vidtheque.

read-only demo

Videos AMiyLItEtLA

fighting slop with slop — Vaibhav Gupta, Boundary

index_state ready data_status ok

AI Engineer· published 2026-07-31· 0:21:32· en-US· indexed 2026-08-10 19:37

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 0:17, 1 of 1 keyframes kept
  5. Shot 4, 0:17 to 0:26, 1 of 1 keyframes kept
  6. Shot 5, 0:26 to 0:38, 1 of 1 keyframes kept
  7. Shot 6, 0:38 to 0:41, 1 of 1 keyframes kept
  8. Shot 7, 0:41 to 0:48, 1 of 1 keyframes kept
  9. Shot 8, 0:48 to 1:08, 1 of 1 keyframes kept
  10. Shot 9, 1:08 to 1:16, 1 of 1 keyframes kept
  11. Shot 10, 1:16 to 1:26, 1 of 1 keyframes kept
  12. Shot 11, 1:26 to 1:42, 1 of 1 keyframes kept
  13. Shot 12, 1:42 to 1:51, 1 of 1 keyframes kept
  14. Shot 13, 1:51 to 2:09, 1 of 1 keyframes kept
  15. Shot 14, 2:09 to 2:37, 1 of 1 keyframes kept
  16. Shot 15, 2:37 to 2:45, 1 of 1 keyframes kept
  17. Shot 16, 2:45 to 2:53, 1 of 1 keyframes kept
  18. Shot 17, 2:53 to 3:01, 1 of 1 keyframes kept
  19. Shot 18, 3:01 to 3:07, 1 of 1 keyframes kept
  20. Shot 19, 3:07 to 3:28, 1 of 1 keyframes kept
  21. Shot 20, 3:28 to 3:48, 1 of 1 keyframes kept
  22. Shot 21, 3:48 to 3:58, 1 of 1 keyframes kept
  23. Shot 22, 3:58 to 4:04, 1 of 1 keyframes kept
  24. Shot 23, 4:04 to 4:07, 1 of 1 keyframes kept
  25. Shot 24, 4:07 to 4:09, 1 of 1 keyframes kept
  26. Shot 25, 4:09 to 4:11, 0 of 1 keyframes kept
  27. Shot 26, 4:11 to 4:14, 1 of 1 keyframes kept
  28. Shot 27, 4:14 to 4:20, 1 of 1 keyframes kept
  29. Shot 28, 4:20 to 4:31, 1 of 1 keyframes kept
  30. Shot 29, 4:31 to 4:35, 0 of 1 keyframes kept
  31. Shot 30, 4:35 to 4:36, 0 of 1 keyframes kept
  32. Shot 31, 4:36 to 5:01, 1 of 1 keyframes kept
  33. Shot 32, 5:01 to 5:11, 1 of 1 keyframes kept
  34. Shot 33, 5:11 to 5:17, 1 of 1 keyframes kept
  35. Shot 34, 5:17 to 5:21, 1 of 1 keyframes kept
  36. Shot 35, 5:21 to 5:25, 1 of 1 keyframes kept
  37. Shot 36, 5:25 to 5:26, 0 of 1 keyframes kept
  38. Shot 37, 5:26 to 5:39, 1 of 1 keyframes kept
  39. Shot 38, 5:39 to 5:59, 1 of 1 keyframes kept
  40. Shot 39, 5:59 to 6:17, 0 of 1 keyframes kept
  41. Shot 40, 6:17 to 6:24, 0 of 1 keyframes kept
  42. Shot 41, 6:24 to 7:00, 1 of 1 keyframes kept
  43. Shot 42, 7:00 to 7:10, 1 of 1 keyframes kept
  44. Shot 43, 7:10 to 7:22, 0 of 1 keyframes kept
  45. Shot 44, 7:22 to 7:32, 1 of 1 keyframes kept
  46. Shot 45, 7:32 to 7:54, 1 of 1 keyframes kept
  47. Shot 46, 7:54 to 8:37, 1 of 1 keyframes kept
  48. Shot 47, 8:37 to 8:39, 1 of 1 keyframes kept
  49. Shot 48, 8:39 to 8:41, 1 of 1 keyframes kept
  50. Shot 49, 8:41 to 9:13, 1 of 1 keyframes kept
  51. Shot 50, 9:13 to 9:37, 1 of 1 keyframes kept
  52. Shot 51, 9:37 to 9:38, 1 of 1 keyframes kept
