vidtheque.

read-only demo

Videos am_oeAoUhew

Harness Engineering: How to Build Software When Humans Steer, Agents Execute — Ryan Lopopolo, OpenAI

index_state ready data_status ok

AI Engineer· published 2026-04-17· 0:46:20· en-US· indexed 2026-08-10 19:44

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:09, 1 of 1 keyframes kept
  2. Shot 1, 0:09 to 0:15, 1 of 1 keyframes kept
  3. Shot 2, 0:15 to 0:32, 1 of 1 keyframes kept
  4. Shot 3, 0:32 to 0:34, 1 of 1 keyframes kept
  5. Shot 4, 0:34 to 0:36, 1 of 1 keyframes kept
  6. Shot 5, 0:36 to 0:41, 1 of 1 keyframes kept
  7. Shot 6, 0:41 to 0:42, 1 of 1 keyframes kept
  8. Shot 7, 0:42 to 0:46, 1 of 1 keyframes kept
  9. Shot 8, 0:46 to 1:12, 1 of 1 keyframes kept
  10. Shot 9, 1:12 to 1:38, 0 of 1 keyframes kept
  11. Shot 10, 1:38 to 1:49, 1 of 1 keyframes kept
  12. Shot 11, 1:49 to 2:24, 1 of 1 keyframes kept
  13. Shot 12, 2:24 to 3:14, 1 of 1 keyframes kept
  14. Shot 13, 3:14 to 3:42, 1 of 1 keyframes kept
  15. Shot 14, 3:42 to 4:09, 1 of 1 keyframes kept
  16. Shot 15, 4:09 to 4:40, 1 of 1 keyframes kept
  17. Shot 16, 4:40 to 5:11, 1 of 1 keyframes kept
  18. Shot 17, 5:11 to 5:40, 1 of 1 keyframes kept
  19. Shot 18, 5:40 to 6:08, 1 of 1 keyframes kept
  20. Shot 19, 6:08 to 6:37, 1 of 1 keyframes kept
  21. Shot 20, 6:37 to 7:02, 1 of 1 keyframes kept
  22. Shot 21, 7:02 to 7:28, 1 of 1 keyframes kept
  23. Shot 22, 7:28 to 7:53, 1 of 1 keyframes kept
  24. Shot 23, 7:53 to 8:18, 1 of 1 keyframes kept
  25. Shot 24, 8:18 to 8:25, 1 of 1 keyframes kept
  26. Shot 25, 8:25 to 8:56, 1 of 1 keyframes kept
  27. Shot 26, 8:56 to 9:26, 1 of 1 keyframes kept
  28. Shot 27, 9:26 to 10:00, 1 of 1 keyframes kept
  29. Shot 28, 10:00 to 10:34, 1 of 1 keyframes kept
  30. Shot 29, 10:34 to 11:06, 1 of 1 keyframes kept
  31. Shot 30, 11:06 to 11:39, 1 of 1 keyframes kept
  32. Shot 31, 11:39 to 12:11, 1 of 1 keyframes kept
  33. Shot 32, 12:11 to 12:57, 1 of 1 keyframes kept
  34. Shot 33, 12:57 to 13:25, 0 of 1 keyframes kept
  35. Shot 34, 13:25 to 13:53, 1 of 1 keyframes kept
  36. Shot 35, 13:53 to 14:18, 1 of 1 keyframes kept
  37. Shot 36, 14:18 to 14:44, 1 of 1 keyframes kept
  38. Shot 37, 14:44 to 15:15, 1 of 1 keyframes kept
  39. Shot 38, 15:15 to 15:46, 1 of 1 keyframes kept
  40. Shot 39, 15:46 to 16:17, 1 of 1 keyframes kept
  41. Shot 40, 16:17 to 16:48, 1 of 1 keyframes kept
  42. Shot 41, 16:48 to 17:23, 1 of 1 keyframes kept
  43. Shot 42, 17:23 to 17:59, 1 of 1 keyframes kept
  44. Shot 43, 17:59 to 18:34, 1 of 1 keyframes kept
  45. Shot 44, 18:34 to 18:35, 1 of 1 keyframes kept
  46. Shot 45, 18:35 to 19:02, 1 of 1 keyframes kept
  47. Shot 46, 19:02 to 19:29, 0 of 1 keyframes kept
  48. Shot 47, 19:29 to 19:56, 1 of 1 keyframes kept
  49. Shot 48, 19:56 to 20:25, 1 of 1 keyframes kept
  50. Shot 49, 20:25 to 20:54, 0 of 1 keyframes kept
  51. Shot 50, 20:54 to 21:20, 1 of 1 keyframes kept
  52. Shot 51, 21:20 to 21:45, 1 of 1 keyframes kept
  53. Shot 52, 21:45 to 22:11, 0 of 1 keyframes kept
