vidtheque.

read-only demo

Videos _Zcw_sVF6hU

The Friction is Your Judgment — Armin Ronacher & Cristina Poncela Cubeiro, Earendil

index_state ready data_status ok

AI Engineer· published 2026-04-18· 0:18:37· en-US· indexed 2026-08-10 19:52

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 0:17, 1 of 1 keyframes kept
  5. Shot 4, 0:17 to 0:18, 1 of 1 keyframes kept
  6. Shot 5, 0:18 to 0:24, 1 of 1 keyframes kept
  7. Shot 6, 0:24 to 0:57, 1 of 1 keyframes kept
  8. Shot 7, 0:57 to 1:23, 1 of 1 keyframes kept
  9. Shot 8, 1:23 to 1:49, 0 of 1 keyframes kept
  10. Shot 9, 1:49 to 2:17, 1 of 1 keyframes kept
  11. Shot 10, 2:17 to 2:54, 1 of 1 keyframes kept
  12. Shot 11, 2:54 to 3:31, 1 of 1 keyframes kept
  13. Shot 12, 3:31 to 3:33, 1 of 1 keyframes kept
  14. Shot 13, 3:33 to 3:54, 1 of 1 keyframes kept
  15. Shot 14, 3:54 to 4:06, 1 of 1 keyframes kept
  16. Shot 15, 4:06 to 4:35, 1 of 1 keyframes kept
  17. Shot 16, 4:35 to 4:56, 1 of 1 keyframes kept
  18. Shot 17, 4:56 to 5:28, 1 of 1 keyframes kept
  19. Shot 18, 5:28 to 5:30, 1 of 1 keyframes kept
  20. Shot 19, 5:30 to 6:09, 1 of 1 keyframes kept
  21. Shot 20, 6:09 to 6:24, 1 of 1 keyframes kept
  22. Shot 21, 6:24 to 6:54, 0 of 1 keyframes kept
  23. Shot 22, 6:54 to 6:58, 1 of 1 keyframes kept
  24. Shot 23, 6:58 to 7:11, 1 of 1 keyframes kept
  25. Shot 24, 7:11 to 7:22, 1 of 1 keyframes kept
  26. Shot 25, 7:22 to 7:40, 1 of 1 keyframes kept
  27. Shot 26, 7:40 to 8:03, 1 of 1 keyframes kept
  28. Shot 27, 8:03 to 8:28, 1 of 1 keyframes kept
  29. Shot 28, 8:28 to 8:39, 1 of 1 keyframes kept
  30. Shot 29, 8:39 to 9:10, 1 of 1 keyframes kept
  31. Shot 30, 9:10 to 9:13, 1 of 1 keyframes kept
  32. Shot 31, 9:13 to 9:46, 1 of 1 keyframes kept
  33. Shot 32, 9:46 to 10:16, 1 of 1 keyframes kept
  34. Shot 33, 10:16 to 10:33, 1 of 1 keyframes kept
  35. Shot 34, 10:33 to 10:50, 1 of 1 keyframes kept
  36. Shot 35, 10:50 to 10:52, 1 of 1 keyframes kept
  37. Shot 36, 10:52 to 11:26, 1 of 1 keyframes kept
  38. Shot 37, 11:26 to 12:07, 1 of 1 keyframes kept
  39. Shot 38, 12:07 to 12:45, 1 of 1 keyframes kept
  40. Shot 39, 12:45 to 12:54, 1 of 1 keyframes kept
  41. Shot 40, 12:54 to 13:08, 1 of 1 keyframes kept
  42. Shot 41, 13:08 to 13:58, 0 of 1 keyframes kept
  43. Shot 42, 13:58 to 14:09, 1 of 1 keyframes kept
  44. Shot 43, 14:09 to 14:10, 1 of 1 keyframes kept
  45. Shot 44, 14:10 to 14:12, 1 of 1 keyframes kept
  46. Shot 45, 14:12 to 14:39, 1 of 1 keyframes kept
  47. Shot 46, 14:39 to 15:05, 1 of 1 keyframes kept
  48. Shot 47, 15:05 to 15:15, 1 of 1 keyframes kept
  49. Shot 48, 15:15 to 15:21, 1 of 1 keyframes kept
  50. Shot 49, 15:21 to 15:25, 1 of 1 keyframes kept
  51. Shot 50, 15:25 to 15:46, 1 of 1 keyframes kept
  52. Shot 51, 15:46 to 16:00, 1 of 1 keyframes kept
  53. Shot 52, 16:00 to 16:20, 1 of 1 keyframes kept
  54. Shot 53, 16:20 to 16:27, 1 of 1 keyframes kept
  55. Shot 54, 16:27 to 16:47, 1 of 1 keyframes kept
  56. Shot 55, 16:47 to 17:21, 1 of 1 keyframes kept
  57. Shot 56, 17:21 to 17:56, 1 of 1 keyframes kept
  58. Shot 57, 17:56 to 18:21, 1 of 1 keyframes kept
  59. Shot 58, 18:21 to 18:37, 1 of 1 keyframes kept

