vidtheque.

read-only demo

Videos so9l_MwS2yg

Your Attention Is the Bottleneck, Not Your Agents — Zack Proser, WorkOS

index_state ready data_status ok

AI Engineer· published 2026-06-11· 0:25:17· en-US· indexed 2026-08-10 19:52

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 0:25, 1 of 1 keyframes kept
  5. Shot 4, 0:25 to 0:34, 1 of 1 keyframes kept
  6. Shot 5, 0:34 to 0:59, 1 of 1 keyframes kept
  7. Shot 6, 0:59 to 1:30, 1 of 1 keyframes kept
  8. Shot 7, 1:30 to 2:01, 1 of 1 keyframes kept
  9. Shot 8, 2:01 to 2:40, 1 of 1 keyframes kept
  10. Shot 9, 2:40 to 2:53, 1 of 1 keyframes kept
  11. Shot 10, 2:53 to 3:05, 1 of 1 keyframes kept
  12. Shot 11, 3:05 to 3:33, 1 of 1 keyframes kept
  13. Shot 12, 3:33 to 4:01, 1 of 1 keyframes kept
  14. Shot 13, 4:01 to 4:33, 0 of 1 keyframes kept
  15. Shot 14, 4:33 to 5:11, 1 of 1 keyframes kept
  16. Shot 15, 5:11 to 5:57, 1 of 1 keyframes kept
  17. Shot 16, 5:57 to 6:22, 0 of 1 keyframes kept
  18. Shot 17, 6:22 to 6:48, 0 of 1 keyframes kept
  19. Shot 18, 6:48 to 7:01, 1 of 1 keyframes kept
  20. Shot 19, 7:01 to 7:33, 1 of 1 keyframes kept
  21. Shot 20, 7:33 to 8:13, 1 of 1 keyframes kept
  22. Shot 21, 8:13 to 8:41, 0 of 1 keyframes kept
  23. Shot 22, 8:41 to 9:08, 1 of 1 keyframes kept
  24. Shot 23, 9:08 to 9:36, 0 of 1 keyframes kept
  25. Shot 24, 9:36 to 10:15, 0 of 1 keyframes kept
  26. Shot 25, 10:15 to 10:16, 0 of 1 keyframes kept
  27. Shot 26, 10:16 to 10:43, 1 of 1 keyframes kept
  28. Shot 27, 10:43 to 11:10, 0 of 1 keyframes kept
  29. Shot 28, 11:10 to 11:36, 0 of 1 keyframes kept
  30. Shot 29, 11:36 to 12:02, 0 of 1 keyframes kept
  31. Shot 30, 12:02 to 12:29, 0 of 1 keyframes kept
  32. Shot 31, 12:29 to 12:55, 0 of 1 keyframes kept
  33. Shot 32, 12:55 to 13:22, 1 of 1 keyframes kept
  34. Shot 33, 13:22 to 13:23, 0 of 1 keyframes kept
  35. Shot 34, 13:23 to 13:31, 0 of 1 keyframes kept
  36. Shot 35, 13:31 to 13:58, 0 of 1 keyframes kept
  37. Shot 36, 13:58 to 14:28, 0 of 1 keyframes kept
  38. Shot 37, 14:28 to 14:57, 0 of 1 keyframes kept
  39. Shot 38, 14:57 to 15:29, 0 of 1 keyframes kept
  40. Shot 39, 15:29 to 15:55, 1 of 1 keyframes kept
  41. Shot 40, 15:55 to 16:22, 0 of 1 keyframes kept
  42. Shot 41, 16:22 to 16:48, 0 of 1 keyframes kept
  43. Shot 42, 16:48 to 17:19, 1 of 1 keyframes kept
  44. Shot 43, 17:19 to 17:34, 1 of 1 keyframes kept
  45. Shot 44, 17:34 to 18:01, 1 of 1 keyframes kept
  46. Shot 45, 18:01 to 18:29, 1 of 1 keyframes kept
  47. Shot 46, 18:29 to 18:56, 0 of 1 keyframes kept
  48. Shot 47, 18:56 to 19:24, 0 of 1 keyframes kept
  49. Shot 48, 19:24 to 19:51, 0 of 1 keyframes kept
  50. Shot 49, 19:51 to 20:04, 0 of 1 keyframes kept
  51. Shot 50, 20:04 to 20:19, 1 of 1 keyframes kept
  52. Shot 51, 20:19 to 20:23, 1 of 1 keyframes kept
  53. Shot 52, 20:23 to 20:27, 1 of 1 keyframes kept
  54. Shot 53, 20:27 to 20:39, 1 of 1 keyframes kept
  55. Shot 54, 20:39 to 20:57, 1 of 1 keyframes kept
  56. Shot 55, 20:57 to 21:07, 1 of 1 keyframes kept
  57. Shot 56, 21:07 to 21:13, 0 of 1 keyframes kept
  58. Shot 57, 21:13 to 21:38, 0 of 1 keyframes kept
  59. Shot 58, 21:38 to 22:04, 1 of 1 keyframes kept
  60. Shot 59, 22:04 to 22:29, 0 of 1 keyframes kept
  61. Shot 60, 22:29 to 22:55, 0 of 1 keyframes kept
  62. Shot 61, 22:55 to 23:20, 0 of 1 keyframes kept
  63. Shot 62, 23:20 to 23:45, 1 of 1 keyframes kept
  64. Shot 63, 23:45 to 24:11, 1 of 1 keyframes kept
  65. Shot 64, 24:11 to 24:36, 1 of 1 keyframes kept
  66. Shot 65, 24:36 to 25:02, 0 of 1 keyframes kept
  67. Shot 66, 25:02 to 25:16, 1 of 1 keyframes kept
  68. Shot 67, 25:16 to 25:16, 0 of 1 keyframes kept

