vidtheque.

read-only demo

Videos SKDJo2CopRs

Why Eval++ Is the Next Great Compute Primitive — Sunil Pai & Matt Carey, Cloudflare

index_state ready data_status ok

AI Engineer· published 2026-06-08· 0:24:50· en-US· indexed 2026-08-11 11:13

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 0:20, 1 of 1 keyframes kept
  5. Shot 4, 0:20 to 0:30, 1 of 1 keyframes kept
  6. Shot 5, 0:30 to 0:58, 1 of 1 keyframes kept
  7. Shot 6, 0:58 to 1:26, 1 of 1 keyframes kept
  8. Shot 7, 1:26 to 1:54, 1 of 1 keyframes kept
  9. Shot 8, 1:54 to 2:22, 0 of 1 keyframes kept
  10. Shot 9, 2:22 to 2:50, 1 of 1 keyframes kept
  11. Shot 10, 2:50 to 2:57, 1 of 1 keyframes kept
  12. Shot 11, 2:57 to 3:23, 0 of 1 keyframes kept
  13. Shot 12, 3:23 to 3:49, 0 of 1 keyframes kept
  14. Shot 13, 3:49 to 4:15, 1 of 1 keyframes kept
  15. Shot 14, 4:15 to 4:41, 1 of 1 keyframes kept
  16. Shot 15, 4:41 to 5:07, 1 of 1 keyframes kept
  17. Shot 16, 5:07 to 5:33, 1 of 1 keyframes kept
  18. Shot 17, 5:33 to 5:59, 1 of 1 keyframes kept
  19. Shot 18, 5:59 to 6:25, 0 of 1 keyframes kept
  20. Shot 19, 6:25 to 6:51, 0 of 1 keyframes kept
  21. Shot 20, 6:51 to 7:17, 0 of 1 keyframes kept
  22. Shot 21, 7:17 to 7:22, 1 of 1 keyframes kept
  23. Shot 22, 7:22 to 7:23, 1 of 1 keyframes kept
  24. Shot 23, 7:23 to 7:26, 1 of 1 keyframes kept
  25. Shot 24, 7:26 to 7:29, 1 of 1 keyframes kept
  26. Shot 25, 7:29 to 7:32, 1 of 1 keyframes kept
  27. Shot 26, 7:32 to 7:35, 1 of 1 keyframes kept
  28. Shot 27, 7:35 to 7:38, 1 of 1 keyframes kept
  29. Shot 28, 7:38 to 7:40, 1 of 1 keyframes kept
  30. Shot 29, 7:40 to 7:54, 0 of 1 keyframes kept
  31. Shot 30, 7:54 to 7:57, 0 of 1 keyframes kept
  32. Shot 31, 7:57 to 8:18, 1 of 1 keyframes kept
  33. Shot 32, 8:18 to 8:48, 0 of 1 keyframes kept
  34. Shot 33, 8:48 to 8:55, 1 of 1 keyframes kept
  35. Shot 34, 8:55 to 9:35, 1 of 1 keyframes kept
  36. Shot 35, 9:35 to 9:53, 0 of 1 keyframes kept
  37. Shot 36, 9:53 to 10:19, 1 of 1 keyframes kept
  38. Shot 37, 10:19 to 10:44, 1 of 1 keyframes kept
  39. Shot 38, 10:44 to 11:10, 0 of 1 keyframes kept
  40. Shot 39, 11:10 to 11:36, 1 of 1 keyframes kept
  41. Shot 40, 11:36 to 12:01, 0 of 1 keyframes kept
  42. Shot 41, 12:01 to 12:27, 1 of 1 keyframes kept
  43. Shot 42, 12:27 to 12:53, 0 of 1 keyframes kept
  44. Shot 43, 12:53 to 13:19, 0 of 1 keyframes kept
  45. Shot 44, 13:19 to 13:44, 0 of 1 keyframes kept
  46. Shot 45, 13:44 to 14:10, 0 of 1 keyframes kept
  47. Shot 46, 14:10 to 14:36, 0 of 1 keyframes kept
  48. Shot 47, 14:36 to 15:02, 0 of 1 keyframes kept
  49. Shot 48, 15:02 to 15:27, 1 of 1 keyframes kept
  50. Shot 49, 15:27 to 15:53, 1 of 1 keyframes kept
  51. Shot 50, 15:53 to 16:19, 0 of 1 keyframes kept
  52. Shot 51, 16:19 to 16:45, 0 of 1 keyframes kept
  53. Shot 52, 16:45 to 17:10, 0 of 1 keyframes kept
  54. Shot 53, 17:10 to 17:36, 1 of 1 keyframes kept
  55. Shot 54, 17:36 to 18:02, 0 of 1 keyframes kept
  56. Shot 55, 18:02 to 18:03, 1 of 1 keyframes kept
  57. Shot 56, 18:03 to 18:29, 1 of 1 keyframes kept
  58. Shot 57, 18:29 to 18:56, 1 of 1 keyframes kept
  59. Shot 58, 18:56 to 19:22, 0 of 1 keyframes kept
  60. Shot 59, 19:22 to 19:48, 0 of 1 keyframes kept
  61. Shot 60, 19:48 to 20:14, 1 of 1 keyframes kept
  62. Shot 61, 20:14 to 20:40, 1 of 1 keyframes kept
  63. Shot 62, 20:40 to 21:06, 0 of 1 keyframes kept
  64. Shot 63, 21:06 to 21:32, 0 of 1 keyframes kept
  65. Shot 64, 21:32 to 21:58, 0 of 1 keyframes kept
  66. Shot 65, 21:58 to 22:25, 0 of 1 keyframes kept
  67. Shot 66, 22:25 to 22:51, 0 of 1 keyframes kept
  68. Shot 67, 22:51 to 23:17, 1 of 1 keyframes kept
  69. Shot 68, 23:17 to 23:43, 0 of 1 keyframes kept
  70. Shot 69, 23:43 to 24:09, 1 of 1 keyframes kept
  71. Shot 70, 24:09 to 24:35, 0 of 1 keyframes kept
  72. Shot 71, 24:35 to 24:50, 1 of 1 keyframes kept