59 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
223
whisperx 223
chunks
33
from 223 cues
keyframes
56
kept of 59 captured
frames with text
56
1,265 lines read
chapters
7
from the source metadata
keyframe bytes
7.3 MB
word timings on 223 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 17:08 1m 38s
stt done — 2026-08-10 17:09 29s
chunk done — 2026-08-10 17:10 0s
text_embed done — 2026-08-10 19:52 1s
keyframe done — 2026-08-10 17:10 1m 44s
ocr done — 2026-08-10 17:11 28s
frame_embed done — 2026-08-10 19:52 9s

Frames, and what the machine read

  • 0:03 #0 done2 line(s)

    shot 0·sharpness 658.4

    1. AlEngineer0.98
    2. EUROPE1.00

  • 0:08 #1 done2 line(s)

    shot 1·sharpness 827.9

    1. PRESENTINGSPONSOR1.00
    2. Google DeepMind1.00

  • 0:13 #2 done3 line(s)

    shot 2·sharpness 907.6

    1. PLATINUM SPONSORS0.98
    2. # Braintrust0.96
    3. WorkOS OpenAI0.95

  • 0:15 #3 done10 line(s)

    shot 3·sharpness 616.6

    1. HER1.00
    2. CUBEIRO1.00
    3. FOUNDER1.00
    4. SOFTWARE ENGINEER1.00
    5. ENDIL1.00
    6. IL1.00
    7. AlEngineer0.97
    8. EUROPE0.99
    9. PRESENTED BY1.00
    10. Google DeepMind0.97

  • 0:18 #4 done34 line(s)

    shot 4·sharpness 1136.8

    1. R0.82
    2. meel0.71
    3. Unblocked1.00
    4. AlEngineer0.99
    5. Goo!0.96
    6. OPE0.86
    7. L0.76
    8. NDER1.00
    9. AAEEggineer0.97
    10. WonkOs0.72
    11. LEUROPE0.98
    12. 70.99
    13. epMliadl0.61
    14. AAEEggineer0.91
    15. QpenAlI0.52
    16. CLBUROPE0.86
    17. L0.73
    18. mw4j0.71
    19. SENTRY0.60
    20. AAEBggineer0.88
    21. 」0.57
    22. I0.66
    23. AlEngineer1.00
    24. TEesssl0.70
    25. neer0.90
    26. EUROPE1.00
    27. WEBgijineer0.90
    28. PRESENTED BY0.99
    29. EUROIGPE0.83
    30. Google DeepMind0.99
    31. BANCORD0.76
    32. )ssoroar0.57
    33. AEgigineer0.80
    34. FURORIPE0.74

