vidtheque.

read-only demo

Videos bRnoEpoK5m4

The Pipeline Is Dead - Iris ten Teije, Sky Valley Ambient Computing

index_state ready data_status ok

AI Engineer· published 2026-07-07· 0:19:49· en-US· indexed 2026-08-11 05:06

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:11, 1 of 1 keyframes kept
  2. Shot 1, 0:11 to 0:20, 1 of 1 keyframes kept
  3. Shot 2, 0:20 to 0:33, 1 of 1 keyframes kept
  4. Shot 3, 0:33 to 1:01, 1 of 1 keyframes kept
  5. Shot 4, 1:01 to 1:30, 0 of 1 keyframes kept
  6. Shot 5, 1:30 to 1:55, 1 of 1 keyframes kept
  7. Shot 6, 1:55 to 2:21, 0 of 1 keyframes kept
  8. Shot 7, 2:21 to 2:46, 0 of 1 keyframes kept
  9. Shot 8, 2:46 to 3:19, 1 of 1 keyframes kept
  10. Shot 9, 3:19 to 3:52, 0 of 1 keyframes kept
  11. Shot 10, 3:52 to 4:34, 1 of 1 keyframes kept
  12. Shot 11, 4:34 to 5:00, 1 of 1 keyframes kept
  13. Shot 12, 5:00 to 5:27, 1 of 1 keyframes kept
  14. Shot 13, 5:27 to 5:55, 0 of 1 keyframes kept
  15. Shot 14, 5:55 to 6:22, 0 of 1 keyframes kept
  16. Shot 15, 6:22 to 6:49, 0 of 1 keyframes kept
  17. Shot 16, 6:49 to 7:09, 1 of 1 keyframes kept
  18. Shot 17, 7:09 to 7:46, 1 of 1 keyframes kept
  19. Shot 18, 7:46 to 8:13, 1 of 1 keyframes kept
  20. Shot 19, 8:13 to 8:40, 1 of 1 keyframes kept
  21. Shot 20, 8:40 to 9:07, 0 of 1 keyframes kept
  22. Shot 21, 9:07 to 9:37, 1 of 1 keyframes kept
  23. Shot 22, 9:37 to 10:07, 1 of 1 keyframes kept
  24. Shot 23, 10:07 to 10:37, 0 of 1 keyframes kept
  25. Shot 24, 10:37 to 11:14, 1 of 1 keyframes kept
  26. Shot 25, 11:14 to 11:39, 1 of 1 keyframes kept
  27. Shot 26, 11:39 to 11:48, 1 of 1 keyframes kept
  28. Shot 27, 11:48 to 11:51, 0 of 1 keyframes kept
  29. Shot 28, 11:51 to 12:07, 1 of 1 keyframes kept
  30. Shot 29, 12:07 to 12:17, 1 of 1 keyframes kept
  31. Shot 30, 12:17 to 12:35, 1 of 1 keyframes kept
  32. Shot 31, 12:35 to 13:12, 1 of 1 keyframes kept
  33. Shot 32, 13:12 to 13:43, 1 of 1 keyframes kept
  34. Shot 33, 13:43 to 14:12, 1 of 1 keyframes kept
  35. Shot 34, 14:12 to 14:42, 0 of 1 keyframes kept
  36. Shot 35, 14:42 to 15:08, 0 of 1 keyframes kept
  37. Shot 36, 15:08 to 15:33, 1 of 1 keyframes kept
  38. Shot 37, 15:33 to 15:58, 0 of 1 keyframes kept
  39. Shot 38, 15:58 to 16:24, 1 of 1 keyframes kept
  40. Shot 39, 16:24 to 16:49, 1 of 1 keyframes kept
  41. Shot 40, 16:49 to 17:15, 1 of 1 keyframes kept
  42. Shot 41, 17:15 to 17:40, 0 of 1 keyframes kept
  43. Shot 42, 17:40 to 18:05, 1 of 1 keyframes kept
  44. Shot 43, 18:05 to 18:22, 1 of 1 keyframes kept
  45. Shot 44, 18:22 to 18:58, 1 of 1 keyframes kept
  46. Shot 45, 18:58 to 19:28, 0 of 1 keyframes kept
  47. Shot 46, 19:28 to 19:49, 1 of 1 keyframes kept

47 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
213
whisperx 213
chunks
36
from 213 cues
keyframes
32
kept of 47 captured
frames with text
32
448 lines read
chapters
0
from the source metadata
keyframe bytes
2.8 MB
word timings on 213 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-11 05:03 1m 25s
stt done — 2026-08-11 05:04 18s
chunk done — 2026-08-11 05:04 0s
text_embed done — 2026-08-11 05:04 1s
keyframe done — 2026-08-11 05:04 1m 01s
ocr done — 2026-08-11 05:05 11s
frame_embed done — 2026-08-11 05:06 5s