72 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
449
whisperx 449
chunks
45
from 449 cues
keyframes
41
kept of 72 captured
frames with text
41
670 lines read
chapters
0
from the source metadata
keyframe bytes
8.6 MB
word timings on 449 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-11 11:09 1m 38s
stt done — 2026-08-11 11:10 28s
chunk done — 2026-08-11 11:11 0s
text_embed done — 2026-08-11 11:11 1s
keyframe done — 2026-08-11 11:11 2m 15s
ocr done — 2026-08-11 11:13 15s
frame_embed done — 2026-08-11 11:13 6s

Frames, and what the machine read

  • 0:03 #0 done2 line(s)

    shot 0·sharpness 662.1

    1. AlEngineer0.98
    2. EUROPE1.00

  • 0:08 #1 done2 line(s)

    shot 1·sharpness 814.9

    1. PRESENTINGSPONSOR1.00
    2. Google DeepMind1.00

  • 0:12 #2 done3 line(s)

    shot 2·sharpness 905.6

    1. PLATINUM SPONSORS0.98
    2. # Braintrust0.96
    3. WorkOS OpenAI0.94

  • 0:20 #3 done4 line(s)

    shot 3·sharpness 368.7

    1. The Platform For0.97
    2. Building Agents.1.00
    3. AlEngineer0.98
    4. EUROPE1.00

  • 0:24 #4 done30 line(s)

    shot 4·sharpness 1778.6

    1. npm i agents1.00
    2. The Platform For1.00
    3. Building Agents.1.00
    4. Build agents on Cloudflare — the platform designed for durable0.98
    5. execution, serverless inference, and pricing that scales up (and down).1.00
    6. AIE1.00
    7. Get Started1.00
    8. View Docs1.00
    9. ★1.00
    10. ★1.00
    11. Generative1.00
    12. Implement and open a pull request for the first issue1.00
    13. Implement and open a pull request for the first issue0.99
    14. I have assigned to me.1.00
    15. I have assigned to me.1.00
    16. Sure thing! I can guide you through implementing1.00
    17. Sure thing!1.00
    18. the isue and opening a pull rquest. First. I need0.99
    19. some details:1.00
    20. Look at issues assigned to you0.98
    21. Write code to implement issue0.99
    22. 1. What git repository are you using?0.99
    23. Open pull request0.99
    24. 2. What is the issue about?1.00
    25. 3. What programming language is involved?0.99
    26. Once I have this info, I can walk you through the0.99
    27. implementation and PR process.0.98
    28. Google DeepMind1.00
    29. AlEngineer1.00
    30. 0260.95

