vidtheque.

read-only demo

Videos BiG2ssibKGc

Stop babysitting your agents... — Brandon Waselnuk, Unblocked

index_state ready data_status ok

AI Engineer· published 2026-05-26· 0:18:54· en-US· indexed 2026-08-10 19:54

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 0:45, 1 of 1 keyframes kept
  5. Shot 4, 0:45 to 1:12, 1 of 1 keyframes kept
  6. Shot 5, 1:12 to 1:39, 1 of 1 keyframes kept
  7. Shot 6, 1:39 to 2:05, 1 of 1 keyframes kept
  8. Shot 7, 2:05 to 2:31, 1 of 1 keyframes kept
  9. Shot 8, 2:31 to 2:57, 0 of 1 keyframes kept
  10. Shot 9, 2:57 to 3:23, 0 of 1 keyframes kept
  11. Shot 10, 3:23 to 3:48, 1 of 1 keyframes kept
  12. Shot 11, 3:48 to 4:14, 0 of 1 keyframes kept
  13. Shot 12, 4:14 to 4:40, 0 of 1 keyframes kept
  14. Shot 13, 4:40 to 5:05, 1 of 1 keyframes kept
  15. Shot 14, 5:05 to 5:31, 1 of 1 keyframes kept
  16. Shot 15, 5:31 to 5:57, 0 of 1 keyframes kept
  17. Shot 16, 5:57 to 6:25, 1 of 1 keyframes kept
  18. Shot 17, 6:25 to 6:54, 1 of 1 keyframes kept
  19. Shot 18, 6:54 to 7:23, 1 of 1 keyframes kept
  20. Shot 19, 7:23 to 7:51, 0 of 1 keyframes kept
  21. Shot 20, 7:51 to 8:19, 1 of 1 keyframes kept
  22. Shot 21, 8:19 to 8:47, 1 of 1 keyframes kept
  23. Shot 22, 8:47 to 9:15, 0 of 1 keyframes kept
  24. Shot 23, 9:15 to 9:43, 0 of 1 keyframes kept
  25. Shot 24, 9:43 to 10:11, 0 of 1 keyframes kept
  26. Shot 25, 10:11 to 10:12, 1 of 1 keyframes kept
  27. Shot 26, 10:12 to 10:15, 0 of 1 keyframes kept
  28. Shot 27, 10:15 to 10:56, 1 of 1 keyframes kept
  29. Shot 28, 10:56 to 11:46, 1 of 1 keyframes kept
  30. Shot 29, 11:46 to 12:34, 1 of 1 keyframes kept
  31. Shot 30, 12:34 to 13:06, 1 of 1 keyframes kept
  32. Shot 31, 13:06 to 13:39, 0 of 1 keyframes kept
  33. Shot 32, 13:39 to 14:27, 1 of 1 keyframes kept
  34. Shot 33, 14:27 to 14:48, 0 of 1 keyframes kept
  35. Shot 34, 14:48 to 15:09, 0 of 1 keyframes kept
  36. Shot 35, 15:09 to 15:26, 1 of 1 keyframes kept
  37. Shot 36, 15:26 to 15:51, 1 of 1 keyframes kept
  38. Shot 37, 15:51 to 16:17, 1 of 1 keyframes kept
  39. Shot 38, 16:17 to 16:44, 0 of 1 keyframes kept
  40. Shot 39, 16:44 to 16:45, 1 of 1 keyframes kept
  41. Shot 40, 16:45 to 16:51, 1 of 1 keyframes kept
  42. Shot 41, 16:51 to 16:52, 0 of 1 keyframes kept
  43. Shot 42, 16:52 to 16:53, 1 of 1 keyframes kept
  44. Shot 43, 16:53 to 16:58, 0 of 1 keyframes kept
  45. Shot 44, 16:58 to 17:06, 0 of 1 keyframes kept
  46. Shot 45, 17:06 to 17:13, 1 of 1 keyframes kept
  47. Shot 46, 17:13 to 17:16, 0 of 1 keyframes kept
  48. Shot 47, 17:16 to 17:49, 1 of 1 keyframes kept
  49. Shot 48, 17:49 to 18:04, 1 of 1 keyframes kept
  50. Shot 49, 18:04 to 18:39, 1 of 1 keyframes kept
  51. Shot 50, 18:39 to 18:53, 1 of 1 keyframes kept

