read-only demo

Videos ClWD8OEYgp8

Collaborative AI Engineering: One Dev, Two Dozen Agents, Zero Alignment — Maggie Appleton, GitHub

index_state ready data_status ok

AI Engineer· published 2026-04-26· 0:17:42· en· indexed 2026-08-10 19:49

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:05, 1 of 1 keyframes kept
  2. Shot 1, 0:05 to 0:09, 1 of 1 keyframes kept
  3. Shot 2, 0:09 to 0:14, 1 of 1 keyframes kept
  4. Shot 3, 0:14 to 0:34, 1 of 1 keyframes kept
  5. Shot 4, 0:34 to 0:49, 1 of 1 keyframes kept
  6. Shot 5, 0:49 to 0:58, 1 of 1 keyframes kept
  7. Shot 6, 0:58 to 1:05, 1 of 1 keyframes kept
  8. Shot 7, 1:05 to 1:22, 1 of 1 keyframes kept
  9. Shot 8, 1:22 to 1:41, 1 of 1 keyframes kept
  10. Shot 9, 1:41 to 1:52, 1 of 1 keyframes kept
  11. Shot 10, 1:52 to 2:30, 1 of 1 keyframes kept
  12. Shot 11, 2:30 to 2:42, 1 of 1 keyframes kept
  13. Shot 12, 2:42 to 3:18, 1 of 1 keyframes kept
  14. Shot 13, 3:18 to 3:52, 1 of 1 keyframes kept
  15. Shot 14, 3:52 to 4:26, 1 of 1 keyframes kept
  16. Shot 15, 4:26 to 4:53, 1 of 1 keyframes kept
  17. Shot 16, 4:53 to 5:26, 1 of 1 keyframes kept
  18. Shot 17, 5:26 to 6:00, 1 of 1 keyframes kept
  19. Shot 18, 6:00 to 6:19, 1 of 1 keyframes kept
  20. Shot 19, 6:19 to 6:51, 0 of 1 keyframes kept
  21. Shot 20, 6:51 to 7:07, 1 of 1 keyframes kept
  22. Shot 21, 7:07 to 7:20, 1 of 1 keyframes kept
  23. Shot 22, 7:20 to 7:23, 1 of 1 keyframes kept
  24. Shot 23, 7:23 to 7:32, 0 of 1 keyframes kept
  25. Shot 24, 7:32 to 7:33, 1 of 1 keyframes kept
  26. Shot 25, 7:33 to 7:36, 0 of 1 keyframes kept
  27. Shot 26, 7:36 to 7:39, 1 of 1 keyframes kept
  28. Shot 27, 7:39 to 8:07, 1 of 1 keyframes kept
  29. Shot 28, 8:07 to 8:08, 1 of 1 keyframes kept
  30. Shot 29, 8:08 to 8:10, 0 of 1 keyframes kept
  31. Shot 30, 8:10 to 8:13, 1 of 1 keyframes kept
  32. Shot 31, 8:13 to 8:21, 0 of 1 keyframes kept
  33. Shot 32, 8:21 to 8:23, 1 of 1 keyframes kept
  34. Shot 33, 8:23 to 8:37, 1 of 1 keyframes kept
  35. Shot 34, 8:37 to 8:38, 0 of 1 keyframes kept
  36. Shot 35, 8:38 to 8:46, 1 of 1 keyframes kept
  37. Shot 36, 8:46 to 9:08, 1 of 1 keyframes kept
  38. Shot 37, 9:08 to 9:20, 1 of 1 keyframes kept
  39. Shot 38, 9:20 to 9:23, 1 of 1 keyframes kept
  40. Shot 39, 9:23 to 9:28, 1 of 1 keyframes kept
  41. Shot 40, 9:28 to 9:31, 0 of 1 keyframes kept