68 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
264
whisperx 264
chunks
44
from 264 cues
keyframes
37
kept of 68 captured
frames with text
37
615 lines read
chapters
16
from the source metadata
keyframe bytes
6.8 MB
word timings on 264 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 14:54 2m 05s
stt done — 2026-08-10 14:56 38s
chunk done — 2026-08-10 14:57 0s
text_embed done — 2026-08-10 19:51 1s
keyframe done — 2026-08-10 14:57 2m 12s
ocr done — 2026-08-10 14:59 14s
frame_embed done — 2026-08-10 19:51 6s

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

  • 0:21 #3 done6 line(s)

    shot 3·sharpness 601.7

    1. UNTETHERED1.00
    2. PRODUCTIVITY1.00
    3. Staying Healthy, Creative, and Shipping in the Al Coding Era1.00
    4. Zack Proser0.99
    5. AlEngineer0.99
    6. EUROPE1.00

  • 0:30 #4 done12 line(s)

    shot 4·sharpness 1737.2

    1. Untathered Productivity - S0.95
    2. Ask Gemini0.96
    3. ①localhost:3031/10.92
    4. Work0.83
    5. UNTETHERED1.00
    6. PRODUCTIVITY1.00
    7. AIE1.00
    8. ★0.99
    9. Staying Healthy, Creative, and Shipping in the Al Coding Era1.00
    10. Zack Proser1.00
    11. Applied AI· WorkOS · @zackproser0.95
    12. Google DeepMind1.00

  • 0:56 #5 done38 line(s)

    shot 5·sharpness 1598.9

    1. Untathered Productivity - Si0.96
    2. Ask Gemini0.98
    3. ①localhost:3031/20.97
    4. a0.51
    5. Work0.99
    6. THE DAILY REALITY1.00
    7. Slack1.00
    8. #eng-requests0.98
    9. Slack-DMs0.96
    10. Sarah Chen1.00
    11. "@channel Auth fix deployed, But now SSO is broken0.98
    12. ★1.00
    13. accounts.1.00
    14. AIE1.00
    15. ★1.00
    16. GitHub0.95
    17. ★1.00
    18. GONE1.00
    19. ★1.00
    20. ★1.00
    21. ★1.00
    22. 6 files chapgedich4ZgnA20.88
    23. Merge conflicts with main0.99
    24. Slack1.00
    25. #incidents0.98
    26. Last updated: 4 hours ago1.00
    27. "Rollback complete but 3 customers saw errors. Postmortem1.00
    28. Terminal1.00
    29. needed."0.98
    30. $ claude status0.98
    31. Agent A: feature/auth - waiting for review0.99
    32. Agent B: fix/timeout - conflicts detected0.97
    33. Agent C: dashboard - 78% complete0.97
    34. ATTENTION1.00
    35. This is how most of us manage agent work right now.1.00
    36. Alt-tab. Read. Switch. Forget. Repeat. The tools got faster - we didn't.0.99
    37. Braintrust1.00
    38. WorkOS OpenAI0.97

