript

This page contains the full transcript of Automate tasks with the Codex app by OpenAI, generated from the public captions YouTube serves with the video. The transcript has 907 words across 132 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.