  42. Shot 41, 9:31 to 9:32, 1 of 1 keyframes kept
  43. Shot 42, 9:32 to 9:38, 1 of 1 keyframes kept
  44. Shot 43, 9:38 to 9:43, 1 of 1 keyframes kept
  45. Shot 44, 9:43 to 9:53, 1 of 1 keyframes kept
  46. Shot 45, 9:53 to 9:54, 1 of 1 keyframes kept
  47. Shot 46, 9:54 to 9:58, 1 of 1 keyframes kept
  48. Shot 47, 9:58 to 10:04, 1 of 1 keyframes kept
  49. Shot 48, 10:04 to 10:08, 0 of 1 keyframes kept
  50. Shot 49, 10:08 to 10:11, 1 of 1 keyframes kept
  51. Shot 50, 10:11 to 10:18, 1 of 1 keyframes kept
  52. Shot 51, 10:18 to 10:43, 1 of 1 keyframes kept
  53. Shot 52, 10:43 to 11:01, 0 of 1 keyframes kept
  54. Shot 53, 11:01 to 11:10, 0 of 1 keyframes kept
  55. Shot 54, 11:10 to 11:17, 1 of 1 keyframes kept
  56. Shot 55, 11:17 to 11:21, 0 of 1 keyframes kept
  57. Shot 56, 11:21 to 11:34, 1 of 1 keyframes kept
  58. Shot 57, 11:34 to 11:42, 0 of 1 keyframes kept
  59. Shot 58, 11:42 to 11:45, 1 of 1 keyframes kept
  60. Shot 59, 11:45 to 12:17, 1 of 1 keyframes kept
  61. Shot 60, 12:17 to 12:26, 0 of 1 keyframes kept
  62. Shot 61, 12:26 to 12:34, 1 of 1 keyframes kept
  63. Shot 62, 12:34 to 12:36, 0 of 1 keyframes kept
  64. Shot 63, 12:36 to 12:38, 1 of 1 keyframes kept
  65. Shot 64, 12:38 to 12:45, 0 of 1 keyframes kept
  66. Shot 65, 12:45 to 12:47, 1 of 1 keyframes kept
  67. Shot 66, 12:47 to 12:56, 1 of 1 keyframes kept
  68. Shot 67, 12:56 to 12:58, 0 of 1 keyframes kept
  69. Shot 68, 12:58 to 13:04, 1 of 1 keyframes kept
  70. Shot 69, 13:04 to 13:07, 0 of 1 keyframes kept
  71. Shot 70, 13:07 to 13:29, 1 of 1 keyframes kept
  72. Shot 71, 13:29 to 13:39, 1 of 1 keyframes kept
  73. Shot 72, 13:39 to 13:42, 1 of 1 keyframes kept
  74. Shot 73, 13:42 to 13:48, 1 of 1 keyframes kept
  75. Shot 74, 13:48 to 13:59, 0 of 1 keyframes kept
  76. Shot 75, 13:59 to 14:09, 1 of 1 keyframes kept
  77. Shot 76, 14:09 to 14:44, 0 of 1 keyframes kept
  78. Shot 77, 14:44 to 15:09, 0 of 1 keyframes kept
  79. Shot 78, 15:09 to 15:35, 1 of 1 keyframes kept
  80. Shot 79, 15:35 to 16:00, 1 of 1 keyframes kept
  81. Shot 80, 16:00 to 16:29, 1 of 1 keyframes kept
  82. Shot 81, 16:29 to 16:58, 0 of 1 keyframes kept
  83. Shot 82, 16:58 to 17:27, 1 of 1 keyframes kept
  84. Shot 83, 17:27 to 17:41, 1 of 1 keyframes kept
  85. Shot 84, 17:41 to 17:42, 0 of 1 keyframes kept