  • 1:03 #6 done16 line(s)

    shot 6·sharpness 2208.1

    1. Untathered Productivlty- 510.89
    2. +Ask Geminl0.94
    3. localhost:3031/30.97
    4. Work0.99
    5. CLOSING THE LOOP0.98
    6. · Slack bot at WorkOS – anyone requests a blog draft, gets it in 90 seconds0.99
    7. *★★0.54
    8. ·Needed a bug fix – acronyms like SCIM and SSO getting mangled0.99
    9. AIE1.00
    10. ★1.00
    11. ★1.00
    12. ·Normally: that window chaos you just saw0.98
    13. · Instead: gave Claude Code direct access to Slack - it fixed, deployed, tested, and0.98
    14. verified without me0.97
    15. Braintrust1.00
    16. WorkOS OpenAI0.96

  • 1:52 #7 done17 line(s)

    shot 7·sharpness 2117.7

    1. Untathered Productivity - S0.97
    2. +Ask Geminl0.92
    3. ①localhost:3031/30.97
    4. Work0.99
    5. CLOSING THE LOOP0.99
    6. · Slack bot at WorkOS – anyone requests a blog draft, gets it in 90 seconds0.98
    7. ·Needed a bug fix – acronyms like SCIM and SSO getting mangled0.97
    8. AIE1.00
    9. ★1.00
    10. ★1.00
    11. ★1.00
    12. ★1.00
    13. · Normally: that window chaos you just saw0.98
    14. · Instead: gave Claude Code direct access to Slack - it fixed, deployed, tested, and0.98
    15. verified without me0.97
    16. AlEngineer0.96
    17. EUROPE1.00

  • 2:28 #8 done50 line(s)

    shot 8·sharpness 2098.1

    1. Untathered Productivity - Sl0.96
    2. Ask Gemini0.98
    3. ①localhost:3031/40.96
    4. Work1.00
    5. CLOSING THE LOOP In Action0.98
    6. Claude Code1.00
    7. #blog-drafts0.98
    8. via Claude Slack Connector1.00
    9. ~/blog-bot $ claude0.99
    10. zack1.00
    11. just now0.99
    12. *★★0.55
    13. ★1.00
    14. > Fix the sentence case enforcer - it's0.98
    15. Write a post about how SCIM provisioning and SSO configuration work0.99
    16. AIE1.00
    17. ★1.00
    18. uppercasing acronyms like "SCIM" and "SSO"0.99
    19. together for enterprise User Management.0.99
    20. ★1.00
    21. "Reading src/pipeline/sentence-case.ts ...0.97
    22. " Found the bug: regex doesn't preserve0.97
    23. Sent using @Claude0.98
    24. ★1.00
    25. *0.87
    26. ★1.00
    27. known acronyms in ACRONYM_LIST0.99
    28. Edit src/pipeline/sentence-case.ts1.00
    29. Blog Bot1.00
    30. Classifying your input0.98
    31. + const preserved = ACRONYM_LIST.filter(0.99
    32. V0.55
    33. Generating draft with Claude Opus0.97
    34. + a → input.toUpperCase().includes(a)0.99
    35. Reviewing technical accuracy1.00
    36. Removing Al writing tells1.00
    37. Enforcing sentence case ← SCIM + SSO preserved0.98
    38. ✓ Tests passing (14/14)0.96
    39. Running voice and style checks0.99
    40. Deployed to Cloudflare Workers1.00
    41. Delivering to Slack0.99
    42. > Now testing via Slack...0.96
    43. Draft ready - How SCIM provisioning and SSO configuration work together0.99
    44. "Your app works great for small teams. Then a Fortune 500 prospect asks: 'Do1.00
    45. you support SCIM provisioning? Can our IT admin configure SSO1.00
    46. themselves?'...*0.96
    47. Category: guides-Slug: scim-sso-enterprise-user-management0.99
    48. Claude Code fixed the bug → deployed to CF → tested via Slack → acronyms preserved. The loop verified itself.0.99
    49. Replay1.00
    50. Engineering the future of Al0.99