  53. Shot 52, 9:38 to 9:44, 1 of 1 keyframes kept
  54. Shot 53, 9:44 to 9:47, 0 of 1 keyframes kept
  55. Shot 54, 9:47 to 10:01, 1 of 1 keyframes kept
  56. Shot 55, 10:01 to 10:46, 1 of 1 keyframes kept
  57. Shot 56, 10:46 to 10:52, 1 of 1 keyframes kept
  58. Shot 57, 10:52 to 11:10, 1 of 1 keyframes kept
  59. Shot 58, 11:10 to 11:24, 1 of 1 keyframes kept
  60. Shot 59, 11:24 to 11:26, 1 of 1 keyframes kept
  61. Shot 60, 11:26 to 12:06, 1 of 1 keyframes kept
  62. Shot 61, 12:06 to 12:08, 1 of 1 keyframes kept
  63. Shot 62, 12:08 to 12:09, 1 of 1 keyframes kept
  64. Shot 63, 12:09 to 12:14, 1 of 1 keyframes kept
  65. Shot 64, 12:14 to 12:21, 1 of 1 keyframes kept
  66. Shot 65, 12:21 to 12:24, 1 of 1 keyframes kept
  67. Shot 66, 12:24 to 12:28, 0 of 1 keyframes kept
  68. Shot 67, 12:28 to 12:34, 0 of 1 keyframes kept
  69. Shot 68, 12:34 to 12:49, 1 of 1 keyframes kept
  70. Shot 69, 12:49 to 13:21, 1 of 1 keyframes kept
  71. Shot 70, 13:21 to 13:49, 1 of 1 keyframes kept
  72. Shot 71, 13:49 to 14:23, 1 of 1 keyframes kept
  73. Shot 72, 14:23 to 14:56, 1 of 1 keyframes kept
  74. Shot 73, 14:56 to 14:57, 1 of 1 keyframes kept
  75. Shot 74, 14:57 to 15:11, 0 of 1 keyframes kept
  76. Shot 75, 15:11 to 15:33, 0 of 1 keyframes kept
  77. Shot 76, 15:33 to 15:34, 1 of 1 keyframes kept
  78. Shot 77, 15:34 to 15:41, 1 of 1 keyframes kept
  79. Shot 78, 15:41 to 15:42, 0 of 1 keyframes kept
  80. Shot 79, 15:42 to 15:46, 1 of 1 keyframes kept
  81. Shot 80, 15:46 to 15:56, 1 of 1 keyframes kept
  82. Shot 81, 15:56 to 15:58, 1 of 1 keyframes kept
  83. Shot 82, 15:58 to 16:46, 0 of 1 keyframes kept
  84. Shot 83, 16:46 to 17:06, 1 of 1 keyframes kept
  85. Shot 84, 17:06 to 17:08, 0 of 1 keyframes kept
  86. Shot 85, 17:08 to 17:25, 1 of 1 keyframes kept
  87. Shot 86, 17:25 to 17:27, 0 of 1 keyframes kept
  88. Shot 87, 17:27 to 17:57, 1 of 1 keyframes kept
  89. Shot 88, 17:57 to 17:58, 1 of 1 keyframes kept
  90. Shot 89, 17:58 to 18:00, 1 of 1 keyframes kept
  91. Shot 90, 18:00 to 18:02, 1 of 1 keyframes kept
  92. Shot 91, 18:02 to 18:42, 1 of 1 keyframes kept
  93. Shot 92, 18:42 to 18:46, 1 of 1 keyframes kept
  94. Shot 93, 18:46 to 18:48, 1 of 1 keyframes kept
  95. Shot 94, 18:48 to 19:23, 1 of 1 keyframes kept
  96. Shot 95, 19:23 to 19:24, 1 of 1 keyframes kept
  97. Shot 96, 19:24 to 19:26, 1 of 1 keyframes kept
  98. Shot 97, 19:26 to 19:53, 0 of 1 keyframes kept
  99. Shot 98, 19:53 to 20:34, 1 of 1 keyframes kept
  100. Shot 99, 20:34 to 21:14, 1 of 1 keyframes kept
  101. Shot 100, 21:14 to 21:15, 1 of 1 keyframes kept
  102. Shot 101, 21:15 to 21:31, 0 of 1 keyframes kept
  103. Shot 102, 21:31 to 21:31, 1 of 1 keyframes kept