  • 0:21 #5 done36 line(s)

    shot 5·sharpness 1181.8

    1. EURO0.90
    2. AIEngineer0.97
    3. together.ai0.96
    4. AlEngineer0.97
    5. EUROPE1.00
    6. EUROPE1.00
    7. SAFE1.00
    8. AlEngineer0.98
    9. Braintrust1.00
    10. AlEngineer0.98
    11. Google De1.00
    12. INTELLIGENCE1.00
    13. EUROPE1.00
    14. EUROPE1.00
    15. arize1.00
    16. Engineer1.00
    17. Microsoft1.00
    18. AIEngir0.90
    19. EUROPE0.99
    20. EUROPE1.00
    21. AlEngineer0.99
    22. EUROPE1.00
    23. PRESENTED BY0.97
    24. AlEngineer0.98
    25. gger.dev0.96
    26. AIEngineer0.95
    27. Wor0.93
    28. Google DeepMind0.98
    29. EUROPE1.00
    30. EUROPE1.00
    31. Modal1.00
    32. AlEngineer0.99
    33. tailscale1.00
    34. AlEngir0.93
    35. EUROPE1.00
    36. EUROP0.92

  • 0:37 #6 done35 line(s)

    shot 6·sharpness 2902.8

    1. WorkOS0.99
    2. AIEngine0.94
    3. EUROPE1.00
    4. EARENDIL1.00
    5. (web app host) "accidentally enables CDN"0.99
    6. causing massive data breaches0.99
    7. NEngi0.91
    8. Braint1.00
    9. EU0.99
    10. a1.00
    11. AlEngine0.94
    12. EUROPE1.00
    13. NEngin0.93
    14. Triggel0.96
    15. SHIP1.00
    16. EUROPE1.00
    17. WITHOUT1.00
    18. FRICTION1.00
    19. Frontend1.00
    20. Mo1.00
    21. AlEngine0.97
    22. EURODE0.91
    23. Open1.00
    24. ARMIN RONACHER & CRISTINA PONCELA0.98
    25. EARENDIL INC.0.99
    26. THE FRICTION IS YOUR JUDGMENT1.00
    27. AlEngineer0.96
    28. EUROPE1.00
    29. PRESENTED BY1.00
    30. ARMIN RONACHER1.00
    31. CRISTINA PONCELA CUBEIRO1.00
    32. Google DeepMind1.00
    33. Founder1.00
    34. Software Engineer0.97
    35. EARENDIL1.00

  • 1:08 #7 done36 line(s)

    shot 7·sharpness 2264.1

    1. WorkOS0.99
    2. AIEngin0.95
    3. EUROPE1.00
    4. EARENDIL1.00
    5. AIEngin0.93
    6. Brain1.00
    7. EURO1.00
    8. Who We Are1.00
    9. ar0.88
    10. AlEngin0.93
    11. EUROPE1.00
    12. ARMIN RONACHER1.00
    13. CRISTINA PONCELA1.00
    14. • Creator of Flask0.99
    15. • AI Engineer at Earendil0.97
    16. AIEngin0.96
    17. Trigge0.99
    18. • Founder at Earendil0.95
    19. • Previously at Bending Spoons1.00
    20. EUROPE1.00
    21. • 10 years at Sentry0.99
    22. Mo1.00
    23. AlEngin0.89
    24. EUROPE1.00
    25. ARMIN RONACHER & CRISTINA PONCELA0.97
    26. EARENDIL INC.1.00
    27. THE FRICTION IS YOUR JUDGMENT1.00
    28. AlEngineer0.98
    29. EUROPE1.00
    30. PRESENTED BY1.00
    31. ARMIN RONACHER1.00
    32. CRISTINA PONCELA CUBEIRO1.00
    33. Google DeepMind1.00
    34. Founder1.00
    35. Software Engineer0.98
    36. EARENDIL1.00