  • 2:48 #9 done9 line(s)

    shot 9·sharpness 1353.0

    1. Untethered Productivity - Sii0.93
    2. Ask Gemini0.96
    3. ①localhost:3031/50.97
    4. Work0.99
    5. AIE1.00
    6. That felt incredible.1.00
    7. ★1.00
    8. It also scared me.1.00
    9. Engineering the future of Al1.00

  • 2:56 #10 done11 line(s)

    shot 10·sharpness 1329.7

    1. Untethered Productivity - Si0.96
    2. Ask Gemini0.96
    3. localhost:3031/61.00
    4. Work0.94
    5. The tools are nuclear.0.98
    6. AIE1.00
    7. Your nervous system is relatively ancient.0.99
    8. ★1.00
    9. ★1.00
    10. This is the tension. The answer is developer balance.1.00
    11. Engineering the future of Al1.00

  • 3:30 #11 done34 line(s)

    shot 11·sharpness 1736.0

    1. Untethered Productivity - Si0.96
    2. Ask Gemini0.97
    3. ①localhost:3031/70.97
    4. ☆0.86
    5. B0.51
    6. a0.59
    7. Work1.00
    8. SO WHY NOT STACK 150 LOOPS?1.00
    9. ACTIVE AGENT LOOPS1.00
    10. Auth refactor1.00
    11. Dashboard UI0.98
    12. API migration1.00
    13. Test suite1.00
    14. ★1.00
    15. AIE1.00
    16. Blog post pipeline1.00
    17. Onboarding flow1.00
    18. Billing integration1.00
    19. Search indexer0.99
    20. I can't track any of these1.00
    21. ★1.00
    22. ★1.00
    23. ★1.00
    24. ★1.00
    25. Email templates1.00
    26. CI/CD pipeline1.00
    27. Monitoring alerts1.00
    28. SDK update1.00
    29. 12 loops. Approving without1.00
    30. reading. This is the trap.1.00
    31. Loops running: 12 Completed: 100.99
    32. Replay the spiral0.98
    33. AlEngineer0.96
    34. EUROPE1.00

  • 3:37 #12 done21 line(s)

    shot 12·sharpness 1747.9

    1. Untethered Productivity - Si0.95
    2. Ask Germini0.89
    3. localhost:3031/80.97
    4. Work1.00
    5. The agents aren't the bottleneck.0.99
    6. We are.1.00
    7. AIE1.00
    8. ★1.00
    9. ★1.00
    10. ★1.00
    11. Agents1.00
    12. Verification0.99
    13. Your attention0.99
    14. Scale infinitely. Never tired. Don't context-0.99
    15. Can be automated. Lint, test, screenshot,1.00
    16. Fixed. Finite. Degrades under load. This is1.00
    17. switch.1.00
    18. review gates.1.00
    19. the constraint.1.00
    20. AlEngineer0.97
    21. EUROPE1.00

  • 4:29 #13 skipped

    shot 13·duplicate of #8

  • 4:55 #14 done20 line(s)

    shot 14·sharpness 2750.6

    1. Untathered Productivity - Si0.93
    2. Ask Gemini0.96
    3. ①localhost:3031/100.97
    4. Work0.99
    5. DEVELOPER BALANCE1.00
    6. We need to bring the human developer into balance with this new0.99
    7. *★★0.57
    8. ★1.00
    9. way of working.0.99
    10. AIE1.00
    11. ★0.91
    12. ★1.00
    13. It has never been faster to burn yourself out. Getting this right is critical.0.99
    14. ★1.00
    15. ★★0.95
    16. Agents for what they do best: speed, tireless execution, pattern detection.1.00
    17. Us for what we still do best: judgment, taste, and knowing when to stop.0.99
    18. Build verification into every loop so you can walk away with confidence.0.99
    19. Here's the stack that lets me do that.1.00
    20. Engineering the future of Al1.00

