WTF Is the Context Layer? The Missing Infrastructure for Production Agents — Prukalpa Sankar

What was stored

cues

255

whisperx 255

chunks

36

from 255 cues

keyframes

35

kept of 51 captured

frames with text

35

988 lines read

chapters

0

from the source metadata

keyframe bytes

7.4 MB

word timings on 255 cues

Provenance

Frames, and what the machine read

• 

  0:02 #0 done2 line(s)

  shot 0·sharpness 448.8

  1. AlEngineer0.95
  2. World's Fair0.97
• 

  0:03 #1 done2 line(s)

  shot 1·sharpness 663.0

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

  0:10 #2 done24 line(s)

  shot 2·sharpness 2732.4

  1. LAB & PLATINUM SPONSORS0.99
  2. Amazon AGI Lab0.98
  3. ANTHROP\C1.00
  4. Google DeepMind1.00
  5. MINIMAX0.96
  6. OpenAI0.92
  7. Akamai1.00
  8. arize1.00
  9. aws1.00
  10. Braintrust bright data0.98
  11. B1.00
  12. Browserbase1.00
  13. docker1.00
  14. :neo4j0.91
  15. ORACLE1.00
  16. PayPal1.00
  17. qodo1.00
  18. reducto1.00
  19. Sonar1.00
  20. Makers of0.99
  21. togetherai1.00
  22. Unblocked1.00
  23. WorkOS1.00
  24. SonarQube1.00
• 

  0:31 #3 done2 line(s)

  shot 3·sharpness 351.5

  1. atlan0.99
  2. World's Fair0.97
• 

  0:50 #4 done9 line(s)

  shot 4·sharpness 1314.8

  1. AlEngineer0.99
  2. World's Fair0.97
  3. atlan1.00
  4. PRESENTED BY0.99
  5. WTF is the Context Layer1.00
  6. Microsoft1.00
  7. The Missing Infrastructure1.00
  8. for Production Agents1.00
  9. Engineering the future of Al0.99
• 

  0:53 #5 done24 line(s)

  shot 5·sharpness 2160.9

  1. AlEngineer0.98
  2. Al doesn't know your business, we fix that1.00
  3. atlan1.00
  4. World'sFair1.00
  5. mastercard1.00
  6. GitLab1.00
  7. zoom1.00
  8. Hub$pot0.95
  9. PRESENTED BY1.00
  10. Microsoft1.00
  11. elastic1.00
  12. Nasdaq1.00
  13. RIOT0.96
  14. Medtronic1.00
  15. GAMES1.00
  16. GRAINGER1.00
  17. FOX1.00
  18. affirm1.00
  19. Vugia0.71
  20. workday.0.98
  21. Discord1.00
  22. general motors1.00
  23. Dropbox1.00
  24. Engineering the future of Al0.98
• 

  0:55 #6 done12 line(s)

  shot 6·sharpness 1365.7

  1. AlEngineer0.99
  2. World's Fair0.98
  3. Content1.00
  4. Context1.00
  5. is King1.00
  6. PRESENTED BY1.00
  7. Microsoft1.00
  8. Bill Gates 19960.97
  9. is King.0.97
  10. Jun 20251.00
  11. atlan1.00
  12. Engineering the future of Al1.00
• 

  1:14 #7 done24 line(s)

  shot 7·sharpness 2182.5

  1. AlEngineer0.99
  2. Al doesn't know your business, we fix that1.00
  3. atlan1.00
  4. World'sFair1.00
  5. mastercard1.00
  6. GitLab1.00
  7. zoom1.00
  8. HubSpot0.92
  9. PRESENTED BY0.99
  10. Microsoft1.00
  11. elastic1.00
  12. Nasdaq0.99
  13. RIOT0.98
  14. Medtronic1.00
  15. GAMES1.00
  16. GRAINGER1.00
  17. FOX1.00
  18. affirm1.00
  19. workday.0.97
  20. Discord1.00
  21. general motors1.00
  22. Dropbox1.00
  23. TRACK 8· JULY 1, 20260.95
  24. Context Engineering0.99
• 

  1:24 #8 done13 line(s)

  shot 8·sharpness 1389.9

  1. AlEngineer1.00
  2. World's Fair0.97
  3. Content1.00
  4. Context1.00
  5. is King1.00
  6. PRESENTED BY1.00
  7. Microsoft1.00
  8. Bill Gates 19961.00
  9. is King.0.94
  10. Jun 20250.95
  11. atlan1.00
  12. TRACK 8 · JULY 1, 20260.92
  13. Context Engineering0.99
• 

