vidtheque.

read-only demo

Videos eW_vxrjvERk

Connecting the Dots with Context Graphs — Stephen Chin, Neo4j

index_state ready data_status ok

AI Engineer· published 2026-05-16· 0:17:38· en-US· indexed 2026-08-10 19:53

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 0:40, 1 of 1 keyframes kept
  5. Shot 4, 0:40 to 1:14, 1 of 1 keyframes kept
  6. Shot 5, 1:14 to 1:26, 1 of 1 keyframes kept
  7. Shot 6, 1:26 to 1:53, 1 of 1 keyframes kept
  8. Shot 7, 1:53 to 2:11, 1 of 1 keyframes kept
  9. Shot 8, 2:11 to 2:34, 1 of 1 keyframes kept
  10. Shot 9, 2:34 to 3:03, 1 of 1 keyframes kept
  11. Shot 10, 3:03 to 3:37, 1 of 1 keyframes kept
  12. Shot 11, 3:37 to 4:11, 1 of 1 keyframes kept
  13. Shot 12, 4:11 to 4:24, 1 of 1 keyframes kept
  14. Shot 13, 4:24 to 4:50, 1 of 1 keyframes kept
  15. Shot 14, 4:50 to 5:16, 1 of 1 keyframes kept
  16. Shot 15, 5:16 to 5:43, 0 of 1 keyframes kept
  17. Shot 16, 5:43 to 6:09, 1 of 1 keyframes kept
  18. Shot 17, 6:09 to 6:23, 1 of 1 keyframes kept
  19. Shot 18, 6:23 to 6:39, 1 of 1 keyframes kept
  20. Shot 19, 6:39 to 7:08, 1 of 1 keyframes kept
  21. Shot 20, 7:08 to 7:36, 1 of 1 keyframes kept
  22. Shot 21, 7:36 to 8:05, 0 of 1 keyframes kept
  23. Shot 22, 8:05 to 8:36, 1 of 1 keyframes kept
  24. Shot 23, 8:36 to 9:04, 1 of 1 keyframes kept
  25. Shot 24, 9:04 to 9:32, 0 of 1 keyframes kept
  26. Shot 25, 9:32 to 10:01, 1 of 1 keyframes kept
  27. Shot 26, 10:01 to 10:29, 1 of 1 keyframes kept
  28. Shot 27, 10:29 to 10:33, 1 of 1 keyframes kept
  29. Shot 28, 10:33 to 11:02, 1 of 1 keyframes kept
  30. Shot 29, 11:02 to 11:30, 0 of 1 keyframes kept
  31. Shot 30, 11:30 to 11:37, 1 of 1 keyframes kept
  32. Shot 31, 11:37 to 11:51, 0 of 1 keyframes kept
  33. Shot 32, 11:51 to 12:29, 1 of 1 keyframes kept
  34. Shot 33, 12:29 to 13:11, 1 of 1 keyframes kept
  35. Shot 34, 13:11 to 13:26, 1 of 1 keyframes kept
  36. Shot 35, 13:26 to 13:52, 1 of 1 keyframes kept
  37. Shot 36, 13:52 to 14:17, 1 of 1 keyframes kept
  38. Shot 37, 14:17 to 14:19, 1 of 1 keyframes kept
  39. Shot 38, 14:19 to 15:06, 1 of 1 keyframes kept
  40. Shot 39, 15:06 to 15:28, 1 of 1 keyframes kept
  41. Shot 40, 15:28 to 15:42, 1 of 1 keyframes kept
  42. Shot 41, 15:42 to 15:54, 1 of 1 keyframes kept
  43. Shot 42, 15:54 to 16:13, 1 of 1 keyframes kept
  44. Shot 43, 16:13 to 16:57, 1 of 1 keyframes kept
  45. Shot 44, 16:57 to 17:23, 1 of 1 keyframes kept
  46. Shot 45, 17:23 to 17:38, 1 of 1 keyframes kept