  • 5:16 #15 done23 line(s)

    shot 15·sharpness 1983.5

    1. Untathered Productivity - Si0.88
    2. Ask Gemini0.97
    3. ①localhost:3031/110.97
    4. a0.67
    5. Work0.98
    6. THE ENABLING STACK1.00
    7. Four capabilities that make untethered work possible:0.99
    8. *★*0.54
    9. AIE1.00
    10. Signal Layers0.95
    11. Voice-First Flows1.00
    12. ★1.00
    13. ★1.00
    14. ★1.00
    15. ★1.00
    16. Never open Slack. Only see the delta.0.98
    17. Dictate intent. Your voice is the new keyboard.1.00
    18. Remote Control0.99
    19. The System Improves Itself0.98
    20. Walk away. Agents keep working on your machine.0.99
    21. Finds your patterns. Recommends what to automate.0.99
    22. Let me walk through each one.0.99
    23. Engineering the future of Al1.00

  • 6:07 #16 skipped

    shot 16·duplicate of #7

  • 6:33 #17 skipped

    shot 17·duplicate of #15

  • 7:00 #18 done42 line(s)

    shot 18·sharpness 1731.6

    1. Untathered Productivity - Sl0.95
    2. Ask Geminl0.95
    3. ①localhost:3031/120.97
    4. a0.57
    5. Work0.99
    6. SIGNAL LAYERS0.97
    7. Slack1.00
    8. Linear1.00
    9. Slack MCP0.92
    10. Linear MCP0.98
    11. AIE-1421.00
    12. ★1.00
    13. Claude Code1.00
    14. Auth bug blocking deploy1.00
    15. AIE1.00
    16. ★1.00
    17. ★1.00
    18. ★1.00
    19. ★1.00
    20. Scanning Slack channels...0.99
    21. - #eng-requests: auth bug - action needed0.97
    22. $ morning brief1.00
    23. AIE-1431.00
    24. API latency investigation0.98
    25. - #incidents: API latency - action needed0.98
    26. + #product-asks: dashboard timeline - deduped w/ AIE-1390.99
    27. AIE-1441.00
    28. - DM Sarah: deploy question -surfaced0.97
    29. Dashboard shipping timeline1.00
    30. - #general: lunch - filtered0.96
    31. Creating Linear tickets...0.99
    32. Clean. Organized. Deduped.0.99
    33. AIE-142 created from #eng-requests0.98
    34. AIE-143 created from #incidents0.99
    35. noise filtored out0.96
    36. AIE-144 deduped → existing AIE-1390.96
    37. Done. 16 messages - 2 new tickets, 1 deduped.0.96
    38. You never opened Slack. You never opened Linear.1.00
    39. Two MCP connectors. 90% reduction in perceived context-switching cost.1.00
    40. Slack + Linear flow into Claude Code. One interface. Zero alt-tabbing.1.00
    41. Replay1.00
    42. Engineering the future of Al1.00

  • 7:17 #19 done27 line(s)

    shot 19·sharpness 1899.3

    1. Untathered Productivity - S0.95
    2. Ask Geminl0.96
    3. localhost:3031/131.00
    4. a0.60
    5. Work0.99
    6. VOICE-FIRST FLOWSThe Pipeline0.99
    7. Voice Capture0.99
    8. Speech Recognition1.00
    9. Al Enhancement0.97
    10. Polished Output0.99
    11. AIE1.00
    12. • Text inserted at cursor!0.96
    13. ★1.00
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. Refactor the auth module: the current implementation uses deprecated methods. Add improved error handling.1.00
    18. Clean, professional text- no editing needed0.98
    19. Replay1.00
    20. 901.00
    21. 1791.00
    22. 2x1.00
    23. WPM TYPING1.00
    24. WPMVOICE1.00
    25. FASTER1.00
    26. 00.81
    27. Engineering the future of Al0.99