  • 1:46 #8 skipped

    shot 8·duplicate of #7

  • 1:52 #9 done26 line(s)

    shot 9·sharpness 2193.9

    1. Snork0.96
    2. EARENDIL1.00
    3. AIE1.00
    4. Who We Are1.00
    5. Micr0.99
    6. ARMIN RONACHER1.00
    7. CRISTINA PONCELA1.00
    8. • Creator of Flask0.99
    9. • AI Engineer at Earendil0.99
    10. • Founder at Earendil0.99
    11. • Previously at Bending Spoons1.00
    12. • 10 years at Sentry0.98
    13. AlEnginee0.94
    14. EUROPE0.95
    15. ARMIN RONACHER & CRISTINA PONCELA0.96
    16. EARENDIL INC.1.00
    17. THE FRICTION IS YOUR JUDGMENT1.00
    18. AlEngineer0.98
    19. EUROPE1.00
    20. PRESENTED BY1.00
    21. ARMIN RONACHER1.00
    22. CRISTINA PONCELA CUBEIRO1.00
    23. Google DeepMind1.00
    24. Founder1.00
    25. Software Engineer0.99
    26. EARENDIL1.00

  • 2:43 #10 done33 line(s)

    shot 10·sharpness 3245.9

    1. DeepMind0.97
    2. AlEngineer0.95
    3. EUROPE1.00
    4. EARENDIL1.00
    5. Engineer0.97
    6. togethe1.00
    7. EUROPE1.00
    8. From Practice, Not Theory1.00
    9. gge0.99
    10. AIEngineer0.97
    11. EUROPE1.00
    12. • We have been building with agents for over 12 months0.98
    13. • We had moments of huge leverage and moments of real mess0.99
    14. ariz1.00
    15. Engineer0.99
    16. • Two distinct classes of problems keep recurring and they need different solutions0.99
    17. EUROPE1.00
    18. • Take this as part of a learning journey, not a finished playbook0.99
    19. AIEngineer0.95
    20. CLOUDELA0.84
    21. EUROPE1.00
    22. ARMIN RONACHER & CRISTINA PONCELA0.97
    23. EARENDIL INC.0.99
    24. THE FRICTION IS YOUR JUDGMENT1.00
    25. AlEngineer0.99
    26. EUROPE1.00
    27. PRESENTED BY1.00
    28. ARMIN RONACHER1.00
    29. CRISTINA PONCELA CUBEIRO1.00
    30. Google DeepMind1.00
    31. Founder1.00
    32. Software Engineer1.00
    33. EARENDIL1.00

  • 3:02 #11 done34 line(s)

    shot 11·sharpness 3237.6

    1. DeepMind0.97
    2. AlEngineer0.99
    3. EUROPE1.00
    4. EARENDIL1.00
    5. Engin0.91
    6. togethe1.00
    7. EURO1.00
    8. From Practice, Not Theory1.00
    9. gge0.97
    10. AIEngineer0.96
    11. EUROPE1.00
    12. • We have been building with agents for over 12 months0.99
    13. • We had moments of huge leverage and moments of real mess0.99
    14. Engine0.99
    15. Λariz0.90
    16. • Two distinct classes of problems keep recurring and they need different solutions1.00
    17. EUROPE1.00
    18. • Take this as part of a learning journey, not a finished playbook0.99
    19. AlEngineer0.95
    20. LOUD1.00
    21. RF0.99
    22. EUROPE0.92
    23. ARMIN RONACHER & CRISTINA PONCELA0.97
    24. EARENDIL INC.0.99
    25. THE FRICTION IS YOUR JUDGMENT1.00
    26. AlEngineer0.98
    27. EUROPE1.00
    28. PRESENTED BY1.00
    29. ARMIN RONACHER1.00
    30. CRISTINA PONCELA CUBEIRO1.00
    31. Google DeepMind1.00
    32. Founder1.00
    33. Software Engineer1.00
    34. EARENDIL1.00

  • 3:31 #12 done13 line(s)

    shot 12·sharpness 1625.7

    1. EARENDIL1.00
    2. PART I0.99
    3. The Psychology Problem0.99
    4. ARMIN RONACHER G CRISTINA PONCELA0.96
    5. EARENDIL INC.1.00
    6. THE FRICTION IS YOUR JUDGMENT1.00
    7. AlEngineer0.98
    8. EUROPE1.00
    9. PRESENTED BY0.99
    10. Google DeepMind1.00
    11. Braintrust1.00
    12. WorkOS1.00
    13. OpenAl0.95

