vidtheque.

read-only demo

Videos n97BCfyFIvw

"The engineer of the future is the person who is able to choose what is worth doing." — Addy Osmani

index_state ready data_status ok

AI Engineer· published 2026-07-14· 0:18:26· en-US· indexed 2026-08-10 19:48

Open on YouTube

Scene timeline

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

49 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
218
whisperx 218
chunks
33
from 218 cues
keyframes
46
kept of 49 captured
frames with text
45
1,241 lines read
chapters
16
from the source metadata
keyframe bytes
6.7 MB
word timings on 218 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 11:22 2m 30s
stt done — 2026-08-10 11:24 24s
chunk done — 2026-08-10 11:25 0s
text_embed done — 2026-08-10 19:48 0s
keyframe done — 2026-08-10 11:25 1m 58s
ocr done — 2026-08-10 11:27 23s
frame_embed done — 2026-08-10 19:48 9s

Frames, and what the machine read

  • 0:02 #0 done2 line(s)

    shot 0·sharpness 455.1

    1. AlEngineer0.96
    2. World's Fair0.97

  • 0:03 #1 done2 line(s)

    shot 1·sharpness 659.0

    1. AlEngineer0.96
    2. World's Fair0.99

  • 0:10 #2 done24 line(s)

    shot 2·sharpness 2734.5

    1. LAB & PLATINUM SPONSORS0.99
    2. Amazon AGI Lab0.98
    3. ANTHROP\C1.00
    4. Google DeepMind1.00
    5. MINIMAX0.96
    6. OpenAI0.92
    7. Akamai1.00
    8. arize1.00
    9. aws1.00
    10. Braintrust bright data0.98
    11. B1.00
    12. Browserbase1.00
    13. docker1.00
    14. :neo4j0.93
    15. ORACLE1.00
    16. PayPal1.00
    17. qodo1.00
    18. reducto1.00
    19. Sonar1.00
    20. Makers of0.99
    21. togetherai1.00
    22. Unblocked1.00
    23. WorkOS1.00
    24. SonarQube1.00

  • 0:17 #3 done8 line(s)

    shot 3·sharpness 930.5

    1. ADDYOSMANI1.00
    2. promptol0.97
    3. ineer1.00
    4. 'sFair0.94
    5. FORMER DIRECTOR OF GOOGLE CLOUD AI0.99
    6. Google1.00
    7. AIE1.00
    8. BeNCORD0.76

  • 0:20 #4 done62 line(s)

    shot 4·sharpness 4121.0

    1. World'sFai0.99
    2. ORACLE0.99
    3. arize1.00
    4. Google DeepMind1.00
    5. World'sF0.96
    6. :neo4j0.89
    7. Z.AI0.99
    8. bright data1.00
    9. extend1.00
    10. vast.ai1.00
    11. Ref.1.00
    12. RedHat1.00
    13. mezmo1.00
    14. stigg1.00
    15. World'sF0.98
    16. REL1.00
    17. World'sFa0.92
    18. World'sFe0.93
    19. Modal1.00
    20. promptqt0.95
    21. World'sFai0.93
    22. THE VELOCITYROOM0.98
    23. fiddler1.00
    24. forid'sFa0.85
    25. Surreal0.99
    26. ZERO1.00
    27. comet1.00
    28. authe0.99
    29. SOIO.IO0.92
    30. World's Fair1.00
    31. Modular1.00
    32. MERGE1.00
    33. AUTOMATTIC1.00
    34. Zed1.00
    35. Meticulous1.00
    36. Daytona1.00
    37. GRAVITEE1.00
    38. OINNGEST0.93
    39. BAND1.00
    40. World's Fai0.94
    41. Cleric1.00
    42. A ATLASSIAN0.97
    43. FACTORY1.00
    44. baseten1.00
    45. Z.AI0.99
    46. qodo1.00
    47. PayPal1.00
    48. :neo4j0.93
    49. World'sF0.99
    50. arize1.00
    51. reducto1.00
    52. OpenAl0.99
    53. Amazon AGI Lab1.00
    54. Microsoft1.00
    55. Braintrust1.00
    56. OpenAl0.98
    57. aws0.99
    58. DATA1.00
    59. LanceDB1.00
    60. ilder.io0.96
    61. descupe0.96
    62. TOPK1.00