  54. Shot 53, 22:11 to 22:36, 1 of 1 keyframes kept
  55. Shot 54, 22:36 to 23:01, 0 of 1 keyframes kept
  56. Shot 55, 23:01 to 23:27, 0 of 1 keyframes kept
  57. Shot 56, 23:27 to 23:52, 0 of 1 keyframes kept
  58. Shot 57, 23:52 to 24:18, 1 of 1 keyframes kept
  59. Shot 58, 24:18 to 24:43, 1 of 1 keyframes kept
  60. Shot 59, 24:43 to 25:08, 0 of 1 keyframes kept
  61. Shot 60, 25:08 to 25:33, 1 of 1 keyframes kept
  62. Shot 61, 25:33 to 25:59, 0 of 1 keyframes kept
  63. Shot 62, 25:59 to 26:24, 0 of 1 keyframes kept
  64. Shot 63, 26:24 to 26:49, 1 of 1 keyframes kept
  65. Shot 64, 26:49 to 27:15, 1 of 1 keyframes kept
  66. Shot 65, 27:15 to 27:40, 0 of 1 keyframes kept
  67. Shot 66, 27:40 to 28:05, 0 of 1 keyframes kept
  68. Shot 67, 28:05 to 28:30, 1 of 1 keyframes kept
  69. Shot 68, 28:30 to 28:56, 1 of 1 keyframes kept
  70. Shot 69, 28:56 to 29:21, 1 of 1 keyframes kept
  71. Shot 70, 29:21 to 29:46, 0 of 1 keyframes kept
  72. Shot 71, 29:46 to 30:11, 0 of 1 keyframes kept
  73. Shot 72, 30:11 to 30:37, 1 of 1 keyframes kept
  74. Shot 73, 30:37 to 31:02, 0 of 1 keyframes kept
  75. Shot 74, 31:02 to 31:27, 1 of 1 keyframes kept
  76. Shot 75, 31:27 to 31:53, 1 of 1 keyframes kept
  77. Shot 76, 31:53 to 32:18, 0 of 1 keyframes kept
  78. Shot 77, 32:18 to 32:43, 1 of 1 keyframes kept
  79. Shot 78, 32:43 to 33:08, 1 of 1 keyframes kept
  80. Shot 79, 33:08 to 33:34, 0 of 1 keyframes kept
  81. Shot 80, 33:34 to 33:59, 1 of 1 keyframes kept
  82. Shot 81, 33:59 to 34:24, 0 of 1 keyframes kept
  83. Shot 82, 34:24 to 34:49, 0 of 1 keyframes kept
  84. Shot 83, 34:49 to 35:15, 1 of 1 keyframes kept
  85. Shot 84, 35:15 to 35:40, 1 of 1 keyframes kept
  86. Shot 85, 35:40 to 36:05, 0 of 1 keyframes kept
  87. Shot 86, 36:05 to 36:31, 1 of 1 keyframes kept
  88. Shot 87, 36:31 to 36:56, 1 of 1 keyframes kept
  89. Shot 88, 36:56 to 37:21, 1 of 1 keyframes kept
  90. Shot 89, 37:21 to 37:46, 0 of 1 keyframes kept
  91. Shot 90, 37:46 to 38:12, 0 of 1 keyframes kept
  92. Shot 91, 38:12 to 38:37, 1 of 1 keyframes kept
  93. Shot 92, 38:37 to 39:02, 0 of 1 keyframes kept
  94. Shot 93, 39:02 to 39:27, 1 of 1 keyframes kept
  95. Shot 94, 39:27 to 39:53, 1 of 1 keyframes kept
  96. Shot 95, 39:53 to 40:18, 0 of 1 keyframes kept
  97. Shot 96, 40:18 to 40:43, 0 of 1 keyframes kept
  98. Shot 97, 40:43 to 41:09, 1 of 1 keyframes kept
  99. Shot 98, 41:09 to 41:34, 1 of 1 keyframes kept
  100. Shot 99, 41:34 to 41:59, 1 of 1 keyframes kept
  101. Shot 100, 41:59 to 42:26, 1 of 1 keyframes kept
  102. Shot 101, 42:26 to 42:53, 1 of 1 keyframes kept
  103. Shot 102, 42:53 to 43:20, 1 of 1 keyframes kept
  104. Shot 103, 43:20 to 43:48, 1 of 1 keyframes kept
  105. Shot 104, 43:48 to 44:15, 1 of 1 keyframes kept
  106. Shot 105, 44:15 to 44:43, 1 of 1 keyframes kept
  107. Shot 106, 44:43 to 45:10, 0 of 1 keyframes kept
  108. Shot 107, 45:10 to 45:38, 1 of 1 keyframes kept
  109. Shot 108, 45:38 to 46:05, 1 of 1 keyframes kept
  110. Shot 109, 46:05 to 46:20, 1 of 1 keyframes kept