  • 3:50 #13 done20 line(s)

    shot 13·sharpness 2821.0

    1. :neo4j0.89
    2. EARENDIL1.00
    3. AlEngineer0.96
    4. The Shift0.99
    5. Tessl1.00
    6. • Early on, agentic coding felt like a secret unlock: you got more done, had breathing room0.99
    7. • Then it went mainstream and the expectation shifted0.98
    8. • Instead of reclaiming time, everyone is under pressure to ship faster0.99
    9. • The tools didn't buy us slack but they raised the baseline expectation0.99
    10. AlEngineer0.95
    11. ARMIN RONACHER & CRISTINA PONCELA0.99
    12. EARENDIL INC.1.00
    13. THE FRICTION IS YOUR JUDGMENT1.00
    14. AlEngineer0.98
    15. EUROPE1.00
    16. PRESENTED BY1.00
    17. Google DeepMind1.00
    18. Braintrust1.00
    19. WorkOS1.00
    20. OpenAl0.94

  • 4:05 #14 done26 line(s)

    shot 14·sharpness 3029.3

    1. Google DepMind0.90
    2. AlEngineer0.99
    3. EARENDIL1.00
    4. runpod1.00
    5. AlEnga0.80
    6. togetherai0.99
    7. ATri0.94
    8. AlEngine0.98
    9. The Shift1.00
    10. Microsoft0.98
    11. arize1.00
    12. • Early on, agentic coding felt like a secret unlock: you got more done, had breathing room0.99
    13. • Then it went mainstream and the expectation shifted1.00
    14. AlEng0.98
    15. • Instead of reclaiming time, everyone is under pressure to ship faster1.00
    16. • The tools didn't buy us slack but they raised the baseline expectation1.00
    17. ARMIN RONACHER & CRISTINA PONCELA0.98
    18. EARENDIL INC.1.00
    19. THE FRICTION IS YOUR JUDGMENT1.00
    20. AlEngineer0.98
    21. EUROPE1.00
    22. PRESENTED BY1.00
    23. Google DeepMind1.00
    24. Braintrust1.00
    25. WorkOS1.00
    26. OpenAl0.92

  • 4:10 #15 done22 line(s)

    shot 15·sharpness 2817.1

    1. EARENDIL1.00
    2. neod'0.80
    3. AlEngine0.98
    4. EUROPE1.00
    5. The Shift1.00
    6. Tes1.00
    7. • Early on, agentic coding felt like a secret unlock: you got more done, had breathing room1.00
    8. • Then it went mainstream and the expectation shifted0.99
    9. • Instead of reclaiming time, everyone is under pressure to ship faster0.99
    10. AlEngineer0.99
    11. • The tools didn't buy us slack but they raised the baseline expectation1.00
    12. EUROPE1.00
    13. ARMIN RONACHER & CRISTINA PONCELA0.98
    14. EARENDIL INC.1.00
    15. THE FRICTION IS YOUR JUDGMENT1.00
    16. AlEngineer0.98
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. Google DeepMind1.00
    20. # Braintrust0.95
    21. WorkOS1.00
    22. OpenAI0.93

  • 4:44 #16 done21 line(s)

    shot 16·sharpness 2880.4

    1. Braintrust1.00
    2. AlEngineer0.99
    3. EARENDIL1.00
    4. :neo4j0.91
    5. Tessi1.00
    6. orkel0.99
    7. AlEnginee0.97
    8. The Trap1.00
    9. Λarize0.91
    10. SENTRY1.00
    11. Agents produce output fast, and output feels like progress. So you stop0.99
    12. questioning, stop designing, stop saying "wait.1.00
    13. sonar0.96
    14. AlEngine0.96
    15. • Your brain rewards you for velocity: shipping feels good1.00
    16. • You get speed early, then quality drift later1.00
    17. • The agent doesn't know when to stop ... and neither do you, once you're in the flow0.99
    18. ARMIN RONACHER & CRISTINA PONCELA0.97
    19. EARENDIL INC.1.00
    20. Engineering the future of Al0.99
    21. AlEngineer0.99