51 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
328
whisperx 328
chunks
34
from 328 cues
keyframes
33
kept of 51 captured
frames with text
33
1,132 lines read
chapters
0
from the source metadata
keyframe bytes
5.4 MB
word timings on 328 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 18:03 1m 55s
stt done — 2026-08-10 18:05 24s
chunk done — 2026-08-10 18:05 0s
text_embed done — 2026-08-10 19:54 1s
keyframe done — 2026-08-10 18:05 1m 30s
ocr done — 2026-08-10 18:07 23s
frame_embed done — 2026-08-10 19:54 5s

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 829.1

    1. PRESENTINGSPONSOR1.00
    2. Google DeepMind1.00

  • 0:13 #2 done3 line(s)

    shot 2·sharpness 907.7

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

  • 0:42 #3 done11 line(s)

    shot 3·sharpness 3033.2

    1. CONTEXT ENGINEERING0.98
    2. Stop babysitting your agents: building a0.99
    3. AIE0.98
    4. context engine for mergeable code1.00
    5. ★1.00
    6. Brandon Waselnuk1.00
    7. Unblocked|getunblocked.com1.00
    8. AIE Europe 20260.97
    9. Engineering the future of Al0.98
    10. AlEngineer0.93
    11. 0260.71

  • 0:56 #4 done11 line(s)

    shot 4·sharpness 2009.5

    1. What we'll cover today1.00
    2. AIE1.00
    3. ★1.00
    4. ★1.00
    5. 31.00
    6. 31.00
    7. Myths1.00
    8. Lessons1.00
    9. Braintrust1.00
    10. WorkOS OpenAI0.97
    11. 0260.79

  • 1:25 #5 done7 line(s)

    shot 5·sharpness 1997.8

    1. Not long ago, you1.00
    2. AIE1.00
    3. were the context0.99
    4. engine.1.00
    5. AlEngineer0.97
    6. L0260.62
    7. EUROPE1.00

  • 1:57 #6 done20 line(s)

    shot 6·sharpness 3553.3

    1. Remember how you built context?1.00
    2. Day l at your job you had almost no context.0.98
    3. Over time, you accumulated it:0.98
    4. Mentorship1.00
    5. AIE1.00
    6. Planning & code reviews1.00
    7. Architecture decisions1.00
    8. ★1.00
    9. ★1.00
    10. ★1.00
    11. Incidents & outages0.98
    12. Experiments & rollouts0.99
    13. Pushing a lot of PRs0.97
    14. Now, you're good at yourjob because:0.98
    15. Asked good questions1.00
    16. Gathered accurate context0.98
    17. AlEngineer0.97
    18. AlEingineer0.93
    19. C0260.79
    20. EUROPE1.00

  • 2:18 #7 done32 line(s)

    shot 7·sharpness 2245.9

    1. Now your agents need a context engine0.99
    2. Levels of Al maturity in engineering organizations1.00
    3. AIE1.00
    4. ★1.00
    5. 50.99
    6. ★1.00
    7. ★1.00
    8. ★1.00
    9. 21.00
    10. Tab1.00
    11. Agent1.00
    12. Context1.00
    13. Parallel1.00
    14. MCP1.00
    15. Harness1.00
    16. Background1.00
    17. Agent1.00
    18. complete1.00
    19. IDE1.00
    20. engineering1.00
    21. agents1.00
    22. + skills0.97
    23. engineering1.00
    24. agents1.00
    25. teams1.00
    26. You are the context1.00
    27. Curated context0.98
    28. Context engine0.97
    29. Unblocked1.00
    30. Engineering the future of Al0.99
    31. Al lEngineer0.90
    32. 20260.78

  • 2:37 #8 skipped

    shot 8·duplicate of #7

  • 3:05 #9 skipped

    shot 9·duplicate of #7

  • 3:26 #10 done32 line(s)

    shot 10·sharpness 2238.2

    1. Now your agents need a context engine0.99
    2. Levels of Al maturity in engineering organizations0.99
    3. AIE1.00
    4. ★1.00
    5. 50.99
    6. ★1.00
    7. ★1.00
    8. ★1.00
    9. 21.00
    10. Tab1.00
    11. Agent1.00
    12. Context1.00
    13. Parallel1.00
    14. MCP1.00
    15. Harness1.00
    16. Background1.00
    17. Agent1.00
    18. complete1.00
    19. IDE1.00
    20. engineering1.00
    21. agents1.00
    22. + skills0.96
    23. engineering1.00
    24. agents1.00
    25. teams1.00
    26. You are the context1.00
    27. Curated context0.98
    28. Context engine0.98
    29. Unblocked1.00
    30. AlEngineer0.97
    31. AlEn0.84
    32. EUROPE1.00