110 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
412
whisperx 412
chunks
81
from 412 cues
keyframes
83
kept of 110 captured
frames with text
83
962 lines read
chapters
0
from the source metadata
keyframe bytes
12.2 MB
word timings on 412 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 00:47 1m 34s
stt done — 2026-08-10 00:49 48s
chunk done — 2026-08-10 00:49 0s
text_embed done — 2026-08-10 19:44 1s
keyframe done — 2026-08-10 00:49 2m 26s
ocr done — 2026-08-10 00:52 22s
frame_embed done — 2026-08-10 19:44 14s

Frames, and what the machine read

  • 0:08 #0 done2 line(s)

    shot 0·sharpness 832.2

    1. PRESENTING SPONSOR0.98
    2. Google DeepMind1.00

  • 0:12 #1 done3 line(s)

    shot 1·sharpness 914.1

    1. PLATINUMSPONSORS1.00
    2. # Braintrust0.96
    3. WorkOS OpenAI0.96

  • 0:30 #2 done5 line(s)

    shot 2·sharpness 470.9

    1. AlEngineer0.97
    2. OpenAl0.94
    3. AlEngineer0.96
    4. EUROPE1.00
    5. EUROPE1.00

  • 0:33 #3 done4 line(s)

    shot 3·sharpness 926.9

    1. RYAN1.00
    2. LOPOPOLO1.00
    3. MENBER OF TECHICE STAF0.87
    4. OpenAl0.92

  • 0:35 #4 done5 line(s)

    shot 4·sharpness 800.0

    1. RYAN1.00
    2. AIE0.98
    3. LOPOPOLO1.00
    4. MEMBER OF TECHNICAL STAFF1.00
    5. AIE0.99

  • 0:36 #5 done9 line(s)

    shot 5·sharpness 643.7

    1. RYAN1.00
    2. AELOPOLO0.94
    3. AIE0.99
    4. LOPOPOLO1.00
    5. MEMBER OF TECHNICAL STAFF1.00
    6. Or0.81
    7. AIE1.00
    8. AlEngineer0.96
    9. Google DepMind0.98

  • 0:42 #6 done3 line(s)

    shot 6·sharpness 437.4

    1. MEMBER1.00
    2. TECHNICAL STAFF1.00
    3. enAl0.97

  • 0:44 #7 done35 line(s)

    shot 7·sharpness 1249.5

    1. EUROPE1.00
    2. therai1.00
    3. AlEngineer0.98
    4. Al0.79
    5. EUROPE1.00
    6. neer1.00
    7. Braintrust1.00
    8. AlEngineer0.97
    9. le DeepMind0.96
    10. AlEngineer0.98
    11. EUROPE1.00
    12. EUROPE1.00
    13. ize1.00
    14. AlEngineer0.98
    15. Micros1.00
    16. gineer1.00
    17. neo4j0.99
    18. All0.88
    19. EUROPE1.00
    20. ROPE1.00
    21. neer1.00
    22. Trigger.dev1.00
    23. AlEngineer0.96
    24. VorkOS0.95
    25. AlEngineer0.99
    26. EUROPE1.00
    27. EUROPE1.00
    28. lodal1.00
    29. AlEngir0.98
    30. tailscale1.00
    31. AlEngineer0.98
    32. Sonar1.00
    33. All0.90
    34. EUROPE1.00
    35. EUROPE0.99