103 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
277
whisperx 277
chunks
39
from 277 cues
keyframes
85
kept of 103 captured
frames with text
84
4,097 lines read
chapters
13
from the source metadata
keyframe bytes
15.5 MB
word timings on 277 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-09 21:50 0s
stt done — 2026-08-09 02:57 24s
chunk done — 2026-08-09 02:57 0s
text_embed done — 2026-08-10 19:37 1s
keyframe done — 2026-08-09 02:57 3m 32s
ocr done — 2026-08-09 03:01 58s
frame_embed done — 2026-08-10 19:37 14s

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 664.7

    1. AlEngineer0.95
    2. World's Fair0.99

  • 0:11 #2 done24 line(s)

    shot 2·sharpness 2742.1

    1. LAB & PLATINUM SPONSORS0.97
    2. Amazon AGI Lab0.99
    3. ANTHROP\C1.00
    4. Google DeepMind1.00
    5. MINIMAX0.98
    6. OpenAl0.92
    7. Akamai1.00
    8. arize1.00
    9. aws1.00
    10. Braintrust bright data0.99
    11. B1.00
    12. Browserbase1.00
    13. docker1.00
    14. :neo4j0.91
    15. ORACLE1.00
    16. PayPal1.00
    17. qodo0.99
    18. reducto1.00
    19. Sonar1.00
    20. Makers of1.00
    21. together.ai0.98
    22. Unblocked1.00
    23. WorkOS1.00
    24. SonarQube1.00

  • 0:15 #3 done40 line(s)

    shot 3·sharpness 1570.2

    1. AlEngineer0.99
    2. AlEngineer0.99
    3. WorkOS1.00
    4. World's Fair1.00
    5. Amazon AGI Lab1.00
    6. World's Fair0.98
    7. World's Fair0.96
    8. AlEngineer0.99
    9. Braintrı0.93
    10. World's Fair0.99
    11. AIEngineer0.95
    12. OpenAI0.93
    13. W0.99
    14. AlEng'0.87
    15. AlEngineer0.99
    16. aws1.00
    17. Worl0.93
    18. fight slop with slop0.99
    19. togetherai1.00
    20. World's Fair0.98
    21. AlEngineer1.00
    22. AlEngineer0.97
    23. World's Fair1.00
    24. World's Fair0.99
    25. DATADOG1.00
    26. W0.99
    27. CLOUDFL0.99
    28. AlEnginee1.00
    29. AlEngineer1.00
    30. LanceDB1.00
    31. World's1.00
    32. builder.io1.00
    33. World's Fair0.99
    34. World's Fair0.96
    35. AlEngineer1.00
    36. TOFHA0.74
    37. World's Fair0.99
    38. AlEngineer0.99
    39. cognee1.00
    40. W0.98

  • 0:22 #4 done15 line(s)

    shot 4·sharpness 1694.0

    1. AlEngineer0.98
    2. World's Fair0.96
    3. Fight slop with slop1.00
    4. or1.00
    5. How we ship serious software1.00
    6. without ever reading the code1.00
    7. AlEngineer0.99
    8. Jst0.90
    9. World'sFai1.00
    10. BAML1.00
    11. ail0.82
    12. fight slop with slop0.98
    13. gether1.00
    14. AlEngineer0.99
    15. 'd'sFai1.00

  • 0:30 #5 done19 line(s)

    shot 5·sharpness 1475.9

    1. AlEngineer0.99
    2. World's Fair0.98
    3. No code reviews.1.00
    4. 010.99
    5. You must work on things in parallel.1.00
    6. 020.95
    7. Use Al however you want.0.98
    8. 031.00
    9. AlEnginee0.97
    10. ntrust1.00
    11. World's1.00
    12. PBAML0.93
    13. neer1.00
    14. 'sFai0.91
    15. fight slop with slop0.98
    16. get1.00
    17. Enginee0.96
    18. d's0.94
    19. LARE1.00

  • 0:41 #6 done11 line(s)

    shot 6·sharpness 1058.8

    1. we're probably a1.00
    2. AlEnginieer0.94
    3. World's Fair0.98
    4. zoomer YC startup0.98
    5. Braintru0.98
    6. Wc0.99
    7. BAML1.00
    8. — AlEngine0.93
    9. lorld'0.90
    10. Wc0.98
    11. LOUDF0.94