  • 4:04 #11 skipped

    shot 11·duplicate of #6

  • 4:17 #12 skipped

    shot 12·duplicate of #7

  • 4:57 #13 done12 line(s)

    shot 13·sharpness 2663.5

    1. The problem:0.97
    2. access≠understanding1.00
    3. AIE1.00
    4. We connect agents to code, logs, docs, tickets,0.99
    5. ★1.00
    6. ★1.00
    7. and more...0.99
    8. Which gives us plausible output that often1.00
    9. compiles...1.00
    10. But it fails human review and real world0.99
    11. expectations.1.00
    12. Engineering the future of Al0.98

  • 5:28 #14 done30 line(s)

    shot 14·sharpness 3697.1

    1. The problem:0.99
    2. THE TASK1.00
    3. "Add adaptive thinking mode for Claude 4.6 models to a1.00
    4. access≠understanding1.00
    5. Kotlin SDK. The SDK is a monorepo with three modules:0.98
    6. direct API, Bedrock, and batch processing. Maintain0.99
    7. ★1.00
    8. backwards compatibility for existing callers using the1.00
    9. AIE1.00
    10. ★0.99
    11. We connect agents to code, logs, docs, tickets,0.99
    12. old budget_token method."1.00
    13. ★1.00
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. and more...0.95
    18. 1 feat: add adaptive thinking mode for Claude 4.6 models #8470.99
    19. Which gives us plausible output that often1.00
    20. All checks have passed1.00
    21. compiles...0.97
    22. SR1.00
    23. sarah-richards Senior Engineer1.00
    24. But it fails human review and real world0.99
    25. Changes requested. This auto-detects adaptive mode for 4.6 models,0.99
    26. expectations.1.00
    27. which breaks every existing caller using budget_token . Our convention0.99
    28. is opt-in for new features with backwards-compatible defaults.1.00
    29. AlEngineer0.97
    30. EUROPE1.00

  • 5:37 #15 skipped

    shot 15·duplicate of #14

  • 6:22 #16 done8 line(s)

    shot 16·sharpness 2287.7

    1. Three myths about building context for agents1.00
    2. *★★0.69
    3. AIE1.00
    4. 1. Naive RAG over my docs is a context engine.1.00
    5. ★1.00
    6. ★1.00
    7. Unblocked1.00
    8. Engineering the future of Al0.99

  • 6:40 #17 done8 line(s)

    shot 17·sharpness 2293.3

    1. Three myths about building context for agents1.00
    2. *★★0.67
    3. AIE1.00
    4. 1. Naive RAG over my docs is a context engine.1.00
    5. ★1.00
    6. ★0.99
    7. ★1.00
    8. Engineering the future of Al0.99

  • 7:19 #18 done9 line(s)

    shot 18·sharpness 2889.1

    1. Three myths about building context for agents1.00
    2. AIE1.00
    3. ★0.99
    4. 1. Naive RAG over my docs is a context engine.0.98
    5. ★1.00
    6. ★1.00
    7. 2. If I just connect enough MCPs, I'm done.0.97
    8. 3. A bigger context window will solve this.0.99
    9. Google DeepMind1.00

  • 7:48 #19 skipped

    shot 19·duplicate of #4

  • 7:55 #20 done21 line(s)

    shot 20·sharpness 2492.4

    1. CODE THAT COMPILES0.96
    2. AIE1.00
    3. ORIGINAL INTENT1.00
    4. What your agent can't see0.98
    5. ★1.00
    6. TEAM CONVENTIONS1.00
    7. ★1.00
    8. ★1.00
    9. ★1.00
    10. WHY WAS IT BUILT THIS WAY0.96
    11. PAST DECISIONS1.00
    12. MIGRATION PLANS1.00
    13. TESTING STANDARDS1.00
    14. REJECTED APPROACHES0.99
    15. SLACK DECISIONS1.00
    16. ARCHITECTURE RATIONALE0.98
    17. DESIGN DOC WITH TRADE OFFS0.98
    18. INCIDENT LEARNINGS0.98
    19. Unblocked1.00
    20. Braintrust1.00
    21. WorkOS OpenAI0.97