  • 7:57 #20 done34 line(s)

    shot 20·sharpness 1921.4

    1. Untathered Productivity - Sl0.95
    2. Ask Gemini0.97
    3. ①localhost:3031/140.97
    4. Work0.99
    5. VOICE-FIRST FLOWSTheMultiplier1.00
    6. What happens when you can speak instructions to 3 agents at once?0.99
    7. VOICE-DISPATCHED AGENTS1.00
    8. AIE1.00
    9. ★1.00
    10. ★1.00
    11. Agent A0.96
    12. 100%1.00
    13. Auth module refactor0.99
    14. ✓ Done0.90
    15. Agent B0.98
    16. 100%1.00
    17. Dashboard API endpoints0.98
    18. ✓ Done0.89
    19. ★1.00
    20. ★1.00
    21. ★1.00
    22. ★1.00
    23. Agent C1.00
    24. 100%1.00
    25. Test suite migration1.00
    26. √ Done0.86
    27. KEYBOARD-ONLY DEVELOPER1.00
    28. Typing1.00
    29. 26%1.00
    30. Still typing first task...0.98
    31. Voice dispatched 3 tasks in the time keyboard completed 26% of one1.00
    32. Replay race1.00
    33. Braintrust1.00
    34. WorkOs OpenAI0.93

  • 8:19 #21 skipped

    shot 21·duplicate of #7

  • 8:44 #22 done22 line(s)

    shot 22·sharpness 2211.0

    1. Untathered Productivity - S0.90
    2. Ask Gemini0.97
    3. ①localhost:3031/150.97
    4. Work1.00
    5. REMOTE CONTROL The Idea0.98
    6. You know the shower principle?1.00
    7. *★★0.62
    8. ★1.00
    9. You're stuck on a hard problem at your desk. Focused mode - deep concentration,0.99
    10. AIE1.00
    11. ★1.00
    12. narrowing in. Then you walk away - go for a run, stand in line at the grocery store - and0.98
    13. ★1.00
    14. ★1.00
    15. suddenly the solution hits you. Diffuse mode. Your brain makes connections when you step0.99
    16. away.1.00
    17. We've always known this. But before, stepping away meant stopping work.1.00
    18. Now it doesn't.1.00
    19. Show of hands - who's controlled an agent from their phone while away from the desk? [hands] OK. For the rest of you,0.99
    20. watch this.1.00
    21. AlEngineer0.96
    22. EUROPE1.00