  • 0:46 #7 done37 line(s)

    shot 7·sharpness 1901.3

    1. AlEngineer0.98
    2. World's Fair0.99
    3. WA1.00
    4. WASHINGTON1.00
    5. DRIVER LICENSE0.98
    6. USA1.00
    7. FEDERAL LIMITS APPLY1.00
    8. 4d LIC#1.00
    9. 9CLASS1.00
    10. DONOR1.00
    11. 1GUPTA1.00
    12. clearly millenial0.99
    13. 2VAIBHAV1.00
    14. 3 DOB 02/200.97
    15. /19940.92
    16. 4a Iss 08/14/20220.94
    17. 8 1823 MINOR AVE0.99
    18. 17040.96
    19. SEATTLE WA 98101-09250.98
    20. 15 SEX M0.98
    21. 18 EYES BLK1.00
    22. 16 HGT 5'-07"1.00
    23. 17WGT 130 lb0.92
    24. AlEngineer0.99
    25. 12 RESTRICTIONS1.00
    26. 9a END NONE1.00
    27. trust1.00
    28. Norld'sF0.98
    29. B1.00
    30. 4b EXP 02/20/20270.98
    31. BAML1.00
    32. Fai1.00
    33. fight slop with slop0.97
    34. geth0.92
    35. AlEngineer-0.92
    36. 1d'sF0.95
    37. ARE1.00

  • 1:05 #8 done14 line(s)

    shot 8·sharpness 1739.6

    1. AlEngineer0.99
    2. World'sFair1.00
    3. BAML1.00
    4. AlEngineer0.98
    5. itrust0.99
    6. World'sI0.91
    7. BAML1.00
    8. Fai1.00
    9. fight slop with slop0.99
    10. geth1.00
    11. Engineering the future of Al0.99
    12. AlEngineer1.00
    13. ld'sl0.86
    14. .ARE0.97

  • 1:14 #9 done15 line(s)

    shot 9·sharpness 2662.0

    1. AlEngineer1.00
    2. World'sFair1.00
    3. SLOP1.00
    4. BAML1.00
    5. AlEngineer0.98
    6. trust1.00
    7. World'sF0.99
    8. BAML1.00
    9. Fai1.00
    10. fight slop with slop0.98
    11. geth0.96
    12. Engineering the future of Al1.00
    13. AlEngineer0.98
    14. rld'sF0.97
    15. ARE1.00

  • 1:18 #10 done17 line(s)

    shot 10·sharpness 1885.9

    1. AlEngineer0.98
    2. “To know your slop,0.98
    3. World's Fair0.96
    4. you must become your slop”0.96
    5. - Slop Tzu0.99
    6. AIEngine0.94
    7. intru:0.97
    8. World's1.00
    9. PBAML0.97
    10. ineer0.89
    11. 'sFai0.94
    12. stcslop wit slop0.76
    13. toget0.94
    14. Engineering the future of Al0.99
    15. — AlEngine0.86
    16. Vorld's1.00
    17. FLARE1.00

  • 1:40 #11 done17 line(s)

    shot 11·sharpness 1706.0

    1. AlEngineer0.98
    2. World's Fair0.93
    3. Slop1.00
    4. pronounced släp1.00
    5. Any code you don't read0.99
    6. AlEngin0.94
    7. intrt0.93
    8. World'0.99
    9. PBAML0.89
    10. gineer1.00
    11. I's0.99
    12. fight slop with slop0.97
    13. toge1.00
    14. Engineering the future of Al1.00
    15. AlEngin0.97
    16. Norld'0.90
    17. FLARI0.93

  • 1:48 #12 done19 line(s)

    shot 12·sharpness 3562.3

    1. No code reviews.1.00
    2. AlEngineer0.99
    3. World's Fair0.97
    4. You must work on things in parallel.1.00
    5. Hit0.51
    6. Use Al however you want.1.00
    7. SLOP1.00
    8. SLOP1.00
    9. BAML1.00
    10. Braintr0.98
    11. Worl0.96
    12. PBAML0.95
    13. AlEngin0.99
    14. rld'0.92
    15. fight slop with slop0.98
    16. to1.00
    17. Engineering the future of Al1.00
    18. Worl1.00
    19. UDFLA1.00

  • 1:57 #13 done16 line(s)

    shot 13·sharpness 2193.0

    1. Skirmish of Standards1.00
    2. AlEngineer0.98
    3. World's Fair0.94
    4. casualty: 200M tokens lost1.00
    5. AlEngine0.99
    6. intrust0.99
    7. World's1.00
    8. PBAML0.86
    9. jineer -0.87
    10. I'sF0.91
    11. fight slop with slop0.97
    12. toge1.00
    13. Engineering the future of Al1.00
    14. Engine1.00
    15. Vorld's1.00
    16. FLARE1.00