  1:37 #9 done40 line(s)

  shot 9·sharpness 3485.8

  1. 2026 is the year of Context0.99
  2. AlEngineer1.00
  3. World's Fair0.98
  4. Launches Context0.99
  5. Y Combinator0.97
  6. Mesh to Connect Enterprise0.99
  7. @ycombinator1.00
  8. Data to Al Agents0.98
  9. The Al Operating System for Companies1.00
  10. Raises USD 3 Million to1.00
  11. @sdianahu0.95
  12. Build Context Layer Enterprise Al0.99
  13. The best Al-native companies have made their entire company1.00
  14. PRESENTED BY1.00
  15. queryable: every meeting, ticket, and customer interaction legible to an1.00
  16. intelligence layer that learns from it.1.00
  17. Jaya Gupta@JayaGup10 · Dec 23, 20250.94
  18. Microsoft1.00
  19. Building this today requires brutal integration work, and there's no1.00
  20. product that connects all this context into a single layer that can reason1.00
  21. across it.1.00
  22. Context Graphs: Capturing The "Why"1.00
  23. In The Age of Al0.98
  24. Founder and CTO at HubSpot. Helping millions grow better.0.99
  25. Dharmesh Shah1.00
  26. X Article0.92
  27. January 10, 20260.99
  28. Al's trillion-dollar opportunity: Context graphs1.00
  29. The last generation of enterprise software created a trillion-dollar0.99
  30. ecosystem by becoming systems of record. Salesforce for customers.1.00
  31. to Deploy 1,000 Context Engineers,1.00
  32. Workday for employees. SAP for operations. Own the canonical...1.00
  33. Powered by1.00
  34. to Industrialize Agentic Al1.00
  35. 3961.00
  36. t2 1.3K0.90
  37. O7K0.77
  38. l 4.8M0.86
  39. TRACK 8· JULY 1, 20260.95
  40. Context Engineering1.00
• 

  2:30 #10 done28 line(s)

  shot 10·sharpness 2100.5

  1. AlEngineer0.98
  2. World's Fair0.99
  3. Models are getting exponentially smarter1.00
  4. But not exponentially more useful1.00
  5. 20231.00
  6. 20251.00
  7. Financial benefit1.00
  8. from Al0.96
  9. Graduate science1.00
  10. 35%1.00
  11. 87% (experts: 65%)1.00
  12. ~56% of CEOs1.00
  13. Competition math1.00
  14. 9%1.00
  15. 97%1.00
  16. Novel reasoning0.98
  17. 5%1.00
  18. 87.5% (humans: 85%)0.99
  19. 1/51.00
  20. Report significant0.98
  21. Software engineering1.00
  22. 15%1.00
  23. 60%+1.00
  24. value1.00
  25. ~Gartner0.97
  26. atlan1.00
  27. TRACK 8• JULY 1, 20260.95
  28. Context Engineering1.00
• 

  2:43 #11 done21 line(s)

  shot 11·sharpness 1802.4

  1. AlEngineer0.99
  2. Cognitive Intelligence (IQ) doesn't determine0.99
  3. World's Fair0.98
  4. real world effectiveness.0.98
  5. High1.00
  6. eftecttcnes0.57
  7. Rel wrld0.57
  8. ~10%0.99
  9. of1.00
  10. performance1.00
  11. variance1.00
  12. explained by1.00
  13. IQ1.00
  14. Low1.00
  15. Low1.00
  16. High1.00
  17. Cognitive intelligence (IQ)1.00
  18. Based on: Sternberg & Wagner (1986), Practical Intelligence; Sackett, Zhang, Berry & Lievens (2022), Journal of Applied Psychology.1.00
  19. atlan1.00
  20. TRACK 8·JULY 1,20260.98
  21. ContextEngineering1.00
• 

  3:25 #12 done12 line(s)

  shot 12·sharpness 1782.2

  1. AlEngineer0.99
  2. World'sFair1.00
  3. Context: knowledge, skills, and1.00
  4. tools built through doing the work.1.00
  5. Performance: The outcomes you0.99
  6. actually deliver in the real world.0.99
  7. P=f(I,C)0.97
  8. Intelligence: Cognitive horsepower. How fast0.99
  9. you learn, reason, process. Largely fixed.1.00
  10. atlan1.00
  11. TRACK 8· JULY 1, 20260.94
  12. Context Engineering1.00
• 