  • 5:24 #17 done24 line(s)

    shot 17·sharpness 3371.0

    1. IllElever0.87
    2. EARENDIL1.00
    3. ev0.99
    4. ANIEr0.76
    5. The Trap1.00
    6. Agents produce output fast, and output feels like progress. So you stop0.99
    7. questioning, stop designing, stop saying "wait.0.99
    8. • Your brain rewards you for velocity: shipping feels good0.99
    9. ed1.00
    10. AIlEngin0.90
    11. • You get speed early, then quality drift later0.99
    12. • The agent doesn't know when to stop ... and neither do you, once you're in the flow0.99
    13. ARMIN RONACHER G CRISTINA PONCELA1.00
    14. EARENDIL INC.0.99
    15. THE FRICTION IS YOUR JUDGMENT1.00
    16. AlEngineer0.99
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. ARMIN RONACHER1.00
    20. CRISTINA PONCELA CUBEIRO1.00
    21. Google DeepMind1.00
    22. Founder1.00
    23. Software Engineer0.98
    24. EARENDIL1.00

  • 5:29 #18 done26 line(s)

    shot 18·sharpness 3576.7

    1. EARENDIL1.00
    2. IIElevenLabs0.96
    3. AlEngineer0.99
    4. The Amplification Problem1.00
    5. EUROPE1.00
    6. • Before agents: engineers balanced creating and reviewing code0.99
    7. neo0.96
    8. • Now every engineer has an agent amplifying their output0.99
    9. • But responsibility still rests on the human and it hasn't been amplified0.99
    10. • You are outnumbered by entities that cannot carry responsibility0.99
    11. AlEngineer0.99
    12. • Code reviews are increasingly skipped or rubber-stamped0.99
    13. EUROPE1.00
    14. • Goal: small PRs0.97
    15. ARMIN RONACHER G CRISTINA PONCELA0.98
    16. EARENDIL INC.0.99
    17. THE FRICTION IS YOUR JUDGMENT1.00
    18. AlEngineer0.95
    19. EUROPE1.00
    20. PRESENTED BY1.00
    21. ARMIN RONACHER1.00
    22. CRISTINA PONCELA CUBEIRO1.00
    23. Google DeepMind1.00
    24. Founder1.00
    25. Software Engineer1.00
    26. EARENDIL1.00

  • 5:42 #19 done32 line(s)

    shot 19·sharpness 3596.5

    1. Braintrust1.00
    2. AIEr0.89
    3. EARENDIL1.00
    4. AlEnn'0.74
    5. ne0.92
    6. The Amplification Problem1.00
    7. AlEr0.90
    8. • Before agents: engineers balanced creating and reviewing code0.99
    9. El0.74
    10. • Now every engineer has an agent amplifying their output0.99
    11. • But responsibility still rests on the human and it hasn't been amplified1.00
    12. AIEn0.90
    13. SI0.60
    14. • You are outnumbered by entities that cannot carry responsibility0.99
    15. EUR1.00
    16. • Code reviews are increasingly skipped or rubber-stamped0.99
    17. • Goal: small PRs0.98
    18. S0.96
    19. ar1.00
    20. AIEr0.94
    21. ARMIN RONACHER & CRISTINA PONCELA0.97
    22. EARENDIL INC.0.97
    23. THE FRICTION IS YOUR JUDGMENT1.00
    24. AlEngineer0.98
    25. EUROPE1.00
    26. PRESENTED BY1.00
    27. ARMIN RONACHER1.00
    28. CRISTINA PONCELA CUBEIRO1.00
    29. Google DeepMind1.00
    30. Founder1.00
    31. Software Engineer1.00
    32. EARENDIL1.00