  • 8:23 #21 done15 line(s)

    shot 21·sharpness 3703.5

    1. Why you need a context engine:0.98
    2. ★0.53
    3. ★1.00
    4. AIE1.00
    5. ★0.99
    6. Understands who you are and which information actually matters.1.00
    7. ★1.00
    8. ★1.00
    9. ★1.00
    10. ★1.00
    11. Resolves conflicts between different sources (code, docs, Slack, tickets).0.99
    12. Respects permissions and governance.1.00
    13. Delivers the right context to the right model at the right time.0.99
    14. AlEngineer0.97
    15. EUROPE1.00

  • 9:12 #22 skipped

    shot 22·duplicate of #16

  • 9:34 #23 skipped

    shot 23·duplicate of #16

Transcript

328 cues· 3,873 words· 20,580 chars

  1. 0:15 Hello, everybody.
  2. 0:16 It's good to see you.
  3. 0:18 I'm Brandon.
  4. 0:18 I work at Unblocked.
  5. 0:20 It's a great place.
  6. 0:21 And my goal is to make it so that you don't have to babysit your agents anymore.
  7. 0:24 I'm sure we all have a different take on what that means.
  8. 0:26 What I think of is care and feeding.
  9. 0:28 Basically, agents, whenever you spawn it by typing Claude in your CLI, let's assume, whatever tool you may use, they exist.
  10. 0:35 And it's like a brilliant software engineer has just spawned.
  11. 0:38 And it knows nothing about what it needs to do.
  12. 0:40 It knows nothing about your org.
  13. 0:41 It's completely zero context in its head.
  14. 0:44 So typically what happens is people have to move through building that context, which we'll go through in the beginning.
  15. 0:50 But first what I'm gonna cover today pretty quickly is three myths about how you can stop babysitting your agents.
  16. 0:57 And then three lessons that we learned the hard way, building a context engine at Unblocked.
  17. 1:01 So our product basically provides this context for agents.
  18. 1:04 And we'll tell you a bit about how we built that, techniques to care about.
  19. 1:08 And I'll show you a repo we actually constructed at our workshop yesterday that will be open source at the end of the week, which has one component of what's in a context engine, which you can lift for yourself and bring into your org if you'd like.
  20. 1:19 So not long ago, you were the context engine.
  21. 1:23 If you think about that when you're writing code, you thought about everything.
  22. 1:26 You knew everything.
  23. 1:27 You figured it all out.
  24. 1:28 You were dealing with that.
  25. 1:29 And now what's weird is you're in a weird state where you are actually the context engine for your agents.
  26. 1:35 So a useful way to think about this is how did you build context when you showed up at a company?
  27. 1:39 So day one, you had probably nothing.
  28. 1:42 But you were really smart.
  29. 1:43 You finished school.
  30. 1:43 I don't know, maybe some self-learning.
  31. 1:45 Then over time, you accumulate context by doing stuff at work, meeting market, meeting your team, being like, here's a PR, getting it rejected.
  32. 1:53 All these good things built up a lot of your capabilities.
  33. 1:56 And then finally, you became very good at your job because you asked good questions and you knew how to gather accurate context and shred stuff that wasn't helpful for you.
  34. 2:07 This is where we're at right now.
  35. 2:10 Most people here, if you look at the bottom, is in the you are the context engine stage because you're either dealing with the early phases of AI, which was just fancy autocomplete, or you're in an agentic IDE where you're triggering every job.
  36. 2:22 What we see with all these businesses that work with us in their AI adoption is it's usually at a varying level of this.
  37. 2:27 This has been adapted from Basim Eldeth's work.
  38. 2:30 If you check him out later, great engineer.
  39. 2:32 But basically, this is the type of ladder that we're dreaming for.
  40. 2:35 And far on the side is like,
  41. 2:37 A dreamy future that maybe Codex figured out, I think.
  42. 2:40 I didn't see Ryan's talk, unfortunately.
  43. 2:42 But everyone else is kind of trying to get there.
  44. 2:44 In order to move through this, you want to get to the curated context layer.
  45. 2:49 That is typically what a lot of teams are doing by creating static repos.
  46. 2:52 So static stores have a bunch of context that says key things about their company.
  47. 2:56 These can include, of course, CloudMD files, AgentsMD, those types of tooling.
  48. 3:01 But usually people start to put a bunch of other key corporate context into an area that agents can access to pull data from.
  49. 3:07 The issue with that is those are static content pieces, of course, so someone has to maintain and update them, as well as they don't have the availability of actual raw runtime data.
  50. 3:19 There's just a bunch of information that engineers obviously need that don't go into these static layers that you're starting to see, which typically look like a file system.

Open at this second