  3:53 #13 done13 line(s)

  shot 13·sharpness 1805.7

  1. AlEngineer0.98
  2. The last decade0.99
  3. World'sFair1.00
  4. Intelligence1.00
  5. ~1000x Compounding. Commoditizing.0.99
  6. Context barely moved.1.00
  7. Intelligence (I) — frontier capability.0.98
  8. Context (C) — Situated knowledge of your business. Locked in0.99
  9. Compounding. Commoditizing.1.00
  10. dashboards, Slack threads, and the analyst who's about to leave.1.00
  11. atlan1.00
  12. TRACK 8· JULY 1, 20260.93
  13. Context Engineering0.98
• 

  4:08 #14 done8 line(s)

  shot 14·sharpness 1183.4

  1. AlEngineer0.98
  2. World'sFair1.00
  3. How do we help Al build0.98
  4. context about our business?1.00
  5. Hint: The same way we train and onboard humans.0.99
  6. atlan1.00
  7. TRACK 8· JULY 1, 20260.94
  8. Context Engineering0.99
• 

  4:34 #15 done18 line(s)

  shot 15·sharpness 3518.9

  1. AlEngineer0.98
  2. Meet Maya0.98
  3. World'sFair1.00
  4. She's a data analyst at McContext0.98
  5. Burgers — the person every1.00
  6. operator pings when something1.00
  7. Marcus1.00
  8. looks off. This morning it's Marcus,1.00
  9. Franchise owner1.00
  10. a franchisee with 14 stores0.98
  11. Why is my drive-through time up this week?1.00
  12. Sounds simple. It isn't.1.00
  13. Before Maya can even start, she has to answer a stack0.99
  14. of questions the message doesn't — and none of them1.00
  15. are in the data. They're in her head.1.00
  16. atlan1.00
  17. TRACK 8· JULY 1, 20260.94
  18. Context Engineering1.00
• 

  5:15 #16 done54 line(s)

  shot 16·sharpness 2356.8

  1. AlEngineer0.99
  2. Why is drive-through time up this week?I0.98
  3. KNOWLEDGE1.00
  4. World's Fair0.99
  5. The map of the business0.97
  6. To answer this seemingly simple question, the agent needs multiple types of context1.00
  7. What things mean. Entities, definitions,1.00
  8. metrics, relationships — the ontology of0.99
  9. how the company talks to itself.0.97
  10. KNOWLEDGE1.00
  11. EXPERTISE1.00
  12. the business1.00
  13. The map of1.00
  14. How work1.00
  15. gets done1.00
  16. ☆1.00
  17. EXPERTISE1.00
  18. "Drive-through time"1.00
  19. means avg_dt_secs, not1.00
  20. CONTEXT1.00
  21. Diagnostic playbook —0.97
  22. validate the premise,0.98
  23. How work gets done1.00
  24. The procedural know-how. Playbooks,0.99
  25. the POS number Finance0.99
  26. check seasonality, then0.98
  27. workflows, task decomposition — the0.99
  28. uses.1.00
  29. investigate root causes.1.00
  30. practical skill of running the business.1.00
  31. "This week" = Mon-Sun0.98
  32. Root-cause correlation-0.99
  33. in store local time, closed0.99
  34. NORMS1.00
  35. cross-reference against1.00
  36. rolling 7-day window.0.99
  37. Sunday 11:59 PM. Not a0.98
  38. acceptable action1.00
  39. The rules of1.00
  40. weather, public events.0.98
  41. product launches,1.00
  42. NORMS1.00
  43. The rules of acceptable action0.99
  44. What is allowed. Policies, permissions,1.00
  45. Persona scoping — Store manager only sees his0.98
  46. scoping, approval paths — the guardrails0.99
  47. store data. VP Ops sees the full chain.1.00
  48. that keep agents safe.0.99
  49. Metric authority — Store manager gets0.97
  50. avg_dt_secs; Finance gets POS metrics.1.00
  51. atlan1.00
  52. Wordu Far0.83
  53. TRACK 8· JULY 1, 20260.96
  54. Context Engineering1.00
• 

  5:38 #17 done20 line(s)

  shot 17·sharpness 3682.8

  1. AlEngineer0.98
  2. Maya1.00
  3. World'sFair1.00
  4. Maya nails it0.99
  5. Looked into it0.98
  6. Scoped to his stores. Root1.00
  7. cause found. Fix attached.1.00
  8. Your drive-through time is up ~18s, but only at your0.99
  9. 6 single-lane stores — dual-lane locations are flat.0.99
  10. Cause: the Double Stack launched Monday. Extra1.00
  11. prep step, no second staging lane to absorb it.1.00
  12. Fix: pre-stage Double Stack patties at single-lane0.99
  13. stores during the 11–1 rush. Should claw most of it0.99
  14. back this week.1.00
  15. Marcus1.00
  16. Franchise owner1.00
  17. This is exactly what I needed.0.99
  18. atlan1.00
  19. TRACK 8• JULY 1, 20260.94
  20. Context Engineering0.99
• 