  • 0:32 #5 done9 line(s)

    shot 5·sharpness 733.5

    1. World'sFair0.96
    2. IVicrosoTt0.75
    3. AlEngineer1.00
    4. OpenAl0.92
    5. World's Fair0.98
    6. AlEngineer0.99
    7. World's Fa0.96
    8. Akamai1.00
    9. AlEngineer0.99

  • 0:47 #6 done19 line(s)

    shot 6·sharpness 3051.5

    1. AlEngineer0.98
    2. THE FUTURE OF ENGINEERING0.99
    3. World'sFair1.00
    4. The engineer of the future will choose0.97
    5. PRESENTED BY0.99
    6. what is worth doing, then own the0.99
    7. Microsoft1.00
    8. evidence, understanding and verdict for0.99
    9. work increasingly automated by agents.0.98
    10. ia'stair0.84
    11. oenAl0.83
    12. Wor1.00
    13. Al0.77
    14. @addyosmani1.00
    15. IEnginer0.91
    16. Keynote1.00
    17. Id's0.99
    18. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    19. Google1.00

  • 1:13 #7 done27 line(s)

    shot 7·sharpness 2168.1

    1. AlEngineer0.96
    2. COURT1.00
    3. World'sFair1.00
    4. REDUNDANT1.00
    5. OF1.00
    6. CODE1.00
    7. PRESENTED BY1.00
    8. Microsoft1.00
    9. COURT1.00
    10. REDUNDANT1.00
    11. OF1.00
    12. CODE1.00
    13. id'stair0.79
    14. IVI0.85
    15. Judge1.00
    16. CODE1.00
    17. IS1.00
    18. Bugs1.00
    19. LIFE0.95
    20. AI0.87
    21. enAl0.94
    22. Worl1.00
    23. Enginee1.00
    24. Keynote1.00
    25. Id's0.94
    26. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    27. Google1.00

  • 1:35 #8 done23 line(s)

    shot 8·sharpness 3965.8

    1. Quality = the system of checks that0.97
    2. AlEngineer1.00
    3. World's Fair0.96
    4. produces evidence.0.97
    5. Verdict = the human/accountable0.98
    6. PRESENTED BY1.00
    7. Microsoft1.00
    8. decision made from that evidence.0.99
    9. Answerability = the ability to explain0.97
    10. asFair0.84
    11. IVII0.93
    12. and stand behind the verdict later.0.98
    13. AlEr0.89
    14. enAl0.98
    15. Work0.99
    16. @addyosmani0.97
    17. gineer1.00
    18. Keynote1.00
    19. d'sF0.99
    20. AK0.70
    21. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    22. Google1.00
    23. AIEr0.91

  • 2:02 #9 done36 line(s)

    shot 9·sharpness 5857.6

    1. Boris Cherny0.98
    2. ø …0.62
    3. @bcherny1.00
    4. AlEngineer0.98
    5. World'sFair1.00
    6. As engineering, product, design, DS, etc. melt into a new kind of role, I0.99
    7. was reflecting on what roles might look like in the future. For example,0.99
    8. when I look at the Claude Code team I see what I think is five archetypes:0.99
    9. 1. Prototyper: comes up with brand new ideas; churns out many ideas,1.00
    10. most of which don't ship1.00
    11. PRESENTED BY0.95
    12. 2. Builder: quickly turns a prototype/idea into production-grade0.99
    13. product/infra1.00
    14. Microsoft1.00
    15. 3. Sweeper: cleans up the UI, simplifies the code and system, unships,0.99
    16. optimizes performance1.00
    17. 4. Grower: takes a product that has been built and iterates on it to0.98
    18. improve Product-Market Fit0.99
    19. 5. Maintainer: owns a mature system to make it secure, reliable, fast,1.00
    20. and efficient as it scales1.00
    21. Many people span across 2 roles, and sometimes 3 roles. I also notice0.99
    22. asFair0.88
    23. IVII0.97
    24. that these roles are not really tied to job function -- eg. across Anthropic,0.98
    25. some designers match category 1, some 2, some 3; same for engineers,0.99
    26. PM, DS.0.99
    27. AIEI0.88
    28. enAl0.98
    29. Worl1.00
    30. @addyosmani1.00
    31. ngineer1.00
    32. Keynote1.00
    33. d's Fa0.90
    34. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    35. Google1.00
    36. AIEI0.95