  • 9:25 #23 skipped

    shot 23·duplicate of #15

Transcript

264 cues· 5,125 words· 26,790 chars

  1. 0:14 Hey, everyone.
  2. 0:15 I'm Zach.
  3. 0:15 I work at Work OS.
  4. 0:16 Thanks for coming.
  5. 0:17 Work OS is provides drop in API's that allow you to take your software and go upmarket and sell larger deals to enterprises.
  6. 0:26 But what I'm going to talk about now is sort of the way that I'm finding to try and maintain balance with all the insane new tools that we're getting every day.
  7. 0:35 So show of hands if anyone is AI coding with agents lately and feels a little bit like this and despite getting more done than ever before, you're completely fried at the end of the day and adrenaline dumping constantly.
  8. 0:48 So this has been my experience and I've noticed that some of the worst of it is the context switching was always super expensive for me and now it's worse than ever before.
  9. 0:56 So I think a lot of us are feeling this way
  10. 0:59 Like the tools are like insanely powerful and our skills are more in demand than ever before and yet we're like exhausted by 11 a.m.
  11. 1:07 So I'll make this concrete with a recent story.
  12. 1:10 I was working, I'm on the applied AI team at WorkOS.
  13. 1:12 I was building a Slack bot that kind of democratized uniform blogging for everybody so that anybody that, even if they've never written a blog post before, can come into a Slack channel, make a simple request, and get a uniform blog post that does everything the correct way, right?
  14. 1:27 And there was a bug that one of my colleagues reported that said, hey, we need to use sentence case, and the sentence case pass right now is mangling some of our acronyms like SCIM and SSO.
  15. 1:37 So normally, I'd be living in that kind of window hell that I just showed you.
  16. 1:41 And instead, this time, I made a very minor change that ended up being super impactful.
  17. 1:46 So I gave Cloud Code, which is my current preferred aperture for working, the ability to read and write to Slack as well.
  18. 1:53 And it already had my linear ticket access.
  19. 1:55 And so I told it, you need to fix this.
  20. 1:58 And then you also need to verify your own work.
  21. 2:00 And don't stop until you've done that.
  22. 2:03 And so roughly, this is what it looked like.
  23. 2:04 If my terminals on the left, I ran Claude, and I said, fix the sentence case enforcer.
  24. 2:08 It's mangling acronyms.
  25. 2:10 And so it went and did that, and because it had an MCP connection to Slack, it fired it into this blog post channel, and then the blog bot that it's working on, it's sitting in that code base in my working directory, picked it up and ran all the way through and got to the step that was relevant, and then Claude verified the outcome and then said, okay, now I have definitively fixed this bug.
  26. 2:33 So when I came back to it, I came back to a completed loop that had been fixed, and I didn't have to tell it, hey, this part's still broken.
  27. 2:42 So that felt incredible, and I think that that's an important part of what we're going to all build into our toolkit, and we already are.
  28. 2:49 But it also scared me because I realized that there's no ceiling to this, right?
  29. 2:56 So the tools are nuclear now and our nervous system is still relatively ancient.
  30. 2:59 And so the thing that I'm thinking about recently is how do I find my own kind of developer balance in this world, right?
  31. 3:06 So why not just stack that same process?
  32. 3:09 Why not use that harness and just fix 150 bugs every day at work?
  33. 3:13 Well, the agents can kind of do that, especially if you give them enough context, if you give them the right verification criteria, if you give them the tools to verify their own work.
  34. 3:22 But at the end of the day, I can't actually sit on top of all of that and make sure that the quality is there.
  35. 3:28 and also show up the next day for like another eight hour work session and not be completely destroyed.
  36. 3:34 So what I think I'm finding, and I think a lot of other folks I've been talking to recently are finding, is that the agents are not the bottleneck now, and I think that's going to increasingly be the case, but we are.
  37. 3:44 So agents can scale infinitely, especially now that they're made available as of last night via Cloud API.
  38. 3:50 You can give them verification criteria and the tools that they need in order to match that criteria.
  39. 3:55 But our attention is still in meat space, if you will, and it still degrades under load.
  40. 4:00 It's still the hard constraint, essentially.
  41. 4:03 And this is something that's not just me and not just you, everyone that just raised their hand at the beginning of this talk feeling it.
  42. 4:09 As Simon Wilson said recently, like last week, that he fires up four parallel agents and he's wiped out by 11 a.m. And so he suggests that all of us finding our own individual balance and sort of our personal limits is now something that we all need to do on our own.
  43. 4:24 I'm definitely seeing this also on the Applied AI team.
  44. 4:26 So here's a couple of tips and tricks or things that I've used recently and found success with, and so I'll share them.
  45. 4:33 Some of them I'm sure you've already seen.
  46. 4:35 So we essentially need to bring the human developer into balance with this new way of working, because now that we have these hypercharged tools, it's faster than ever to burn yourself out, especially if you just scale linearly in terms of what you're taking on at work and what you're outputting.
  47. 4:51 And so the way that I'm sort of breaking it down in my mind is that agents are going to do better in terms of infinitely scaling, looping infinitely until they have reached the criteria you've given them.
  48. 5:02 But we still have judgment, taste, knowing that something is actually solved, knowing that the criterion is actually met in terms of human needs and business needs.
  49. 5:12 And so this is kind of an early breakdown of the stack that I'm seeing.
  50. 5:15 So the first I'm calling signal layers, for lack of a better word.

Chapters

  1. 0:00 Introduction and the problem of AI developer burnout
  2. 1:08 A concrete example of using AI agents for bug fixing
  3. 3:36 The bottleneck: human attention vs. infinite agent scalability
  4. 5:13 The proposed stack for sustainable AI development
  5. 6:03 The Signal Layer: Managing notifications and context switching
  6. 7:04 Voice-first flows for coding efficiency
  7. 8:15 The Shower Principle and Remote Control for agents
  8. 11:13 Safety, verification, and testing gates
  9. 12:05 The new, balanced developer workflow
  10. 13:22 System self-improvement using conversation history
  11. 15:31 Holistic well-being (Oura ring integration)
  12. 17:35 Q&A: Addressing skill development early in a career
  13. 20:07 Q&A: Managing JSONL history and long-term conversation logs
  14. 21:30 Q&A: Night shift/background agent execution
  15. 22:17 Q&A: Voice interaction and audio feedback
  16. 23:39 Q&A: Handling complex, multi-stack features

Open at this second