46 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
152
whisperx 152
chunks
30
from 152 cues
keyframes
41
kept of 46 captured
frames with text
41
1,037 lines read
chapters
0
from the source metadata
keyframe bytes
5.7 MB
word timings on 152 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 17:27 1m 26s
stt done — 2026-08-10 17:29 19s
chunk done — 2026-08-10 17:29 0s
text_embed done — 2026-08-10 19:53 1s
keyframe done — 2026-08-10 17:29 1m 27s
ocr done — 2026-08-10 17:30 19s
frame_embed done — 2026-08-10 19:53 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

  • 0:30 #3 done10 line(s)

    shot 3·sharpness 429.4

    1. AlEngineer1.00
    2. Connecting the d1.00
    3. EUROPE1.00
    4. with context grap0.98
    5. Stephen Chin (@steveonjava)1.00
    6. EU/ACO0.88
    7. VP of Developer Relations @ Neo4j0.99
    8. AlEngineer0.99
    9. a(e'el0.51
    10. EUROPE1.00

  • 0:57 #4 done23 line(s)

    shot 4·sharpness 1017.2

    1. yEQ0.55
    2. R0.56
    3. m0.73
    4. *0.71
    5. 8m2110.59
    6. 80.51
    7. nrVon0.58
    8. **0.87
    9. AIE1.00
    10. ★1.00
    11. en0.96
    12. 60.90
    13. *OTR0.63
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. =m O0.65
    18. Dre0.63
    19. mes0.59
    20. Engineer1.00
    21. AlEngineer0.97
    22. D260.60
    23. EUROPE1.00

  • 1:22 #5 done8 line(s)

    shot 5·sharpness 1218.0

    1. AIE1.00
    2. ★1.00
    3. ★0.98
    4. ALLI AM OFFERING IS THE TRUTH0.99
    5. Engineer1.00
    6. AlEngineer0.96
    7. 2860.66
    8. EUROPE1.00

  • 1:49 #6 done28 line(s)

    shot 6·sharpness 1743.2

    1. Today1.00
    2. Scattered, lost reasoning1.00
    3. Why?1.00
    4. Considered?1.00
    5. Alternatives1.00
    6. CRM1.00
    7. AIE1.00
    8. ★1.00
    9. Precedent?1.00
    10. ★1.00
    11. ★1.00
    12. ★1.00
    13. #0.99
    14. Slack1.00
    15. Slack1.00
    16. Zendesk Support0.97
    17. Ticket1.00
    18. Jira Project0.99
    19. Board1.00
    20. pd1.00
    21. PagerDuty1.00
    22. Zoom1.00
    23. Context Lose in People's Heads & Silos0.99
    24. Tribal Knowledge, Cross-System Synthesis,1.00
    25. Approval Chains Outside Systems.1.00
    26. Engineering the future of Al0.97
    27. Engineer1.00
    28. J260.53

  • 2:00 #7 done32 line(s)

    shot 7·sharpness 2416.7

    1. Context Graph1.00
    2. Connected, captured decision traces1.00
    3. *★★0.57
    4. Decision1.00
    5. Agent1.00
    6. Policy1.00
    7. AIE1.00
    8. ★1.00
    9. Discount Approval1.00
    10. Standard Discount Rule1.00
    11. ★1.00
    12. ★1.00
    13. ★1.00
    14. ★1.00
    15. System of1.00
    16. Healthcare Sector1.00
    17. Exception1.00
    18. Reasoning1.00
    19. Similar Deal Q30.99
    20. Precedent1.00
    21. Evidence1.00
    22. Evidence1.00
    23. Churn Risk Data0.97
    24. Similar Deal Q30.99
    25. Outcome1.00
    26. Deal Closed, Positive1.00
    27. Inputs, Policies, Exceptions, Precedents,1.00
    28. Approvers, Approvers, Outcomes.0.99
    29. Decision Trace as First-1.00
    30. Engineering the future of Al0.98
    31. Engineer1.00
    32. J260.66