  • 2:29 #10 done22 line(s)

    shot 10·sharpness 3377.6

    1. THE FUTURE OF CAREERS0.97
    2. AlEngineer0.98
    3. World'sFair1.00
    4. Roles are unbundling from craft1.00
    5. and rebundling around ownership.0.98
    6. PRESENTED BY1.00
    7. Microsoft1.00
    8. Prototype.Build.Sweep.Grow.Maintain.1.00
    9. The title matters less than0.99
    10. the part of the system you can own.0.98
    11. astair0.95
    12. IVII0.99
    13. enAl0.98
    14. AIEr0.90
    15. Work0.99
    16. @addyosmani1.00
    17. ngineer0.97
    18. Keynote1.00
    19. d'sF1.00
    20. AK0.77
    21. Addy Osmani / Former Director of Google Cloud Al/Gemini0.98
    22. Google1.00

  • 2:41 #11 done38 line(s)

    shot 11·sharpness 2566.9

    1. AlEngineer0.98
    2. THE AGENT IS THE SYSTEM AROUND THE MODEL0.99
    3. World's Fair0.97
    4. HARNESS ENGINEERING1.00
    5. The scaffolding that turns a model into an agent.0.99
    6. agent = model + harness0.98
    7. harness: prompts, tools, state, constraints, feedback loops1.00
    8. PRESENTED BY1.00
    9. CONTEXT1.00
    10. ACTION1.00
    11. rules, memory0.99
    12. tools, MCPs1.00
    13. Microsoft1.00
    14. FAILURE1.00
    15. MODEL1.00
    16. RATCHET1.00
    17. agent slipped1.00
    18. reasons / decides1.00
    19. new rule1.00
    20. CONTROL1.00
    21. PERSIST1.00
    22. plans, routing1.00
    23. one chip on the board1.00
    24. files, git0.98
    25. OBSERVE1.00
    26. HOOKS1.00
    27. tests, logs0.99
    28. block, retry0.99
    29. ia'stair0.83
    30. enAl0.99
    31. Wor1.00
    32. Al0.92
    33. Source: addyosmani.com/blog/agent-harness-engineering1.00
    34. Engineer1.00
    35. Id'sF0.99
    36. Keynote1.00
    37. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    38. Google1.00

  • 2:59 #12 done37 line(s)

    shot 12·sharpness 2881.0

    1. AlEngineer1.00
    2. FROM PROMPTING AGENTS TO DESIGNING THE SYSTEM THAT PROMPTS THEM0.99
    3. World'sFair1.00
    4. LOOP ENGINEERING1.00
    5. Design the loop, not the prompt.0.98
    6. loop = goal + cadence + isolated work + verification + state1.00
    7. VERDICT1.00
    8. STATE1.00
    9. AUTOMATE1.00
    10. PRESENTED BY1.00
    11. owns outer loop1.00
    12. cadence finds work1.00
    13. memory lives outside1.00
    14. Microsoft1.00
    15. DECIDE1.00
    16. RECURSIVE1.00
    17. ACT1.00
    18. ship, block, queue0.99
    19. GOAL1.00
    20. agents in worktrees0.99
    21. iterate until done1.00
    22. ISOLATION1.00
    23. LEARNING1.00
    24. VERIFY1.00
    25. parallel, no chaos1.00
    26. maker != checker1.00
    27. tomorrow reads today0.99
    28. rid'sFair0.82
    29. Loops change the work. They do not delete the engineer.1.00
    30. penAl0.96
    31. Wot0.90
    32. Source: addyosmani.com/blog/loop-engineering0.99
    33. AlEngineer –0.92
    34. rld'sF0.89
    35. Keynote1.00
    36. Addy Osmani/ Former Director of Google Cloud Al/Gemini0.98
    37. Google1.00

  • 3:17 #13 done37 line(s)

    shot 13·sharpness 2950.7

    1. AlEngineer0.97
    2. THE SOFTWARE FACTORY, WITH THE LIGHTS ON1.00
    3. World'sFair1.00
    4. AGENTIC SOFTWARE FACTORY1.00
    5. monitor1.00
    6. users1.00
    7. product intent1.00
    8. agent inner loop0.99
    9. PRESENTED BY0.99
    10. incidents1.00
    11. guide / context0.96
    12. evidence1.00
    13. prod1.00
    14. tests1.00
    15. Microsoft1.00
    16. generate1.00
    17. diff summary1.00
    18. risk notes1.00
    19. user feedback1.00
    20. verify / solve0.97
    21. sandbox,1.00
    22. traces,1.00
    23. tests1.00
    24. stuff worth doing1.00
    25. human verdict1.00
    26. rid'srair0.88
    27. lights off fails here1.00
    28. ship / block / redirect0.98
    29. The win is not removing people from the loop.0.98
    30. penAl0.96
    31. Woi0.83
    32. The win is moving human judgment to the highest leverage checkpoint.0.99
    33. AlEnginee0.99
    34. Keynote1.00
    35. rld's0.99
    36. Addy Osmani / Former Director of Google Cloud Al/Gemini0.98
    37. Google1.00