  • 2:21 #14 done32 line(s)

    shot 14·sharpness 2820.8

    1. Situation1.00
    2. Rule1.00
    3. AlEngineer0.99
    4. World'sFair1.00
    5. Adding a feature0.99
    6. Put it in the highest (earliest) layer possible1.00
    7. belong in the AST layer.0.99
    8. Changing AST0.96
    9. Relatively forgiving — this is where most wc1.00
    10. Changing HIR1.00
    11. Discuss with at least one person who works1.00
    12. Changing TIR1.00
    13. Discuss with at least one person who works0.99
    14. Changing MIR or Emit1.00
    15. Discuss with at least two people. You are al0.99
    16. making a mistake unless you have a very sp1.00
    17. Adding a new layer1.00
    18. Requires explicit approval from the tech lea1.00
    19. contributor. No new layers without significa1.00
    20. AlEngi0.95
    21. aintru60.85
    22. World1.00
    23. Architecture.md1.00
    24. BAML1.00
    25. 'nginr0.93
    26. d'0.97
    27. fight slop with slop0.99
    28. toge0.95
    29. Engineering the future of Al1.00
    30. AlEngi0.95
    31. World1.00
    32. DFLA0.88

  • 2:41 #15 done16 line(s)

    shot 15·sharpness 2039.4

    1. Battle of Design1.00
    2. AlEngirieer0.97
    3. World's Fair0.97
    4. casualty: 5B+ tokens1.00
    5. AlEngine0.97
    6. intrust0.99
    7. World's1.00
    8. PBAML0.87
    9. jineer –0.90
    10. 'sF0.95
    11. fight slop with slop1.00
    12. toge1.00
    13. Engineering the future of Al0.99
    14. AlEngine0.95
    15. Vorld's0.95
    16. FLARE1.00

  • 2:52 #16 done17 line(s)

    shot 16·sharpness 1837.6

    1. AlEngineer1.00
    2. World's Fair0.99
    3. "Code can be slop,0.99
    4. writing cannot"0.99
    5. - me (with taste)1.00
    6. AlEngin0.95
    7. intrust1.00
    8. World'0.99
    9. PBAML0.85
    10. gineer —0.93
    11. I'sFi0.90
    12. fight slop with slop0.97
    13. oge0.99
    14. Engineering the future of Al0.99
    15. - AlEngin0.91
    16. lorld'0.94
    17. FLARE1.00

  • 2:56 #17 done146 line(s)