  • 2:25 #8 done14 line(s)

    shot 8·sharpness 269.7

    1. Conne0.97
    2. AlEngineer0.99
    3. EUROPE1.00
    4. Discount Appr0.97
    5. Decision1.00
    6. △0.61
    7. Exception1.00
    8. Healthcare Sector1.00
    9. EU/ACC0.97
    10. Inputs, Policies1.00
    11. Approvers. App0.97
    12. AlEngineer0.99
    13. EUROPE1.00
    14. 1260.90

  • 2:37 #9 done57 line(s)

    shot 9·sharpness 2054.0

    1. Hype cycle for Agentic Al1.00
    2. Multiagent Systems1.00
    3. Al Coding Agents1.00
    4. Enterprise Al Assistants-0.97
    5. Agentic Al Security1.00
    6. Agentic Analytics-0.98
    7. Model Context Protocol1.00
    8. gentic Al0.99
    9. AIE1.00
    10. ★1.00
    11. ★1.00
    12. ★1.00
    13. EXPTONS0.77
    14. Human-Agent Collaboration Workspace1.00
    15. Al Agent Development Platform0.99
    16. Domain-S0.98
    17. Context Graphs1.00
    18. No-Code Agent Builders1.00
    19. Machine Qustomers0.96
    20. ★1.00
    21. ★1.00
    22. ★1.00
    23. Embodied Al1.00
    24. Agent Management Platform0.99
    25. Al Agent Communication Protocols1.00
    26. Agent Marketplace1.00
    27. Al Agents in Software Enginering1.00
    28. Continual Learning1.00
    29. FinOps for Agentic Al-0.98
    30. Agentic Commerce1.00
    31. Agentic Browser1.00
    32. Agent Experience (AX)0.99
    33. Guardian Agent1.00
    34. Agent Development Life Cycle1.00
    35. Agent Orchestration1.00
    36. Plateau will be reached:1.00
    37. <2yrs.1.00
    38. 2.-5 yrs.0.96
    39. 5-10 yrs.1.00
    40. >10 yrs.1.00
    41. Obsolete before plateau0.97
    42. Computer Use for Agents1.00
    43. As of April 20261.00
    44. Innovation1.00
    45. Peak of Inflated0.99
    46. Trough of0.99
    47. Slope of1.00
    48. Plateau of1.00
    49. Trigger1.00
    50. Expectations1.00
    51. Disillusionment1.00
    52. Enlightenment1.00
    53. Productivity1.00
    54. TIME0.98
    55. Engineering the future of Al1.00
    56. Engineer1.00
    57. 2260.70

  • 3:33 #10 done40 line(s)

    shot 10·sharpness 4200.9

    1. Knowledge graph components1.00
    2. Name: “Dan"0.91
    3. Name: “Ann"0.92
    4. Born: May 29, 19700.98
    5. Born: Dec 5, 19750.99
    6. Nodes1.00
    7. Twitter: “@dan"0.91
    8. *★★0.67
    9. KNOWS1.00
    10. AIE1.00
    11. ★1.00
    12. represent entities in the graph1.00
    13. KNOWS1.00
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. ★1.00
    18. Relationships1.00
    19. Person1.00
    20. LIVES WITH1.00
    21. Person1.00
    22. DRIVES1.00
    23. represent associations or0.99
    24. interactions between nodes1.00
    25. Since:1.00
    26. OWNS1.00
    27. Jan 10, 20110.99
    28. Properties1.00
    29. Car1.00
    30. represent attributes of nodes or1.00
    31. Brand: “Volvo"0.94
    32. relationships including vectors, can be indexed1.00
    33. Model: “V70"0.95
    34. Description: “An executive car manufactured and..."0.99
    35. DescEmbedding: [0.1, -0.3, 0.4, ..., -0.7]0.98
    36. DescSource: "https://en.wikipedia.org/wiki/Volvo_V70"0.99
    37. Engineer1.00
    38. AlEngineer0.96
    39. J260.53
    40. EUROPE1.00