  5:50 #18 skipped

  shot 18·duplicate of #16

• 

  6:19 #19 skipped

  shot 19·duplicate of #17

• 

  6:36 #20 done51 line(s)

  shot 20·sharpness 5434.4

  1. Drive-through time'=1.00
  2. Week 10.97
  3. Week 21.00
  4. AlEngineer0.99
  5. avg_dt_secs. Not the0.96
  6. World's Fair0.98
  7. Finance number.1.00
  8. How Maya learned to1.00
  9. be a great analyst1.00
  10. Clicked through every dashboard.0.99
  11. Sample queries ran clean.0.97
  12. PRESENTED BY1.00
  13. The vocabulary1.00
  14. The tools1.00
  15. What's true: the metrics, the0.99
  16. System fluency: where the data lives0.99
  17. Microsoft1.00
  18. Rule out1.00
  19. Prove it's real.1.00
  20. I sent the what. Marcus0.99
  21. needed the why. Never again1.00
  22. definitions, the map of the business.1.00
  23. and how to pull it.0.99
  24. go hunting.0.98
  25. seasonality. Then0.98
  26. Week 40.93
  27. Month 10.99
  28. Month 20.94
  29. Month 31.00
  30. the what.1.00
  31. You showed him0.99
  32. Doas had three lines. So I0.98
  33. chain-wide. He's a1.00
  34. built my own map.0.98
  35. franchisee - 14 stores0.95
  36. Shadowing1.00
  37. First wrong answer0.98
  38. Manager Feedback1.00
  39. Edge Case1.00
  40. The diagnostic playbook: validate →0.98
  41. The answer is the why, not the1.00
  42. Persona scoping & metric authority:0.99
  43. Skills- Knowledge (The cases and0.99
  44. check seasonality → root-cause.0.98
  45. what — the Double Stack jams0.98
  46. single-lane stores.1.00
  47. POS.1.00
  48. Ops gets avg_dt_secs; Finance gets1.00
  49. people that wiki doesn't cover)0.99
  50. TRACK 8· JULY 1, 20260.96
  51. Context Engineering1.00
• 

  7:20 #21 skipped

  shot 21·duplicate of #14

• 

  7:46 #22 skipped

  shot 22·duplicate of #13

• 

  8:37 #23 done51 line(s)

  shot 23·sharpness 2015.6

  1. AlEngineer1.00
  2. One human, ten specialized agents1.00
  3. World's Fair0.98
  4. A HUMAN0.91
  5. Customer Success Manager0.98
  6. Manages all the leads1.00
  7. AGENT1.00
  8. AGENT1.00
  9. © AGENT0.92
  10. Donna, Call Intelligence Lead1.00
  11. callare performed well.0.95
  12. Manages a team to ensure all te tasks after a0.95
  13. Hermione, Health1.00
  14. Intelligence Lead1.00
  15. opportunities. and recommends next steps.0.98
  16. Reviews curent state, identifies risk and0.97
  17. stage is performed at the highest quality.0.98
  18. Sheldon, Customer1.00
  19. Onboarding Lead1.00
  20. AGENT0.97
  21. AGENT0.98
  22. AGENT0.98
  23. AGENT0.99
  24. AGENT0.99
  25. Sherlock, Strategic Success1.00
  26. Analyst1.00
  27. Success plan management, value0.97
  28. ariculation. milestone tracking.0.95
  29. Leslie, Relationship Analyst0.99
  30. Stakeholder engagement, champion0.98
  31. analysis. communication pattems.0.97
  32. Json, Product Analyst0.98
  33. Deep dive on a customer's product0.99
  34. usage data to find insights and0.98
  35. patterns.0.97
  36. Moneypenny, Financial Risk0.99
  37. Analyst1.00
  38. Deep dive on a customer's commercial0.99
  39. financial signals.0.99
  40. eath from contracts and extemal0.86
  41. Nora, Support Analyst0.98
  42. Support ticket trends and current status0.99
  43. of open tickets.0.93
  44. AGENT0.98
  45. Linkin, Relationship Analyst0.98
  46. Gathers Linkedin profile details and0.99
  47. recent posts based on a person's name0.99
  48. atlan1.00
  49. Workde Ffal0.63
  50. TRACK 8 · JULY 1, 20260.94
  51. Context Engineering1.00

Transcript

255 cues· 3,189 words· 17,455 chars