    shot 17·sharpness 2290.9

    1. BAML Enhancement Proposals1.00
    2. hellovai1.00
    3. AlEngineer0.98
    4. World's Fair0.97
    5. All Proposals0.97
    6. Export All BEPs1.00
    7. New BEP0.99
    8. Q Search BEPs...1.00
    9. Draft1.00
    10. Proposed1.00
    11. Pending1.00
    12. Accepted1.00
    13. Implemented1.00
    14. 181.00
    15. Rejected1.00
    16. BEP-0611.00
    17. BEP-0621.00
    18. BEP-0521.00
    19. BEP-0561.00
    20. BEP-0600.99
    21. BEP-0411.00
    22. Filesystem API Coherence1.00
    23. The Function Interface1.00
    24. Regex1.00
    25. Errors across bridges0.98
    26. CSV Standard Library1.00
    27. Async Iterator Protocol0.99
    28. Dhilan Shah0.96
    29. 371.00
    30. 2d ago1.00
    31. 2kai2kal20.96
    32. 口00.68
    33. 5/26/20261.00
    34. rossirpaulo1.00
    35. 口o0.62
    36. 4/27/20261.00
    37. 口00.66
    38. 2d ago0.95
    39. □60.89
    40. 5/26/20261.00
    41. □20.90
    42. 5d ago1.00
    43. BEP-0581.00
    44. BEP-0461.00
    45. BEP-0161.00
    46. BEP-0591.00
    47. Mock1.00
    48. BEP-0431.00
    49. Base64 and Base64URL0.98
    50. BEP-0571.00
    51. Catch Binding Syntax1.00
    52. Functions as tools1.00
    53. Avery Townsend1.00
    54. Core Type Methods1.00
    55. Standard Library1.00
    56. Associated Types for0.99
    57. kai1.00
    58. Johnny Godoy0.99
    59. □130.90
    60. 2d ago1.00
    61. 2kai2kai20.99
    62. rossirpaulo1.00
    63. Interfaces1.00
    64. 口00.65
    65. 4/20/20261.00
    66. □40.80
    67. 6/10/20261.00
    68. □40.73
    69. 5/7/20261.00
    70. □40.82
    71. 5/8/20261.00
    72. D50.86
    73. 6/4/20261.00
    74. BEP-0540.99
    75. BEP-0551.00
    76. Range1.00
    77. Compiler Attributes1.00
    78. 2kai2kai20.99
    79. BEP-0291.00
    80. Mutability Semantics1.00
    81. UUID Standard Library1.00
    82. BEP-0451.00
    83. BEP-0511.00
    84. Iterators1.00
    85. 2kai2kai21.00
    86. □80.80
    87. 5/22/20261.00
    88. 2kai2kai20.98
    89. rossirpaulo1.00
    90. 2kai2kai21.00
    91. D10.69
    92. 5/26/20261.00
    93. □20.79
    94. 6/9/20261.00
    95. □110.84
    96. 5/8/20261.00
    97. □50.84
    98. 6/9/20261.00
    99. BEP-0531.00
    100. BEP-0400.99
    101. Language Semantic Identity0.99
    102. BEP-0261.00
    103. BEP-0381.00
    104. BEP-0491.00
    105. Middleware Blocks1.00
    106. and Runtime Trace Model1.00
    107. print1.00
    108. Native json Type0.99
    109. String Interpolation1.00
    110. aintru1.00
    111. Work0.98
    112. AlEr0.88
    113. 口00.62
    114. 4/26/20261.00
    115. □120.94
    116. rossirpaulo1.00
    117. 5/21/20261.00
    118. aaronvg1.00
    119. □110.92
    120. 4/24/20261.00
    121. 口80.85
    122. 4/29/20261.00
    123. D60.95
    124. Antonio Sarosi1.00
    125. 2d ago1.00
    126. BEP-0050.96
    127. Values vs References Across1.00
    128. the Host Boundary1.00
    129. BEP-0500.97
    130. Metrics and Tracing IDs1.00
    131. BEP-0840.93
    132. Inferred Visibility (public /0.97
    133. kai0.97
    134. BEP-0211.00
    135. Dates and Times1.00
    136. BEP-0441.00
    137. Interfaces1.00
    138. BAML1.00
    139. IEnginer0.97
    140. Id's0.91
    141. fight slop with slop1.00
    142. tog0.77
    143. Engineering the future of Al0.99
    144. AlEr0.94
    145. Work0.99
    146. IDFLAF0.92

  • 3:05 #18 done77 line(s)

    shot 18·sharpness 3450.4

    1. BEP-050 Metrics and Tracing IDs0.99
    2. Copy1.00
    3. ☆0.80
    4. Proposed1.00
    5. AlEngineer0.99
    6. Shepherds: None assigned1.00
    7. 0 Add Stance0.91
    8. History1.00
    9. World's Fair0.98
    10. 三 On th0.88
    11. README1.00
    12. 461.00
    13. Summary1.00
    14. Summar0.99
    15. Motivatic0.98
    16. Aggregation1.00
    17. 21.00
    18. Probab0.99
    19. Function Rename Safety1.00
    20. Any function with a single non-deterministic component anywhere in its call stack becomes1.00
    21. The ga1.00
    22. infrast1.00
    23. Shared LLM-as-Judge1.00
    24. 41.00
    25. non-deterministic. Traditional testing (assert output == expected) breaks down — you0.98
    26. Curren1.00
    27. need statistical evaluation. Today, evaluation of BAML functions is done entirely outside the0.99
    28. Multi-Signal Scoring1.00
    29. Propose1.00
    30. language: separate eval scripts, YAML configs, dashboard rules. If you rename a function,1.00
    31. $id1.00
    32. Multi-Step Evaluation1.00
    33. the metric binding breaks silently.0.99
    34. Loca1.00
    35. Production to Eval0.99
    36. Scop0.98
    37. This proposal makes metrics a language-level concept in BAML. It introduces:1.00
    38. Retroactive Scoring0.99
    39. set1.00
    40. Sampling and Cost Control 20.98
    41. $id — an opt-in trace identity parameter that keeps return types clean, supports0.99
    42. rossirpaulo1.00
    43. May 201.00
    44. parent-child trace trees, and enables retroactive metric attachment1.00
    45. documenting the dag and that each1.00
    46. DAG model0.99
    47. Open Issues1.00
    48. metric blocks — co-located with the function they measure, type-checked against0.99
    49. function in the metric block fires1.00
    50. when their pre reqs are satisfied is a0.99
    51. risk. humans/agents can get1.00
    52. Decisions1.00
    53. its signature, with a reactive DAG model where timing is derived from data0.99
    54. confused of how async i/o might1.00
    55. All Comments1.00
    56. 651.00
    57. dependencies1.00
    58. affect this1.00
    59. Braintru0.96
    60. Wor1.00
    61. Al Assistant1.00
    62. Beta0.99
    63. assertions1.00
    64. after blocks on testsets (extending BEP-023) for multi-dimensional batch0.98
    65. "might affect* knowing that they0.99
    66. rossirpaulo0.99
    67. can trust the dag, etc. can be non-1.00
    68. intuitive, etc1.00
    69. May 201.00
    70. BAML0.99
    71. AlEnginee0.99
    72. rld's1.00
    73. fight slop with slop0.98
    74. tc0.96
    75. Engineering the future of Al1.00
    76. Wor1.00
    77. DUDFLA!0.88