85 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
220
whisperx 220
chunks
32
from 220 cues
keyframes
63
kept of 85 captured
frames with text
63
2,454 lines read
chapters
0
from the source metadata
keyframe bytes
9.4 MB
word timings on 220 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done 2026-08-10 11:46 1m 41s
stt done 2026-08-10 11:48 20s
chunk done 2026-08-10 11:48 0s
text_embed done 2026-08-10 19:49 0s
keyframe done 2026-08-10 11:48 1m 55s
ocr done 2026-08-10 11:50 39s
frame_embed done 2026-08-10 19:49 11s

Frames, and what the machine read

  • 0:03 #0 done2 line(s)

    shot 0·sharpness 666.1

    1. AlEngineer0.98
    2. EUROPE1.00
  • 0:08 #1 done2 line(s)

    shot 1·sharpness 827.0

    1. PRESENTING SPONSOR0.99
    2. Google DeepMind1.00
  • 0:12 #2 done3 line(s)

    shot 2·sharpness 910.6

    1. PLATINUM SPONSORS0.98
    2. # Braintrust0.96
    3. WorkOS OpenAI0.96
  • 0:22 #3 done12 line(s)

    shot 3·sharpness 587.2

    1. Peveloper,0.94
    2. TwaDozen Agents,0.96
    3. AlEngineer0.97
    4. Zero Alignment1.00
    5. EUROPE1.00
    6. Why we Need0.98
    7. Collaborative AI1.00
    8. EU/ACC0.96
    9. Engineering1.00
    10. April 20260.95
    11. AlEngineer1.00
    12. EUROPE1.00
  • 0:39 #4 done12 line(s)

    shot 4·sharpness 1785.2

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.96
    2. Maggie Appleton1.00
    3. Staff Research Engineer1.00
    4. AIE1.00
    5. 1.00
    6. 1.00
    7. GitHub Next1.00
    8. maggieappleton.com @mappletons1.00
    9. GitHub1.00
    10. AlEngineer0.99
    11. Google DeepMind1.00
    12. 20260.95
  • 0:50 #5 done12 line(s)

    shot 5·sharpness 2376.8

    1. Audience - Maggie Al Engineer Talk - Figma Slides0.97
    2. Search("0.95
    3. Search("dlverssty0.92
    4. 0i1ly-dally10p.0.80
    5. AIE1.00
    6. cyclel0.96
    7. 1.00
    8. 1.00
    9. 1.00
    10. 1.00
    11. AlEngineer1.00
    12. Engineering the future of Al0.98
  • 1:01 #6 done13 line(s)

    shot 6·sharpness 2375.1

    1. Audience ·Maggie Al Engineer Talk - Figma Slides0.97
    2. h("cross-cluster diversity0.96
    3. AIE1.00
    4. cycle)0.73
    5. 1.00
    6. 1.00
    7. 1.00
    8. 1.00
    9. lthinking with high effort)0.97
    10. (thinking with high effort)0.96
    11. le (ctrl-o to expand)0.93
    12. AlEngineer0.98
    13. Engineering the future of Al0.98
  • 1:14 #7 done11 line(s)

    shot 7·sharpness 362.3

    1. 1.00
    2. AlEngineer1.00
    3. 0.99
    4. 0.99
    5. EUROPE1.00
    6. One man, two d0.99
    7. EU/ACC0.96
    8. 1.00
    9. 0.99
    10. AlEngineer0.99
    11. EUROPE1.00
  • 1:37 #8 done13 line(s)

    shot 8·sharpness 1456.6

    1. Audience ·Maggie Al Engineer Talk - Figma Slides0.97
    2. 1.00
    3. AIE1.00
    4. 1.00
    5. 1.00
    6. 1.00
    7. 1.00
    8. 0.58
    9. 0.98
    10. AlEngineer0.97
    11. AlEngineer0.97
    12. 20260.89
    13. EUROPE1.00
  • 1:50 #9 done9 line(s)

    shot 9·sharpness 1364.3

    1. Audience·Maggie Al Engineer Talk - Figma Slides0.98
    2. AIE1.00
    3. 1.00
    4. 1.00
    5. 1.00
    6. AlEngineer0.96
    7. AlEngineer0.97
    8. 20260.88
    9. EUROPE1.00
  • 2:14 #10 done11 line(s)

    shot 10·sharpness 1922.1

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.97
    2. Alignment and comunication0.98
    3. is the new bottleneck0.98
    4. AIE1.00
    5. alignment1.00
    6. 1.00
    7. implementation1.00
    8. AlEngineer0.99
    9. AlEngineer0.97
    10. 20260.97
    11. EUROPE1.00
  • 2:39 #11 done11 line(s)