  • 3:47 #14 done33 line(s)

    shot 14·sharpness 2549.3

    1. AIEngineer0.96
    2. THE OUTPUT CURVE1.00
    3. World'sFair1.00
    4. Al code share is no longer marginal0.99
    5. PRESENTED BY0.97
    6. 65%1.00
    7. 55%1.00
    8. 42%0.99
    9. Microsoft1.00
    10. 42%1.00
    11. of committed1.00
    12. code is already1.00
    13. 19%1.00
    14. Al-generated or0.98
    15. significantly1.00
    16. 6%1.00
    17. assisted1.00
    18. rid'stair0.93
    19. 20231.00
    20. The factory is no longer experimental. It is entering the commit0.98
    21. 20241.00
    22. now1.00
    23. 2026 est.0.95
    24. 2027 est.0.95
    25. history.1.00
    26. penAl0.97
    27. Wo1.00
    28. Source: Sonar State of Code Developer Survey report 20261.00
    29. AlEngineer0.99
    30. Keynote1.00
    31. rld's F0.97
    32. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    33. Google1.00

  • 4:13 #15 done16 line(s)

    shot 15·sharpness 2411.9

    1. AlEngineer0.97
    2. World'sFair1.00
    3. Cleancodeis cheaper0.99
    4. PRESENTED BY0.99
    5. foragents to read.0.96
    6. Microsoft1.00
    7. 7-8% fewer tokens. 34% fewer file revisits. Same pass rate.0.99
    8. rd'stair0.88
    9. penAl0.94
    10. Wol0.95
    11. @addyosmani1.00
    12. AlEngineer0.99
    13. Keynote1.00
    14. ld's0.86
    15. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    16. Google1.00

  • 4:47 #16 done28 line(s)

    shot 16·sharpness 3212.5

    1. AlEngineer0.99
    2. TRUST WITHOUT CAPACITY0.99
    3. World's Fair0.98
    4. Reviewers are already overloaded0.98
    5. PRESENTED BY0.96
    6. Do not fully trust Al-generated code0.99
    7. 96%1.00
    8. Microsoft1.00
    9. 2x1.00
    10. Always verify Al code before committing0.99
    11. 48%1.00
    12. is high, but1.00
    13. trust gap: skepticism0.99
    14. verification is not0.99
    15. keeping up1.00
    16. Say reviewing Al code takes longer than human code1.00
    17. 38%1.00
    18. ra'sFair0.82
    19. The risk is not that engineers distrust Al. It is that they distrust it and0.99
    20. still ship faster than they verify.0.99
    21. penAl0.97
    22. Wol0.88
    23. Source: Sonar State of Code Developer Survey report 20260.99
    24. AlEnginee1.00
    25. rld's0.95
    26. Keynote1.00
    27. Addy Osmani / Former Director of Google Cloud Al/Gemini0.98
    28. Google1.00

  • 5:25 #17 done34 line(s)

    shot 17·sharpness 3489.4

    1. AlEngineer0.99
    2. THE GOVERNANCE GAP0.98
    3. World'sFair1.00
    4. Generation moved faster than control0.98
    5. Review/validation is now the bottleneck1.00
    6. 85%1.00
    7. Governance after creation is the challenge0.99
    8. 84%1.00
    9. 92%0.99
    10. Al code risks new technical debt1.00
    11. 82%1.00
    12. report some1.00
    13. governance1.00
    14. Adopted Al faster than policy0.99
    15. challenge with0.97
    16. 80%1.00
    17. Al-generated code1.00
    18. Cannot reliably distinguish Al vs human code0.97
    19. 43%1.00
    20. rstair0.84
    21. Amazor0.98
    22. This is the argument for the HumanVerdict: provenance, intent, and ownership need1.00
    23. AlEr0.94
    24. a first-class interface.1.00
    25. enAl0.98
    26. Work0.94
    27. Source: GitLab AI accountability research, June 20261.00
    28. ogineer0.90
    29. d'sF0.97
    30. Mi1.00
    31. Keynote1.00
    32. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    33. Google1.00
    34. AlEr0.91