  • 3:25 #19 done40 line(s)

    shot 19·sharpness 2574.5

    1. 90.54
    2. boundary-bot APPMay 18th at 11:47 PM0.98
    3. AlEngineer0.98
    4. New BEP Created: BEP-501.00
    5. World'sFair1.00
    6. Metrics and Tracing IDs0.98
    7. Status:draft1.00
    8. Author: @Vaibhav Gupta0.97
    9. Shepherds: None assigned1.00
    10. Last Updated: May 18 at 11:47 PM0.99
    11. 0 comments | View BEP-50 →0.99
    12. PRESENTED BY1.00
    13. beps.boundaryml.com1.00
    14. BEP Feedback1.00
    15. Microsoft1.00
    16. BAML Enhancement Proposals Feedback Application1.00
    17. 114 replies0.97
    18. # Also sent to the channel0.98
    19. Vaibhav Gupta May 18th at 11:53 PM0.98
    20. recommend taking a look here0.97
    21. for anyone interested. its not 100% done yet, but the direction +1.00
    22. many many examples are there in the sub-pages.0.99
    23. boundary-bot APPMay 20th at 11:40 AM0.98
    24. @Aaron added a discussion on BEP-50: Metrics and Tracing IDs0.99
    25. > $id - Trace Identity0.97
    26. AlEnginee0.98
    27. i think we should reserve underscore-underscore instead perhaps.1.00
    28. ntrust1.00
    29. World's1.00
    30. $ starts looking wonky0.98
    31. BAML1.00
    32. Discussion|May 20 at 11:40 AM0.96
    33. neer0.94
    34. sFa1.00
    35. fight slop with slop0.98
    36. oget0.96
    37. Engineering the future of Al1.00
    38. — AlEnginee0.96
    39. orld's0.88
    40. LARE1.00

  • 3:42 #20 done24 line(s)

    shot 20·sharpness 3954.5

    1. BEP-001-match0.98
    2. bep ×0.96
    3. BEP-002-universal-catch1.00
    4. AlEngineer1.00
    5. World's Fair0.99
    6. BEP-003-safe1.00
    7. bep1.00
    8. BEP-004-inferred-visibility-public-private1.00
    9. >BEP-005-values-vs-references-across-the-host-boundary0.99
    10. 11.00
    11. #!/usr/bin/env -S uv run --script0.99
    12. BEP-006-semantic-streaming-but-better1.00
    13. 21.00
    14. # /// script0.98
    15. BEP-007-typed-throws-contracts-and-inference-alternative-t1.00
    16. 31.00
    17. # requires-python = ">=3.10"0.98
    18. BEP-008-namespaces0.97
    19. BEP-009-streaming0.98
    20. 41.00
    21. # dependencies = ["httpx"]0.99
    22. PRESENTED BY1.00
    23. > BEP-010-symmetric-matchcatch-patterns-with-type-pattern-su0.99
    24. 51.00

  • 3:56 #21 done

    shot 21·sharpness 2031.9

  • 4:03 #22 done

    shot 22·sharpness 2146.3

  • 4:04 #23 done

    shot 23·sharpness 2815.4

The page's on-screen-text budget of 600 lines is spent, so the last cards in this grid list fewer lines than they hold. Narrow the page with ?frames= to read them.