  • 0:56 #8 done28 line(s)

    shot 8·sharpness 2942.6

    1. EUROPE1.00
    2. OpenAl0.95
    3. Harness Engineering:1.00
    4. ineer1.00
    5. qle DeepMin0.96
    6. OPE0.99
    7. How to Build1.00
    8. cro0.96
    9. ngineer1.00
    10. Software When1.00
    11. EUROPE0.97
    12. Humans Steer and0.99
    13. gineer1.00
    14. WorkOS1.00
    15. OPE1.00
    16. Agents Execute0.97
    17. Iscale0.99
    18. IEngineer0.97
    19. EUROP0.95
    20. Ryan Lopopolo1.00
    21. Member of Technical Staff1.00
    22. AlEngineer0.98
    23. INEERING: HOW TO BUILD SOFTWARE WHEN H0.98
    24. EUROPE1.00
    25. Google DeepMind1.00
    26. PRESENTED BY1.00
    27. RYAN LOPOPOLO / Member of Technical Staff0.99
    28. OpenAl0.96

  • 1:15 #9 skipped

    shot 9·duplicate of #8

  • 1:48 #10 done24 line(s)

    shot 10·sharpness 1891.8

    1. EUROPE1.00
    2. The way we build1.00
    3. eer1.00
    4. DeepMind1.00
    5. software has1.00
    6. osol0.94
    7. gi..eer0.85
    8. changed.1.00
    9. ROPE1.00
    10. eer1.00
    11. orkOS0.98
    12. cale1.00
    13. Engineer1.00
    14. EUROPE1.00
    15. OpenAl0.97
    16. Confidential and proprietary.1.00
    17. AlEngineer0.99
    18. S EXECUTE0.98
    19. HARNESS ENGINEERING: HOW T0.97
    20. EUROPE1.00
    21. PRESENTED BY1.00
    22. OpenAl0.94
    23. Google DeepMind1.00
    24. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 2:07 #11 done20 line(s)

    shot 11·sharpness 1880.7

    1. gether.ai0.98
    2. The way we build1.00
    3. gineer1.00
    4. UROPE1.00
    5. software has1.00
    6. changed.1.00
    7. ENTI0.99
    8. osoft0.87
    9. ngineer0.99
    10. UROPE1.00
    11. NCORD1.00
    12. OpenAl0.97
    13. Confidential and proprietary.1.00
    14. AlEngineer0.98
    15. RE WHEN HUMANS STEER AND AGENTS EXECU0.99
    16. EUROPE1.00
    17. PRESENTED BY1.00
    18. OpenAl0.95
    19. Google DeepMind1.00
    20. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 2:30 #12 done18 line(s)

    shot 12·sharpness 1808.6

    1. EUROPE1.00
    2. The way we build1.00
    3. Go0.95
    4. neepMind0.94
    5. software has1.00
    6. t0.99
    7. changed.1.00
    8. Eny0.96
    9. ·EURO0.90
    10. OpenAl0.93
    11. Confidential and proprietary.1.00
    12. AlEngineer0.98
    13. W TO BUILD SOFTWARE WHEN HUMANS STEER0.98
    14. EUROPE1.00
    15. PRESENTED BY0.99
    16. OpenAl0.94
    17. Google DeepMind1.00
    18. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 3:36 #13 done21 line(s)

    shot 13·sharpness 2051.4

    1. EUROPE1.00
    2. orkOS0.96
    3. The models are0.97
    4. Code is free.0.99
    5. Your role is to1.00
    6. good enough.1.00
    7. unblock your1.00
    8. PE1.00
    9. team.1.00
    10. trust1.00
    11. ale1.00
    12. neer0.99
    13. OpenAl0.90
    14. Confidential and proprietary.1.00
    15. AlEngineer0.98
    16. WARE WHEN HUMANS STEER AND AGENTS EXI1.00
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. OpenAl0.97
    20. Google DeepMind1.00
    21. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 3:45 #14 done23 line(s)

    shot 14·sharpness 2004.7

    1. EUROPE1.00
    2. -kOS0.88
    3. The models are0.99
    4. Code is free.0.99
    5. Your role is to1.00
    6. el0.53
    7. good enough.1.00
    8. unblock your1.00
    9. team.1.00
    10. ust1.00
    11. e1.00
    12. ngi0.94
    13. EURO1.00
    14. OpenAl0.94
    15. Confidential and proprietary.0.99
    16. AlEngineer0.99
    17. ND AGENTS EXECUTE1.00
    18. HARNESS ENGINEERI1.00
    19. EUROPE1.00
    20. PRESENTED BY1.00
    21. OpenAl0.94
    22. Google DeepMind1.00
    23. RYAN LOPOPOLO / Member of Technical Staff0.99