    shot 11·sharpness 2051.1

    1. Audience - Maggie Al Engineer Talk - Figma Slides0.96
    2. always has been1.00
    3. wait, alignment0.98
    4. AIE1.00
    5. is the0.98
    6. 1.00
    7. 1.00
    8. bottleneck?1.00
    9. AlEngineer1.00
    10. Engineering the future of Al1.00
    11. 20260.97
  • 2:53 #12 done8 line(s)

    shot 12·sharpness 1542.2

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.97
    2. 方 方 方 方 方 方0.85
    3. AIE1.00
    4. 1.00
    5. 1.00
    6. AlEngineer0.99
    7. Engineering the future of Al1.00
    8. 20860.80
  • 3:28 #13 done16 line(s)

    shot 13·sharpness 1618.0

    1. Audience - Maggie Al Engineer Talk - Figma Slides0.97
    2. chat threads1.00
    3. docs1.00
    4. issues1.00
    5. PRs1.00
    6. AIE1.00
    7. 1.00
    8. 1.00
    9. 1.00
    10. 1.00
    11. plan1.00
    12. build1.00
    13. review1.00
    14. AlEngineer0.99
    15. Engineering the future of Al0.99
    16. 20260.95
  • 3:56 #14 done16 line(s)

    shot 14·sharpness 1583.8

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.96
    2. chat threads1.00
    3. docs1.00
    4. issues1.00
    5. PRs1.00
    6. AIE1.00
    7. 1.00
    8. 1.00
    9. 1.00
    10. plan1.00
    11. build1.00
    12. review1.00
    13. AlEngineer0.99
    14. AlEngineer0.96
    15. 20260.89
    16. EUROPE1.00
  • 4:29 #15 done7 line(s)

    shot 15·sharpness 369.1

    1. 0.61
    2. AlEngineer1.00
    3. local, unshared plans0.98
    4. EUROPE1.00
    5. EU/ACC0.85
    6. AlEngineer1.00
    7. EUROPE1.00
  • 5:22 #16 done15 line(s)

    shot 16·sharpness 1753.1

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.97
    2. Wasted1.00
    3. Coordination1.00
    4. AIE1.00
    5. work1.00
    6. debt1.00
    7. 1.00
    8. 1.00
    9. building the0.99
    10. cleanup cost of1.00
    11. wrong thing0.97
    12. unaligned efforts0.97
    13. AlEngineer0.99
    14. Engineering the future of Al0.99
    15. 20260.99
  • 5:56 #17 done17 line(s)

    shot 17·sharpness 2042.5

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.96
    2. Bring collaboration and0.98
    3. alignment into the agentic0.99
    4. development flow1.00
    5. AIE1.00
    6. 1.00
    7. 0.99
    8. build1.00
    9. build1.00
    10. build1.00
    11. plan1.00
    12. plan1.00
    13. plan0.92
    14. AlEngineer0.98
    15. AlEngineer0.98
    16. 20260.95
    17. EUROPE1.00
  • 6:17 #18 done19 line(s)

    shot 18·sharpness 2094.7

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.97
    2. Human context is0.99
    3. T0.83
    4. T0.69
    5. not in the1.00
    6. T0.73
    7. codebase1.00
    8. AIE1.00
    9. T0.84
    10. 1.00
    11. 1.00
    12. 1.00
    13. Business context & financial resources1.00
    14. Political dynamics1.00
    15. </>0.75
    16. AlEngineer0.99
    17. AlEngineer0.97
    18. 20260.94
    19. EUROPE1.00
  • 6:35 #19 skipped

    shot 19·duplicate of #15

  • 6:56 #20 done9 line(s)

    shot 20·sharpness 1471.7

    1. Audience ·Maggie Al Engineer Talk - Figma Slides0.97
    2. AIE1.00
    3. 1.00
    4. 1.00
    5. Ace1.00
    6. Agent collaboration environment1.00
    7. AlEngineer0.99
    8. Engineering the future of Al0.97
    9. 20260.89
  • 7:18 #21 done73 line(s)