  • 6:23 #20 done20 line(s)

    shot 20·sharpness 3379.0

    1. EARENDIL1.00
    2. The Amplification Problem1.00
    3. • Before agents: engineers balanced creating and reviewing code0.99
    4. • Now every engineer has an agent amplifying their output0.98
    5. • But responsibility still rests on the human and it hasn't been amplified0.99
    6. • You are outnumbered by entities that cannot carry responsibility0.99
    7. • Code reviews are increasingly skipped or rubber-stamped1.00
    8. • Goal: small PRs0.99
    9. ARMIN RONACHER & CRISTINA PONCELA0.97
    10. EARENDIL INC.0.98
    11. THE FRICTION IS YOUR JUDGMENT1.00
    12. AlEngineer0.98
    13. EUROPE1.00
    14. PRESENTED BY1.00
    15. ARMIN RONACHER1.00
    16. CRISTINA PONCELA CUBEIRO1.00
    17. Google DeepMind0.99
    18. Founder1.00
    19. Software Engineer1.00
    20. EARENDIL1.00

  • 6:31 #21 skipped

    shot 21·duplicate of #19

  • 6:58 #22 done22 line(s)

    shot 22·sharpness 3453.1

    1. EARENDIL1.00
    2. enAl0.98
    3. The Amplification Problem1.00
    4. • Before agents: engineers balanced creating and reviewing code0.99
    5. • Now every engineer has an agent amplifying their output0.98
    6. • But responsibility still rests on the human and it hasn't been amplified0.99
    7. • You are outnumbered by entities that cannot carry responsibility0.99
    8. no0.92
    9. • Code reviews are increasingly skipped or rubber-stamped0.99
    10. • Goal: small PRs0.94
    11. ARMIN RONACHER & CRISTINA PONCELA0.98
    12. EARENDIL INC.0.99
    13. THE FRICTION IS YOUR JUDGMENT1.00
    14. AlEngineer0.98
    15. EUROPE1.00
    16. PRESENTED BY1.00
    17. ARMIN RONACHER1.00
    18. CRISTINA PONCELA CUBEIRO1.00
    19. Google DeepMind0.99
    20. Founder1.00
    21. Software Engineer1.00
    22. EARENDIL1.00

  • 7:06 #23 done18 line(s)

    shot 23·sharpness 2755.1

    1. EARENDIL1.00
    2. OpenAl0.96
    3. The moments where you1.00
    4. want to skip thinking1.00
    5. are exactly the moments1.00
    6. where thinking matters most.1.00
    7. ARMIN RONACHER & CRISTINA PONCELA0.97
    8. EARENDIL INC.1.00
    9. THE FRICTION IS YOUR JUDGMENT1.00
    10. AlEngineer0.98
    11. EUROPE1.00
    12. PRESENTED BY1.00
    13. ARMIN RONACHER1.00
    14. CRISTINA PONCELA CUBEIRO1.00
    15. Google DeepMind1.00
    16. Founder1.00
    17. Software Engineer1.00
    18. EARENDIL1.00