  • 0:55 #5 done4 line(s)

    shot 5·sharpness 405.9

    1. The Platform For1.00
    2. Building Agents.1.00
    3. AlEngineer0.98
    4. EUROPE1.00

  • 1:04 #6 done4 line(s)

    shot 6·sharpness 397.9

    1. The Platform For0.98
    2. Building Agents.1.00
    3. AlEngineer0.98
    4. EUROPE1.00

  • 1:51 #7 done4 line(s)

    shot 7·sharpness 390.1

    1. The Platform For1.00
    2. Building Agents.0.97
    3. IEngineer0.95
    4. EUROPE1.00

  • 2:00 #8 skipped

    shot 8·duplicate of #7

  • 2:39 #9 done4 line(s)

    shot 9·sharpness 390.1

    1. The Platform For1.00
    2. Building Agents.0.97
    3. Engineer1.00
    4. EUROPE1.00

  • 2:54 #10 done4 line(s)

    shot 10·sharpness 323.7

    1. The Pla0.98
    2. Buildins0.89
    3. Engineer1.00
    4. EUROPE1.00

  • 3:07 #11 skipped

    shot 11·duplicate of #9

  • 3:28 #12 skipped

    shot 12·duplicate of #7

  • 3:52 #13 done4 line(s)

    shot 13·sharpness 394.0

    1. The Platform For0.99
    2. Building Agents.0.97
    3. Engineer1.00
    4. EUROPE1.00

  • 4:30 #14 done2 line(s)

    shot 14·sharpness 390.2

    1. Engineer1.00
    2. EUROPE1.00

  • 4:54 #15 done2 line(s)

    shot 15·sharpness 391.6

    1. Engineer1.00
    2. EUROPE1.00

  • 5:20 #16 done2 line(s)

    shot 16·sharpness 397.1

    1. IEngineer0.96
    2. EUROPE1.00

  • 5:48 #17 done4 line(s)

    shot 17·sharpness 393.2

    1. The Platform For0.99
    2. Building Agents.0.97
    3. Engineer1.00
    4. EUROPE1.00

  • 6:10 #18 skipped

    shot 18·duplicate of #17

  • 6:28 #19 skipped

    shot 19·duplicate of #13

  • 7:07 #20 skipped

    shot 20·duplicate of #17

  • 7:21 #21 done35 line(s)

    shot 21·sharpness 1850.5

    1. History Bookmarks Profiles Tab Window Help0.99
    2. Thu 9 Apr 14:191.00
    3. agents.cloudflar.com0.97
    4. Work1.00
    5. The Platform For1.00
    6. Building Agents.1.00
    7. Build agents on Cloudflare — the platform designed for durable0.99
    8. ★1.00
    9. execution, serveriess inference, and pricing that scales up (and down).0.99
    10. AIE1.00
    11. Get Started1.00
    12. View Docs1.00
    13. ★1.00
    14. ★1.00
    15. ★1.00
    16. Generative1.00
    17. Implement and open a pull request for the first issue1.00
    18. Implement and open a pull request for the first issue1.00
    19. I have assigned to me.1.00
    20. I have assigned to me.0.99
    21. Sure thing! I can guide you through implementing1.00
    22. Sure thing!1.00
    23. the issue and opening a pull request. First, I need0.97
    24. some details:1.00
    25. Look at issues assigned to you0.99
    26. Write code to implement issue0.99
    27. 1. What git repository are you using?0.99
    28. Open pull request0.98
    29. 2. What is the issue about?1.00
    30. 3. What programming language is involved?1.00
    31. Once I have this info, I can walk you through the0.99
    32. implementation and PR process.0.99
    33. Google DeepMind1.00
    34. Engineer0.98
    35. 0260.86

  • 7:22 #22 done19 line(s)

    shot 22·sharpness 1399.1

    1. File0.96
    2. Edit1.00
    3. Thu 9 Apr 14:190.99
    4. C Al Mode0.89
    5. Work0.99
    6. pertyserver1.00
    7. pertylt0.97
    8. Google1.00
    9. #agents0.92
    10. Q Search Google or type a URL0.99
    11. t Al Mode0.87
    12. AIE1.00
    13. *0.75
    14. ★1.00
    15. Managesd by your orgenisation | Black0.75
    16. Customise Chrome1.00
    17. Google DeepMind1.00
    18. Engineer0.96
    19. 0260.78

  • 7:25 #23 done57 line(s)