    shot 21·sharpness 2346.7

    1. Audience·Maggie Al Engineer Talk- Figma Slides0.98
    2. calmhn0.99
    3. +0.98
    4. Chat1.00
    5. 3Diff ×0.86
    6. Terminal1.00
    7. +0.99
    8. Connected0.97
    9. My Sessions0.99
    10. Updated accent color from orange to1.00
    11. G0.92
    12. upbeat-target1.00
    13. emerald throughout the App0.99
    14. #0.99
    15. # Ilm-categories-stati...0.94
    16. vital-bolt0.96
    17. component, changing gradient0.99
    18. backgrounds and hover states to0.98
    19. epic-icon0.98
    20. asdfasdf0.97
    21. Maybe we should add multiple themes, actually0.99
    22. iamnbutler 5:05 pm1.00
    23. align with the green theme.0.99
    24. App.tsx1.00
    25. +2-21.00
    26. create-green-theme0.98
    27. noble-marble1.00
    28. add-Ilm-provider0.96
    29. We could do a theme per category...I started a llm service in #add-llm-provider so we could kick0.97
    30. that work off.0.98
    31. style: change accent color from0.98
    32. orange to emerald1.00
    33. 77ad8f61.00
    34. 1.00
    35. (edited)1.00
    36. AIE1.00
    37. 1.00
    38. Other People's Sessions1.00
    39. 010.88
    40. # pr-110.98
    41. 1.00
    42. # steady-train0.99
    43. [→ davidslater left0.93
    44. 1.00
    45. 1.00
    46. 1.00
    47. # glad-socket0.99
    48. idan 5:06 pm1.00
    49. # add-blue-theme0.99
    50. Cute. So like, Purple for Al-themed posts, a1.00
    51. nge for classic HN shitposting, etc etc0.99
    52. # onboarding-design-eng...0.99
    53. # power-arm0.99
    54. # kind-craft0.98
    55. exec abort by idan0.99
    56. I bun dev0.90
    57. # pr-100.99
    58. ☐ exec by idan0.93
    59. stop running...0.93
    60. bun dev1.00
    61. VITE v7.1.10 ready in 754 ms0.98
    62. Local: http://localhost:5173/0.97
    63. Network: http://169.254.0.2:5173/0.99
    64. Mention #sessions, @people, +files, and / for commands0.99
    65. Chat mode ( Tab)0.98
    66. +0.98
    67. I' ##19 style: update color scheme...0.98
    68. 0:00/0:101.00
    69. 1x1.00
    70. 0.77
    71. AlEngineer0.97
    72. Engineering the future of Al0.98
    73. 20260.98
  • 7:21 #22 done50 line(s)

    shot 22·sharpness 3488.1

    1. Audience · Maggie Al Engineer Talk - Figma Slides0.97
    2. calmhn1.00
    3. +0.50
    4. Chat1.00
    5. 因Diff0.89
    6. Terminal0.94
    7. +0.94
    8. My Sessions1.00
    9. # upbeat-target0.94
    10. What an unprocessed photo looks lke0.99
    11. # vital-bolt0.94
    12. #0.86
    13. #Ilm-categories-stati...0.96
    14. # epic-icon0.96
    15. Maybe we should add multiple themes, actually0.99
    16. AIE1.00
    17. 1.00
    18. 1.00
    19. # asdfasdf0.93
    20. # noble-marble0.94
    21. iamnbutler 5:05 pm0.98
    22. 1.00
    23. create-green-theme1.00
    24. We could do a theme per category... I started a llm service in #add-llm-provi0.99
    25. 1.00
    26. 1.00
    27. 1.00
    28. # add-Ilm-provider0.94
    29. that work o0.92
    30. (edited)1.00
    31. Other People's Sessions1.00
    32. 0010.93
    33. # pr-110.93
    34. [→ davdslater left0.94
    35. # steady-train0.99
    36. # glad-socket0.98
    37. idan 5:06 pm1.00
    38. # add-blue-theme0.99
    39. Cute. So like, Purple for Al-themed posts, and orange for classic HN shitpost1.00
    40. # power-arm0.96
    41. # kind-craft0.98
    42. ⊗ exec abort by idan0.96
    43. # onboarding-design-eng.0.96
    44. bun dev1.00
    45. 0:02/0:101.00
    46. 1x1.00
    47. 0.53
    48. AlEngineer0.99
    49. Engineering the future of Al0.99
    50. 20260.98
  • 7:31 #23 skipped