Transcript

223 cues· 3,505 words· 18,782 chars

  1. 0:15 Morning.
  2. 0:15 Thanks for having us.
  3. 0:17 Today I want to talk with Christina about friction a little bit.
  4. 0:23 This is a social preview that came up automatically when someone submitted an issue to
  5. 0:35 Basically, this is a forum post that goes with a security incident that was deployed accidentally.
  6. 0:40 It was a configuration change that caused the problem.
  7. 0:43 And the social preview post had the marketing tagline of the company, which said, ship without friction.
  8. 0:50 And we want to encourage to add a little bit of friction to it.
  9. 0:56 And I'll tell you why.
  10. 0:58 So who are we?
  11. 0:59 I've been doing software development for 20 years, most of it in the open source space.
  12. 1:05 I have created Flask, which is a Python framework, which ironically is so much in the weights that a lot of people are learning about it now because the machines are producing it.
  13. 1:16 And I left my previous company that worked for a century in April last year, which perfectly coincided with me having time and then, obviously, Cloud Code.
  14. 1:25 And so I fell deep into a hole of agentic engineering.
  15. 1:28 And I started writing on my blog.
  16. 1:30 And a lot of people reached out to me over the last year being all excited about this.
  17. 1:36 And then I started with a friend in October, a company called Arendelle, where we are trying to make sense of all the AI things.
  18. 1:46 Yeah, and my name is Christina, and I work with Armin at this company called Arendelle.
  19. 1:51 But importantly, I am what I like to call a native AI engineer.
  20. 1:55 And what that basically means is that these tools have been around longer than I have.
  21. 2:00 So what this means is like they've been super foundational in how I've become a software engineer, not just because obviously I use them to work, but also because this is the means by which I've learned to do what I do.
  22. 2:11 And before Arendelle, I was working at Bending Spoons.
  23. 2:16 So we want to share a little bit from practice, not just theory, but I will readily admit that I don't think I have all the solutions.
  24. 2:23 So we have been building with or on agents for a good 12 months.
  25. 2:27 We had huge leverage and great disappointment.
  26. 2:31 And we really keep running into two types of problems.
  27. 2:35 I think especially if you listen to some earlier talks at this conference, you will have learned a lot about that you should keep using your brain.
  28. 2:44 For some reason, it's really, really hard.
  29. 2:45 So there's a psychological problem.
  30. 2:47 And the other one is the engineering challenges.
  31. 2:49 They seem to be producing worse code for some people and better code for some other people.
  32. 2:53 And what is it that actually makes it work?
  33. 2:56 And so this is really not a solution as it is our part of the journey of how we think so far we have managed.
  34. 3:03 Yeah.
  35. 3:05 So problem number one is the psychology part, which is like why is it, even though everybody told you many times over that you should be using your brain, you should be slowing down, it's actually incredibly hard.
  36. 3:14 It's just one more prompt and we don't sleep that much.
  37. 3:17 What is it that actually makes it so hard?
  38. 3:19 And then would it be that hard if the machines would actually be writing perfect code and we wouldn't have to think quite as much?
  39. 3:26 Is there something we can do to make this a little bit better?
  40. 3:29 So I'll begin by introducing the first part of these problems, the psychology problem.
  41. 3:34 And what I want to talk first about is the shift.
  42. 3:37 So I'm sure a lot of us here who have been playing with these tools for a while now experienced this at some point.
  43. 3:42 We were prompting, prompting, not so good.
  44. 3:44 And then at some point suddenly it clicked and they were really, really useful for us.
  45. 3:49 And it was fun in the beginning, and they gave us a lot of extra time, because not everyone was using them.
  46. 3:55 They were actually tools that made us more productive, that made it more fun to do our jobs.
  47. 3:59 But very quickly, because they were so useful and they got us so hooked, everyone was using them.
  48. 4:03 And so this kind of had the opposite effect, where suddenly the baseline expectation was just that everyone is now using them, and you have to use them.
  49. 4:11 And so this fun and free time translated into pressure.
  50. 4:15 Now we all have to ship faster and produce more code, and it is just not sustainable to review and to actually have time to think.

Chapters

  1. 0:00 The Problem
  2. 3:35 The Psychological Trap: Because AI tools are addictive and make coding feel effortless, engineers often stop taking the time to design, review, and truly understand the code being generated.
  3. 7:15 The Engineering Challenge: Agents are optimized for producing code that runs, not code that is maintainable or architecturally sound. This often results in "slop"—code that creates unexpected failure conditions and entropy
  4. 10:55 Agent-Legible Codebases: To maximize AI effectiveness, the speakers suggest designing codebases as infrastructure. This includes
  5. 12:35 Mechanical Enforcement through strict linting (e.g., no bare catch-alls, unique function names, and avoiding hidden magic like dynamic imports)
  6. 14:27 Reintroducing Friction: The speakers advocate for slowing down. They recommend identifying specific, high-stakes areas (such as database migrations or permission changes) where human judgment is non-negotiable and must be intentionally re-inserted into the development process
  7. 17:25 Conclusion

Open at this second