  • 4:16 #15 done33 line(s)

    shot 15·sharpness 2061.9

    1. AEngineer0.98
    2. Braintru1.00
    3. EUROPE1.00
    4. OpenAl0.89
    5. AlEngineer0.96
    6. EUROPE1.00
    7. The models are0.97
    8. Code is free.0.98
    9. Your role is to0.99
    10. good enough.1.00
    11. unblock your1.00
    12. AlEngin0.93
    13. orki0.85
    14. team.1.00
    15. EUROPE1.00
    16. SEN1.00
    17. AlEngineer0.98
    18. EUROPE1.00
    19. Sond0.90
    20. AlEngineer0.98
    21. EUROPE1.00
    22. OpenAl0.97
    23. Confidential and proprietary.1.00
    24. 31.00
    25. AlEngineer0.99
    26. AGENTS EXECUTE1.00
    27. HARNESS ENGINEERING1.00
    28. EUROPE1.00
    29. PRESENTED BY1.00
    30. Google DeepMind1.00
    31. # Braintrust0.93
    32. WorkOS1.00
    33. OpenAl0.94

  • 4:44 #16 done19 line(s)

    shot 16·sharpness 1681.2

    1. Braintrust1.00
    2. AI0.90
    3. The models are1.00
    4. Code is free.0.99
    5. Your role is to1.00
    6. good enough.1.00
    7. unblock your1.00
    8. team.1.00
    9. eer1.00
    10. TRY1.00
    11. ineer1.00
    12. Sonar1.00
    13. OpenAl0.96
    14. Confidential and proprietary.0.99
    15. R AND AGENTS EXECUTE0.99
    16. HARNESS ENGINE0.97
    17. # Braintrust0.95
    18. WorkOS1.00
    19. OpenAl0.94

  • 5:34 #17 done16 line(s)

    shot 17·sharpness 1382.7

    1. EUROPE0.99
    2. intrust0.92
    3. Human time.1.00
    4. Human and1.00
    5. Model context1.00
    6. eer1.00
    7. model attention.0.97
    8. window.1.00
    9. soft1.00
    10. 10.82
    11. En1.00
    12. =EUR0.87
    13. OpenAl0.93
    14. Confidential and proprietary.1.00
    15. Engineering the future of Al0.99
    16. AlEngineer0.98

  • 5:57 #18 done21 line(s)

    shot 18·sharpness 1983.3

    1. EUROPE0.99
    2. -intrust0.98
    3. Human time.1.00
    4. Human and1.00
    5. Model context0.99
    6. model attention.1.00
    7. window.1.00
    8. osoft1.00
    9. 10.71
    10. VE.0.61
    11. eer0.99
    12. EL0.97
    13. OpenAl0.94
    14. Confidential and proprietary.1.00
    15. AlEngineer0.98
    16. OFTWARE WHEN HUMANS STEER AND AGENTS1.00
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. OpenAl0.97
    20. Google DeepMind1.00
    21. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 6:28 #19 done24 line(s)

    shot 19·sharpness 2038.0

    1. EUROPE0.96
    2. er1.00
    3. Praintrust0.96
    4. Human time.1.00
    5. Human and1.00
    6. Model context0.99
    7. Engineer1.00
    8. model attention.1.00
    9. window.1.00
    10. EUROPE1.00
    11. icrosoft1.00
    12. er1.00
    13. dal1.00
    14. gineer1.00
    15. IOPE0.93
    16. OpenAl0.90
    17. Confidential and proprietary.1.00
    18. AlEngineer0.99
    19. WARE WHEN HUMANS STEER AND AGENTS EXE1.00
    20. EUROPE1.00
    21. Google DeepMind1.00
    22. PRESENTED BY1.00
    23. RYAN LOPOPOLO / Member of Technical Staff0.98
    24. OpenAl0.96