    shot 23·duplicate of #21

Transcript

220 cues· 3,421 words· 18,407 chars

  1. 0:14 Okay, are we all good?
  2. 0:16 Right.
  3. 0:17 So yes, this talk is called One Developer, Two Dozen Agents, Zero Alignment.
  4. 0:21 This is the case for why we need collaborative AI engineering.
  5. 0:25 So first, a very quick intro.
  6. 0:27 I'm Maggie.
  7. 0:27 I work at GitHub as a staff research engineer.
  8. 0:30 At least that's my title.
  9. 0:31 I'm actually a designer back when that was like a separate thing to engineer.
  10. 0:35 And next is the labs team within GitHub.
  11. 0:37 So we work on kind of more experimental risky bets than the rest of the organization.
  12. 0:41 We like to call it the department of fuck around and find out.
  13. 0:44 And like everyone else, we are of course trying to shape new developer agentic tools.
  14. 0:50 So I think this is what many people think peak developer productivity looks like right now, right?
  15. 0:55 This is like a wall of terminal-based coding agents all running in parallel on one person's machine.
  16. 1:03 I like to call this the one man, two dozen clods theory of the future.
  17. 1:07 So the promise that we're given here is that one person with a fleet of agents will do the work of an entire team of developers.
  18. 1:16 The main problem with this dream is it assumes that software is made by one person.
  19. 1:20 All of these tools are single player interfaces and they focus on scaling up the work of the individual.
  20. 1:26 But there is limited value in scaling up one individual.
  21. 1:29 Because software is not made by one person in a vacuum.
  22. 1:33 It is a team sport and everyone building it needs to agree on what they're building and why.
  23. 1:38 Believing individual productivity leads to great software is nine women make a baby in one month logic.
  24. 1:46 More individual output doesn't solve problems that require communication and coordination, it makes them worse.
  25. 1:53 And implementation is rapidly becoming a solved problem, right?
  26. 1:56 Probably everyone here believes that.
  27. 1:58 Writing code is now fast, it's getting cheaper, and quality is going up and to the right.
  28. 2:03 The hard question is no longer how to build it, it's should we build it?
  29. 2:08 Agreeing on what to build is the new bottleneck.
  30. 2:11 So everyone on your team needs to be involved in asking, are we making the right thing?
  31. 2:15 Are we spending our energy in the right place?
  32. 2:17 And how do we have the most impact?
  33. 2:20 When production is cheap, opportunity cost becomes the real cost.
  34. 2:25 You can't build everything and whatever you pick comes at the cost of everything else.
  35. 2:30 Anyone who ships software on a team knows that this isn't a new problem.
  36. 2:34 Alignment has always been a bottleneck, but agents have made the cost of not being aligned as a team much, much higher.
  37. 2:42 What makes it worse is that all our coordination tools are still from another era.
  38. 2:47 So GitHub, Slack, Jira, Linear, and the like, as they currently stand, are not designed for the agentic development world.
  39. 2:55 We are funneling masses of agentic outputs into platforms that were built for an outdated way of building software.
  40. 3:03 I work at GitHub, so that might sound heretical for me to say, but I promise it's not controversial.
  41. 3:09 There are very few people internally who believe that the PR and the issue are the future of software development, and there are lots of us inside the machine trying to explore what comes next.
  42. 3:18 So this is how the development process used to look.
  43. 3:21 We had a planning phase, a building phase, and a review phase.
  44. 3:24 And we had all of these touch points of alignment along the way.
  45. 3:28 And it was slow enough that we had time for conversations in Slack and Zoom meetings, comments on issues and draft PRs so you could discuss the details.
  46. 3:35 And everyone could give their two cents and get advice from expertise across your team and seniors and catch mistakes and course correct if things were going wrong.
  47. 3:44 But by the time the code was reviewed and merged, the whole team had seen the work happening, and they were roughly on the same page.
  48. 3:52 But that implementation window has now collapsed.
  49. 3:54 And because implementation is no longer as expensive and time consuming, we think we don't need to plan as much.
  50. 4:01 So most of those early touch points actually disappear.

Open at this second