  • 5:35 #18 done18 line(s)

    shot 18·sharpness 2083.3

    1. AlEngineer0.98
    2. World'sFair1.00
    3. The agent can ship more than you can1.00
    4. review.1.00
    5. So what are you still for?1.00
    6. a'stair0.92
    7. Amazoi0.86
    8. AIE1.00
    9. enAl0.97
    10. Worl1.00
    11. @addyosmani0.97
    12. ngineer1.00
    13. d's1.00
    14. M1.00
    15. Keynote1.00
    16. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    17. Google1.00
    18. AIE1.00

  • 5:45 #19 done16 line(s)

    shot 19·sharpness 1413.9

    1. AlEngineer0.99
    2. World'sFair1.00
    3. CLAUDE1.00
    4. CODE1.00
    5. lastair0.80
    6. Amazo0.85
    7. Al0.78
    8. enAl0.98
    9. Worl1.00
    10. Engineer0.97
    11. Id'sF0.85
    12. M1.00
    13. Keynote1.00
    14. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    15. Google1.00
    16. AI0.80

  • 5:56 #20 skipped

    shot 20·duplicate of #18

  • 6:23 #21 done21 line(s)

    shot 21·sharpness 2919.6

    1. AlEngineer0.99
    2. World's Fair0.98
    3. Alphaisthegap.1.00
    4. Alpha:1.00
    5. what makes you meaningfully better0.99
    6. than what current models can do.0.99
    7. Decay:1.00
    8. a's Fair0.86
    9. Amazo0.98
    10. how fast the models catch up.0.98
    11. All1.00
    12. enAl0.99
    13. Worl1.00
    14. @addyosmani1.00
    15. d'sF0.99
    16. Engineer1.00
    17. M1.00
    18. Keynote1.00
    19. Addy Osmani / Former Director of Google Cloud Al/Gemini0.99
    20. Google1.00
    21. All0.91

  • 6:56 #22 done26 line(s)

    shot 22·sharpness 3709.2

    1. AIEngineer0.96
    2. World'sFair1.00
    3. Paul Graham0.99
    4. @paulg1.00
    5. Prediction: In the Al age, taste will become even more important. When0.99
    6. anyone can make anything, the big differentiator is what you choose to1.00
    7. make.1.00
    8. paulgraham.com/taste.html1.00
    9. 1:31 AM·Feb 14, 2026· 2.1M Views0.96
    10. 'stair0.93
    11. Amazon0.99
    12. 8601.00
    13. 2.2K1.00
    14. 13K1.00
    15. 7.2K1.00
    16. AIEng0.91
    17. enAl0.86
    18. World1.00
    19. @addyosmani1.00
    20. 's F:0.89
    21. jineer1.00
    22. Mic0.99
    23. Keynote1.00
    24. Addy Osmani / Former Director of Google Cloud Al/Gemini0.98
    25. Google1.00
    26. AIEng0.91

  • 7:27 #23 done29 line(s)

    shot 23·sharpness 4691.0

    1. DefiningTaste1.00
    2. AlEngineer1.00
    3. World'sFair1.00
    4. Mitchell Hashimoto0.97
    5. @mitchellh · Jun 260.94
    6. Q1300.89
    7. t5250.87
    8. 3.8K1.00
    9. 706K0.93
    10. "Taste” is the ability to consistently make high-quality qualitative judgments0.98
    11. where no objective metric exists. It's the creation of something that feels right0.99
    12. intuitively, with no real justifiable way to measure that. But when you do it,0.99
    13. people feel it.0.98
    14. A person with "good taste" is someone who can do this repeatedly,0.99
    15. consistently. The funny thing about taste is that it's hard to create, but its0.99
    16. result is very easy to copy. Once someone makes a tasteful decision, others1.00
    17. 1a'sFair0.89
    18. Amazo0.93
    19. can imitate it almost immediately.1.00
    20. This is usually an argument against the existence of taste: "look how easy I0.99
    21. benAl0.89
    22. Wor1.00
    23. A1.00
    24. @addyosmani1.00
    25. IEngineer0.97
    26. Keynote1.00
    27. Id'sF0.98
    28. Addy Osmani / Former Director of Google Cloud Al/Gemini0.97
    29. Google1.00