  • 6:59 #20 done25 line(s)

    shot 20·sharpness 2069.0

    1. AlEngineer0.99
    2. Google De0.97
    3. EUROPE0.93
    4. AlEngin0.96
    5. EUROPE0.99
    6. Human time.1.00
    7. Human and0.99
    8. Model context0.99
    9. model attention.1.00
    10. window.1.00
    11. ze0.72
    12. AlEngin0.98
    13. EUROPE1.00
    14. :neo0.85
    15. AIEn0.87
    16. E0.64
    17. OpenAl0.98
    18. Confidential and proprietary.1.00
    19. AlEngineer0.99
    20. RE WHEN HUMANS STEER AND AGENTS EXECUT1.00
    21. EUROPE1.00
    22. PRESENTED BY1.00
    23. OpenAl0.93
    24. Google DeepMind0.99
    25. RYAN LOPOPOLO / Member of Technical Staff0.99

  • 7:22 #21 done28 line(s)

    shot 21·sharpness 2107.1

    1. engineer0.99
    2. GoogleDeepMi1.00
    3. EUROPE1.00
    4. gether-0.90
    5. AIEngineer0.83
    6. BOPE0.86
    7. Human time.1.00
    8. Human and0.99
    9. Model context0.99
    10. model attention.1.00
    11. window.1.00
    12. ed1.00
    13. Engin1.00
    14. EUROPE0.93
    15. ariz1.00
    16. EUROPE1.00
    17. neo4j0.96
    18. Engineer1.00
    19. EUROPE0.99
    20. OpenAl0.93
    21. Confidential and proprietary.0.99
    22. AlEngineer0.98
    23. OW TO BUILD SOFTWARE WHEN HUMANS STE0.99
    24. EUROPE1.00
    25. PRESENTED BY1.00
    26. OpenAl0.96
    27. Google DeepMind1.00
    28. RYAN LOPOPOLO / Member of Technical Staff0.98

  • 7:48 #22 done24 line(s)

    shot 22·sharpness 2085.1

    1. engineer0.95
    2. Google DeepMi0.97
    3. EUROPE1.00
    4. geterai0.80
    5. ineer1.00
    6. Human time.1.00
    7. Human and1.00
    8. Model context0.99
    9. model attention.1.00
    10. window.1.00
    11. Engine1.00
    12. EUROPE0.98
    13. arize1.00
    14. Engineer1.00
    15. EUROPE1.00
    16. OpenAl0.94
    17. Confidential and proprietary.1.00
    18. AlEngineer0.98
    19. ENGINEERING: HOW TO BUILD SOFTWARE WHE1.00
    20. EUROPE1.00
    21. PRESENTED BY1.00
    22. OpenAl0.95
    23. Google DeepMind1.00
    24. RYAN LOPOPOLO / Member of Technical Staff0.99

  • 7:56 #23 done21 line(s)

    shot 23·sharpness 2100.1

    1. Google DeepMind0.99
    2. AlEngineer0.99
    3. EUROPE1.00
    4. Human time.1.00
    5. Human and1.00
    6. Model context1.00
    7. model attention.0.99
    8. window.1.00
    9. Zed0.99
    10. AlEngineer0.98
    11. EUROPE1.00
    12. :neo4j0.89
    13. OpenAl0.96
    14. Confidential and proprietary.1.00
    15. AlEngineer0.99
    16. D SOFTWARE WHEN HUMANS STEER AND AGEN0.98
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. OpenAl0.97
    20. Google DeepMind1.00
    21. RYAN LOPOPOLO / Member of Technical Staff0.98