    shot 23·sharpness 2449.1

    1. History1.00
    2. Bookmarks1.00
    3. Profiles Tab0.96
    4. Window Help0.99
    5. Thu 9 Apr 14:190.99
    6. XXO0.68
    7. blog.cloudflare.com0.99
    8. ☆0.87
    9. Work0.99
    10. AI0.74
    11. Developers1.00
    12. Radar0.99
    13. Product News1.00
    14. Security1.00
    15. Policy & Legal1.00
    16. Zero Trust1.00
    17. Speed & Relabity0.97
    18. Ufe at Cloudfiare0.93
    19. Partners0.98
    20. From bytecode to bytes: automated1.00
    21. magic packet generation1.00
    22. 2026-04-081.00
    23. By applying symbolic execution and the Z3 theorem prover to1.00
    24. ★1.00
    25. BPF bytecode, we've automated the generation of malware0.99
    26. ★0.99
    27. AIE1.00
    28. trigger packets, cutting analysis time from hours to seconds...0.99
    29. ★1.00
    30. ★1.00
    31. Continue reading »1.00
    32. ★1.00
    33. ★1.00
    34. ★1.00
    35. Axel Boesenach1.00
    36. Cloudflare targets 2029 for full post-0.98
    37. How we built Organizations to help1.00
    38. quantum security1.00
    39. enterprises manage Cloudflare at1.00
    40. 2026-04-071.00
    41. scale1.00
    42. Post-Quantum Security1.00
    43. 2026-04-061.00
    44. Recent advances in quantum hardware and software have accelerated the0.98
    45. Identity Enterprise0.98
    46. timeline on which quantum attack might happen. Cloudflare is responding by0.99
    47. moving our target for full post-quantum security to 2029....0.97
    48. Cloudflare Organizations is now in public beta, introducing a new0.99
    49. management layer for enterprise customers with multiple accounts. Learn1.00
    50. how we consolidated our authorization systems to enable org-wide1.00
    51. Bas Westerbaan1.00
    52. management1.00
    53. Justin Hutchings1.00
    54. Adam Bouhmad1.00
    55. Google DeepMind1.00
    56. Engineer0.98
    57. 0260.81

Transcript

449 cues· 4,305 words· 22,456 chars

  1. 0:15 Yeah, well, welcome.
  2. 0:16 I'm Matt and this is Sunil.
  3. 0:18 We work on agents at Cloudflare.
  4. 0:20 This, this thing.
  5. 0:21 So yeah, maybe you want to introduce yourself a little bit better than me.
  6. 0:25 This is Matt Carey.
  7. 0:27 I only hire my friends for our team.
  8. 0:30 Say something.
  9. 0:31 Make sure it's on the microphone.
  10. 0:32 Yeah, yeah.
  11. 0:34 Is mine on?
  12. 0:34 Like, I keep thinking about Cloudflare as a friends and family company.
  13. 0:38 We love doing favors, all our closest friends.
  14. 0:41 No one pays for Cloudflare bills, which is already dirt cheap, by the way.
  15. 0:44 It's ridiculous.
  16. 0:46 So we started building out AI agents seriously a little over a year ago, and it's gone remarkably well.
  17. 0:53 How many folks here are like orange-pilled?
  18. 0:55 Which is to say, durable objects.
  19. 0:59 Oh yeah, like worst named technology, but incredible, right?
  20. 1:04 Everyone's like, it's a database.
  21. 1:05 I'm like, no, but it has a database.
  22. 1:07 And I'm like, ugh.
  23. 1:09 But we started using that, and it turned out to work out really well.
  24. 1:12 There's some unique capabilities about it.
  25. 1:15 Some might call it vendor lock-in.
  26. 1:16 Some might call it innovation.
  27. 1:19 And ever since then, the team has grown.
  28. 1:23 We are based in, like, most of us are in London.
  29. 1:25 Well, you're British, but you just moved to Lisbon, which is amazing, by the way.
  30. 1:29 Folks, you spend your days drinking wine, and you do fish, potatoes, bread.
  31. 1:34 It's just king's life.
  32. 1:36 I live in Newcastle.
  33. 1:38 The weather is way worse.
  34. 1:40 But I do like it.
  35. 1:41 I tried watching Geordie Shore.
  36. 1:43 I regret doing that.
  37. 1:45 No, that's not what I think of.
  38. 1:46 Anyway, so we've been building agents and- Back on topic.
  39. 1:49 Sorry, yes, of course.
  40. 1:51 I don't know, we can tell you about our journey so far, some of the tech we're using, some of the new upcoming stuff.
  41. 1:57 Thankfully, none of this is recorded, so we can tell you about some secret features that are coming out.
  42. 2:03 But I'd love to put it out there for a couple of minutes if you have any questions.
  43. 2:07 Otherwise, we have like five topics we thought would be interesting to talk about.
  44. 2:11 But what do you want to know about Cloudflare and AI agents?
  45. 2:16 Amazing.
  46. 2:20 So it starts with durable objects, okay?
  47. 2:22 So you know how everyone believes, the way you build servers nowadays is you write a function that takes a request and returns a response, right?
  48. 2:31 Simplest model.
  49. 2:32 And you decide you want to do, you want to build a hit counter for your website.
  50. 2:36 So what you do is you say let x equals zero, or let counter equals zero, and for every request you say counter plus plus.

Open at this second