Frames, and what the machine read

  • 0:07 #0 done1 line(s)

    shot 0·sharpness 647.8

    1. DISTRIBUTION1.00

  • 0:19 #1 done19 line(s)

    shot 1·sharpness 372.6

    1. 日0.88
    2. DOCKER1.00
    3. JENKINS0.99
    4. NPM1.00
    5. GITHUB ACTIONS1.00
    6. KUBERNETES1.00
    7. CDN1.00
    8. HELM1.00
    9. X0.54
    10. 20.97
    11. HELM1.00
    12. PLAY STORE1.00
    13. TESTFLIGHT1.00
    14. APP STORE1.00
    15. C0.58
    16. VERCEL1.00
    17. CLOUDFLARE1.00
    18. AUTO-UPDATERS0.99
    19. BLUE-GREEN DEPLOYS1.00

  • 0:26 #2 done7 line(s)

    shot 2·sharpness 341.4

    1. 業0.69
    2. BUILT HERE1.00
    3. SOURCE CODE0.99
    4. THE APP0.94
    5. RUN HERE1.00
    6. FROZEN ARTIFACT1.00
    7. USER DEVICE1.00

  • 0:44 #3 done3 line(s)

    shot 3·sharpness 974.0

    1. THE PIPELINE IS DEAD0.99
    2. SOFTWARE DISTRIBUTION WHEN THE AGENT IS THE RUNTIME1.00
    3. Iris ten Teije· Differ0.99

  • 1:21 #4 skipped

    shot 4·duplicate of #3

  • 1:45 #5 done1 line(s)

    shot 5·sharpness 259.0

    1. COST1.00

  • 2:15 #6 skipped

    shot 6·duplicate of #5

  • 2:34 #7 skipped

    shot 7·duplicate of #5

  • 3:03 #8 done1 line(s)

    shot 8·sharpness 420.7

    1. NOBODY CHOSE THIS0.99

  • 3:38 #9 skipped

    shot 9·duplicate of #8

  • 3:57 #10 done17 line(s)

    shot 10·sharpness 304.1

    1. OBSERVED1.00
    2. mara.chengacme.com0.98
    3. last 98 days0.97
    4. "Quick actions" panel was never opened – 0 interactions across 214 sessions.0.98
    5. → Removing panel. Reclaiming 320px for content.0.97
    6. src/layouts/Dashboard.tsx1.00
    7. 12 function Dashboard({ user }: Props) {0.99
    8. 13 return (0.98
    9. 14 <div className="grid grid-cols-[1fr_320px] gap-6">1.00
    10. <div className="grid grid-cols-1">0.99
    11. 15 <MainContent />1.00
    12. - 16 <QuickActions /> {/* 0 opens · 214 sessions */}0.97
    13. 17 </div>0.97
    14. 18)0.96
    15. 19}0.97
    16. ON DEMAND0.99
    17. AT THE POINT OF USE0.97

  • 4:40 #11 done2 line(s)

    shot 11·sharpness 387.4

    1. DEVELOPMENT1.00
    2. DISTRIBUTION1.00

  • 5:24 #12 done13 line(s)

    shot 12·sharpness 245.0

    1. THE DEMAND IS OLD. THE RATIONING CHANGED.0.99
    2. 011.00
    3. EXHIBIT1.00
    4. 021.00
    5. EXHIBIT1.00
    6. 031.00
    7. EXHIBIT1.00
    8. PROFESSIONAL SERVICES1.00
    9. DOTFILES1.00
    10. EXCEL1.00
    11. THE PRICE OF FIT0.99
    12. DEMAND LEAKING OUT1.00
    13. SUBSTRATE, NOT A PROGRAM0.99

  • 5:33 #13 skipped

    shot 13·duplicate of #12

  • 6:06 #14 skipped

    shot 14·duplicate of #12

  • 6:36 #15 skipped

    shot 15·duplicate of #12

  • 6:52 #16 done5 line(s)

    shot 16·sharpness 146.9

    1. CONVERGENT INTERFACE1.00
    2. LIKE1.00
    3. REPLY1.00
    4. SHARE1.00
    5. DIVERGENT BEHAVIOR0.99

  • 7:41 #17 done14 line(s)

    shot 17·sharpness 805.6

    1. YOU CAN ONLY HAND-CODE SO MANY BUCKETS.1.00
    2. AI DOESN'T HAVE THAT PROBLEM.0.99
    3. 011.00
    4. 021.00
    5. 031.00
    6. FEATURE FLAGS1.00
    7. SEGMENTATION1.00
    8. ADAPTIVE1.00
    9. WHERE WE STARTED0.98
    10. WHERE WE GOT TO0.99
    11. WHERE THIS GOES0.94
    12. On / off for everyone0.96
    13. Buckets you declare in advance0.98
    14. The rest of this talk0.97