Transcript

277 cues· 3,996 words· 21,228 chars

  1. 0:13 Fighting slop with slop.
  2. 0:16 My name is Vaibhav, and I'm going to talk about something that is a little, I would say, maybe a little silly at first.
  3. 0:23 I'm going to show you our team's engineering practices really quickly.
  4. 0:27 We do no code reviews.
  5. 0:28 We require every engineer to work on things in parallel.
  6. 0:32 And we have no standardization how people do AI.
  7. 0:35 And I know immediately what almost all of you are thinking.
  8. 0:38 We're probably a Zoom or YC startup.
  9. 0:41 And I can guarantee you I'm clearly a millennial.
  10. 0:44 So what do we actually do at our company without code reviews?
  11. 0:47 Well, about three years ago, we decided to build a programming language.
  12. 0:52 That's something that has absolutely no room for slop.
  13. 0:55 It's something that has to work every single time exactly the same way.
  14. 1:00 It's something that you can't just change a year later or a month later because you made a bad design decision.
  15. 1:05 You have to be correct.
  16. 1:06 And for the last three years, we've been in an onslaught of war against slot.
  17. 1:12 And when I first met this enemy, I went to my great mentor, Slop Sue, and he taught me something.
  18. 1:19 To defeat the slop, we must become the slop.
  19. 1:23 So we began, and we prepared, and then we started winning.
  20. 1:27 So when we think about it, what is slop?
  21. 1:30 Slop is just any code you don't read.
  22. 1:32 And whether any of you admit it or not, this is the least amount of slop that your code base will ever have.
  23. 1:39 Cherish it.
  24. 1:41 So we started fighting back against the slop, and we started fighting back with slop.
  25. 1:46 So how do we go ship a stable programming language with these engineering practices?
  26. 1:51 Well, the first skirmish we ever had was the skirmish of standards.
  27. 1:55 The hard part about hiring great engineers is you sadly can't tell them what to do.
  28. 1:59 Some of them want to use Claude, some of them want to use Codex, some of them want to use the latest thing that they just found on Hacker News.
  29. 2:05 So instead of trying to hold standards in our code base, we did something that is an invariant.
  30. 2:10 We built an architecture.md file.
  31. 2:13 Instead of using CloudMD, just pick something that every model can just understand.
  32. 2:17 This file has to be incredibly small, and it can only have things that will not change for months or for years.
  33. 2:24 In our case, it's the layers of the compiler.
  34. 2:26 You go deeper into the compiler, tell the agent to just talk to at least one other person.
  35. 2:31 That slows it down a little bit.
  36. 2:33 So now we have standards so anyone can use whatever they want.
  37. 2:36 But the real foe we faced was actually the battle of design.
  38. 2:40 Everyone here knows that you have to write perfect design docs.
  39. 2:43 And we have a very simple rule in our team.
  40. 2:46 Code can be slopped, writing cannot.
  41. 2:48 And of course, if I tell every engineer this, they write beautiful writing, and they handwrite everything, they don't use AI.
  42. 2:52 Well, sadly not.
  43. 2:54 So we built a design doc tool.
  44. 2:56 What this design doc tool does, it's a replacement for both Notion and GitHub effectively for design docs.
  45. 3:01 It allows versioning, commenting, all the stuff you want.
  46. 3:04 And obviously we do this, people use this.
  47. 3:05 Well, sadly not.
  48. 3:06 We built another tool on top of that.
  49. 3:09 And this tool was a Slack integration for that tool.
  50. 3:11 Every time a design doc got updated, this channel got notifications.

Chapters

  1. 0:00 Fighting slop with slop
  2. 0:27 Code reviews and invariants
  3. 2:08 Building rules that don't change
  4. 3:00 Design docs with notifications
  5. 4:03 CLI tools that catch where things break
  6. 5:08 Agents reading agent transcripts
  7. 5:57 Detecting hallucinations and errors
  8. 9:24 Attacking the foundational layer
  9. 11:03 Execution traces from first principles
  10. 14:26 Type safe tools across platforms
  11. 15:42 Making whole error classes impossible
  12. 17:57 BAML across languages
  13. 20:37 Go build the sloppy tools

Open at this second