Transcript

218 cues· 3,037 words· 16,895 chars

  1. 0:20 Howdy, folks.
  2. 0:23 So good afternoon or good whatever time it is when you're watching this on YouTube.
  3. 0:29 I'm really excited to be here.
  4. 0:31 And today I want to talk to you about really what it takes to keep the human in the loop where engineering is concerned.
  5. 0:40 I really want to start with the human side before we talk about the architecture here.
  6. 0:45 I think that the engineer of the future is going to be really defined by the person who is able to choose what is worth doing.
  7. 0:54 They're gonna own the evidence, they're gonna own the understanding, as well as the verdict around increasingly automated work that's being done by agents.
  8. 1:04 Now, when I use the term verdict, I don't mean that we're suddenly all gonna be Judge Judy.
  9. 1:10 We're not.
  10. 1:11 But what I mean really is something just a little bit different.
  11. 1:15 I mean, we're gonna be accountable for the production decisions.
  12. 1:19 Does something shift?
  13. 1:21 Do we block it?
  14. 1:22 Do we redirect it or accept the risk?
  15. 1:25 Quality is something that we all talk about a lot, but quality produces evidence.
  16. 1:30 A verdict assigns responsibility.
  17. 1:34 And answerability is really what lets us stand behind a verdict.
  18. 1:39 And this, of course, is not the only way that our industry is starting to think about our roles evolving.
  19. 1:46 Boris Cherny recently put some useful language around what many teams are starting to feel.
  20. 1:51 The old craft boundaries are getting blurry and roles are rebundling around the work itself.
  21. 1:58 And the important question here becomes a lot less about what is your title and more what part of the system can you own?
  22. 2:06 Now, I like this taxonomy quite a lot.
  23. 2:10 It's optimistic without being overly vague.
  24. 2:13 So things like prototype, build, sweep, grow, and maintain.
  25. 2:17 And these are real engineering modes.
  26. 2:20 Agents are going to help with all of them, but the scarce thing is not merely doing the task.
  27. 2:25 It's going to be knowing which mode your product needs and what quality bar applies and who owns the result at the end of the day.
  28. 2:33 Now, we've been talking about harnesses and loop engineering in software factories over the last couple of days.
  29. 2:39 We can talk why this shift is happening.
  30. 2:41 We've moved past the model as the whole story, right?
  31. 2:44 With harness engineering, the coding agent is the model plus the harness around it, right?
  32. 2:49 Your context, your tools, your file system, Git.
  33. 2:52 And the harness is what turns intelligence into something that you can delegate to.
  34. 2:56 The next move was loop engineering, where we weren't just prompting one run anymore.
  35. 3:01 We were designing systems that kept prompting, checking, and remembering, and deciding what happened next.
  36. 3:07 And that's really when agents started to feel like infrastructure.
  37. 3:10 And once you start putting all of those things together, you get that software factory.
  38. 3:15 Dex covered this well in his talk.
  39. 3:17 But you have agents that are running inside that inner loop, and evidence that comes out.
  40. 3:22 Humans still end up making the production decisions.
  41. 3:25 in this loop, and the win really isn't moving us from it.
  42. 3:29 The win is moving human judgments the highest leverage checkpoint, I think.
  43. 3:34 And this is why it starts to matter now.
  44. 3:37 AI-generated and AI-assisted code is becoming normal code for a lot of us.
  45. 3:42 One of Sonar's 2026 surveys said that AI-assisted code is no longer marginal.
  46. 3:47 It's increasingly having a large role in our code bases.
  47. 3:51 And once that happens, answerability stops being this philosophical world.
  48. 3:55 It becomes an engineering requirement.
  49. 3:57 And there's a quality point here as well, right?
  50. 3:59 Like we used to care about clean code, code that people could read, but cleaner code is actually not just gonna help the next human and the next person on your teams.

Chapters

  1. 0:00 Introduction and the human side of engineering
  2. 1:46 Rebundling roles and ownership of systems
  3. 2:34 Harnesses, loop engineering, and software factories
  4. 3:34 The shift to answerability as an engineering requirement
  5. 4:26 Reviewing AI-assisted code and organizational bottlenecks
  6. 5:55 Redefining leverage through human judgment
  7. 6:15 Alpha, decay, and the role of "taste"
  8. 8:49 Defining the modern software engineer
  9. 9:50 Risks to avoid: cognitive debt and surrender
  10. 11:51 Orchestration tax and system design
  11. 12:39 Accountability as the foundation for scaling
  12. 13:16 Career math: credibility vs. capability
  13. 14:13 High agency and the decision-making ladder
  14. 15:13 Defining the boundary between agents and humans
  15. 16:13 Operational rule: explain it or don't ship it
  16. 17:20 Future outlook: unlocking latent demand

Open at this second