  • 8:04 #18 done1 line(s)

    shot 18·sharpness 397.4

    1. ONE ARTIFACT FOR EVERYONE1.00

  • 8:37 #19 done2 line(s)

    shot 19·sharpness 1002.6

    1. THE OBJECTION0.99
    2. "I CAN BARELY REASON ABOUT ONE AI-GENERATED CODEBASE. YOU WANT ME TO RUN A MILLION?"1.00

  • 8:46 #20 skipped

    shot 20·duplicate of #19

  • 9:19 #21 done4 line(s)

    shot 21·sharpness 374.0

    1. Change anywhere1.00
    2. Stem + isolated branches1.00
    3. → breaks anywhere0.99
    4. blast radius = one context1.00

  • 9:46 #22 done4 line(s)

    shot 22·sharpness 372.2

    1. Change anywhere1.00
    2. Stem + isolated branches1.00
    3. → breaks anywhere0.99
    4. blast radius = one context0.98

  • 10:22 #23 skipped

    shot 23·duplicate of #21

Transcript

213 cues· 2,710 words· 15,197 chars

  1. 0:00 As we're online today, you can't raise your hands, but just nod at your screen.
  2. 0:04 How many of you have spent a significant part of your career making software move from one computer to another?
  3. 0:12 CI pipelines, package registries, container images, app store reviews.
  4. 0:18 That entire stack exists to solve exactly one problem.
  5. 0:23 get a frozen artifact, dead code, from the machine where it was built to the machine where it runs, safely, reproducibly, once.
  6. 0:34 And here's what I think everyone is missing.
  7. 0:36 That entire stack is built around one idea that is so old that we stopped seeing it as a choice.
  8. 0:43 One version of your software for everyone.
  9. 0:46 We've shipped it that way for so long that almost nobody asks why anymore.
  10. 0:51 I'm Iris.
  11. 0:52 I'm one of the co-founders of Differ.
  12. 0:55 My co-founder Noam was the first engineer at JFrog.
  13. 0:58 He helped build the pipeline that I'm about to tell you is dying.
  14. 1:02 I came at it from the other end.
  15. 1:04 I spent a decade in FinTech.
  16. 1:06 I was early at a digital bank that we scaled and exited.
  17. 1:10 Shipping software in the environments least willing to tolerate what I'm about to propose.
  18. 1:15 So I totally get any reservations and I've had them myself as well.
  19. 1:21 However, I'm telling you, it's coming anyway.
  20. 1:23 And that's not a warning.
  21. 1:26 It's the best thing that's happened to software in a long time.
  22. 1:30 Every piece of distribution infrastructure that you've touched encodes the same assumption.
  23. 1:36 Software is produced in one place, it runs in another, and the thing in the middle, the artifact, is frozen.
  24. 1:45 And that assumption was correct.
  25. 1:47 For decades, it was just true.
  26. 1:50 Why?
  27. 1:52 Because producing a correct change was expensive.
  28. 1:55 It took skilled humans hours or days.
  29. 1:58 So you did it rarely.
  30. 1:59 It was a central event.
  31. 2:00 You verified it and froze it.
  32. 2:03 And you shipped that frozen thing to everyone.
  33. 2:06 So the one-way pipeline isn't arbitrary.
  34. 2:09 It's the direct consequence of production of software being expensive and risky.
  35. 2:15 And of course the Frozen artifact has some advantages.
  36. 2:19 You get reproducibility, reviewability, rollback.
  37. 2:25 And every guarantee that we lean on in production flows from one fact.
  38. 2:29 There is one artifact and it doesn't change after we ship it.
  39. 2:34 That's the deal.
  40. 2:35 One version for everyone, Frozen.
  41. 2:39 We got reliability.
  42. 2:41 But the price was that the software couldn't really be for anyone in particular.
  43. 2:47 And nobody ever made that decision.
  44. 2:49 There was never a meeting where someone put one version for everyone versus a version for each person and picked the first one.
  45. 2:59 And it wasn't because the second option, a version for each person was worse.
  46. 3:04 It just wasn't an option.
  47. 3:06 Giving two users different software meant forking the code base and hand maintaining both.
  48. 3:12 a version per user at any real skill wasn't really a viable option.
  49. 3:18 And one version where everyone never had to win an argument.
  50. 3:21 It was just how it was.

Open at this second