Transcript

412 cues· 7,692 words· 41,579 chars

  1. 0:15 speaker is here to speak about harness engineering, how to build software when humans steer and agents execute.
  2. 0:24 Please join me in welcoming to the stage, member of technical staff at OpenAI, Ryan Lepopolo.
  3. 0:40 Good morning, London.
  4. 0:46 I'm super excited to be here today.
  5. 0:47 I'm Ryan Lopopolo, and for the last nine months, I had the privilege of building software exclusively with agents.
  6. 0:56 I am a token billionaire, and I believe that in order for us to get into our AGI future, we want everybody to be token billionaires, to use the models to do the full job.
  7. 1:08 And what that means is
  8. 1:12 to lean into the idea that the models are capable of being a full software engineer.
  9. 1:17 And I've lived that experience by banning my team from even touching their editors to have to work through the models in order to get the job done.
  10. 1:25 And today I'm going to talk to you a little bit about what it means to lean into that and operationalize the way you work, the code spaces you live in, and the processes on your teams in order to get the agents to do the full job.
  11. 1:40 I believe I'm preaching to the choir here when I say that the way we build software has changed.
  12. 1:45 In the last six months, we have seen coding agents take over the world and capability has continually advanced at a super fast pace to have these models and the harnesses within which they live take more complex actions, do more complicated work with higher reliability over longer time horizons.
  13. 2:06 And the place we've gotten to here is that implementation is no longer the scarce resource of what it means to do the job of software engineering.
  14. 2:15 Code is free.
  15. 2:16 We have an abundance of code to solve the problems that we come across in our day to day as we run our teams, build software, and solve user problems.
  16. 2:27 Hiring the hands on the keyboards as part of our teams is only constrained by GPU capacity and token budgets.
  17. 2:35 And each engineer today in this room has access to five, 50, or 5,000 engineers worth of capacity 24-7 every day of the year.
  18. 2:47 The only thing that needs to happen, our roles, is to figure out how to productively deploy these resources into our code and into our teams to make use of this new capacity.
  19. 2:59 In this world, skill sets are shifting more towards systems thinking, system design, and delegation in order to make use of this abundant capacity to produce code to solve problems.
  20. 3:12 And there are three reasons that this happened, all of which happened in late 2025.
  21. 3:19 For me, the magic moment was GPT 5.2, which, when it came out, was able to do the full job of a software engineer.
  22. 3:26 The models at this point are good enough where they're isomorphic to you and I in terms of the ability to produce code at high quality that solve real user problems in real code bases.
  23. 3:39 Code is free, and I know this is maybe a scary thing to hear because code carries maintenance burden, but it's free to produce, free to refactor, and it is not a thing to get hung up on anymore.
  24. 3:54 We think of code as burden because it's a synchronous attention drain on the human engineers on our team.
  25. 4:01 But the models are incredibly patient.
  26. 4:04 They are infinitely parallel.
  27. 4:05 So the ability to produce, maintain, refactor, and delete code is no longer a forcing function on figuring out how to allocate resources on your engineering teams.
  28. 4:16 So sort of be AGI pilled here is to believe that the models are capable of producing every line of code we could ever possibly need, figuring out when to delete them, figuring out when to refactor them or make them more reliable.
  29. 4:30 And it's your role as software engineers to figure out how to unblock your team of agents and humans driving those agents from being able to drive them over long horizon work to do the full job.
  30. 4:44 The idea here is that every one of you is a staff engineer.
  31. 4:48 You have as many team members as you can possibly drive concurrently and have tokens to support.
  32. 4:55 And you need to look one day, one week, six months into the future to figure out what structures you need to put in place to productively harness this infinite capacity to produce code.
  33. 5:10 The scarce resources in this world that we see today are three things, human time, human and model attention, and model context window.
  34. 5:22 And in the world where human time and attention is scarce, the role is to think about where that time is going, figure out ways to productively automate it, and move that synchronous human time into higher leverage activities.
  35. 5:40 In a world where human time is scarce and human time is required to produce code, we have a stack rank.
  36. 5:48 Things are either P0s or P2s.
  37. 5:50 Those P3s will never get done.
  38. 5:53 However, in a world where code is free and infinitely abundant, all those P3s get kicked off immediately, maybe 4x in parallel.
  39. 6:02 We pick one that solves the problem, and in it goes.
  40. 6:07 I've had the privilege of building a ton of agents internally at OpenAI to improve the productivity of my coworkers.
  41. 6:15 When code is free, all these internal tools can have good localization and internationalization from day one.
  42. 6:23 I can make tools that my colleagues in London, Dublin, Paris, Brussels, Zurich, and Munich are able to experience in their native languages without really having to trade against any of my other team's capacity in order to make high quality tools.
  43. 6:40 We should be working with the assumption that the best parts of software engineering that we all know, live, and breathe are available in any product that we could ever build all the time.
  44. 6:52 Humans no longer need to concern themselves with implementation.
  45. 6:56 The important thing is not the code, but the prompt and the guardrails that got you there.
  46. 7:01 This is why leaving breadcrumbs, documentation, ADRs, persona-oriented documentation around what a good job looks like, all the historical logs of tickets and code reviews, this is the process that got you and your teams to the code and products that you have today.
  47. 7:17 And this is what needs to happen in order to get your agents there as well.
  48. 7:23 Your job is to build systems, software, and structures that enable your team to be successful.
  49. 7:29 And to do that, we need to make them legible to those agents that are driving the implementation.
  50. 7:36 That means structuring them in a way that's native to the agents,

Open at this second