  • 3:51 #11 done39 line(s)

    shot 11·sharpness 4219.1

    1. Knowledge graph components1.00
    2. Name:“Dan"0.95
    3. Name: “Ann"0.93
    4. Born: May 29, 19700.99
    5. Born: Dec 5, 19750.97
    6. Nodes1.00
    7. Twitter:“@dan"0.97
    8. *★★0.67
    9. KNOWS1.00
    10. AIE1.00
    11. ★1.00
    12. represent entities in the graph1.00
    13. KNOWS1.00
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. ★1.00
    18. Relationships1.00
    19. Person1.00
    20. LIVES WITH0.99
    21. Person1.00
    22. DRIVES1.00
    23. represent associations or1.00
    24. interactions between nodes1.00
    25. Since:1.00
    26. OWNS1.00
    27. Jan 10, 20110.98
    28. Properties1.00
    29. Car1.00
    30. represent attributes of nodes or0.98
    31. Brand: “Volvo"0.95
    32. relationships including vectors, can be indexed1.00
    33. Model:“V70"0.99
    34. Description: “An executive car manufactured and..."0.99
    35. DescEmbedding: [0.1, -0.3, 0.4, ..., -0.7]0.99
    36. DescSource: "https://en.wikipedia.org/wiki/Volvo_V70"0.99
    37. Engineering the future of Al1.00
    38. Engineer1.00
    39. J260.52

  • 4:16 #12 done9 line(s)

    shot 12·sharpness 418.1

    1. from graphs eno0.99
    2. How1.00
    3. ousure a high-quality production0.97
    4. AlEngineer0.99
    5. EUROPE1.00
    6. 8880.85
    7. EU/ACC1.00
    8. AlEngineer0.99
    9. EUROPE1.00

  • 4:32 #13 done21 line(s)

    shot 13·sharpness 3521.6

    1. Context from graphs enables explainable Al0.98
    2. How do you ensure a high-quality production environment with Agents?0.99
    3. ***0.64
    4. AIE1.00
    5. Storing1.00
    6. ★1.00
    7. ★1.00
    8. ★1.00
    9. ★1.00
    10. learnings from user/agent interactions in the graph0.99
    11. database as context1.00
    12. 80.65
    13. Visualizing1.00
    14. conversations, flows, reasoning within context0.99
    15. Analysing1.00
    16. Context data for agent system performance,0.99
    17. semi-automatically identify & apply opportunities for0.99
    18. improvement1.00
    19. Engineering the future of Al1.00
    20. Engineer1.00
    21. 2260.57

  • 5:08 #14 done31 line(s)

    shot 14·sharpness 2792.9

    1. Enhance relevance with domain context1.00
    2. What was in the Care Plans associated with Andrea Jenkins's emphysema?1.00
    3. Baseline RAG0.98
    4. *★★0.56
    5. ★1.00
    6. AIE1.00
    7. ★1.00
    8. Llamalndex-RAG backed by GPT 4.01.00
    9. and Chroma vector database1.00
    10. ★1.00
    11. ★1.00
    12. ★1.00
    13. ★1.00
    14. The Care Plan associated with Andrea Jenkins' emphysema included0.99
    15. activities respiratory therapy, specifically deep breathing and0.99
    16. coughingexercises.0.97
    17. LLMDirect1.00
    18. OpenAl over Neo4j function call0.98
    19. Grounded but1.00
    20. The care plan associated with Andrea71.00
    21. incomplete1.00
    22. Jenkins714 emphysema likely included the0.99
    23. followingcomponents:1.00
    24. Not1.00
    25. connected0.99
    26. to real0.99
    27. data0.99
    28. Engineer1.00
    29. AlEngineer0.95
    30. J860.57
    31. EUROPE1.00

  • 5:24 #15 skipped

    shot 15·duplicate of #14

  • 5:48 #16 done37 line(s)

    shot 16·sharpness 2519.4

    1. Enhance relevance with domain context1.00
    2. What was in the Care Plans associated with Andrea Jenkins's emphysema?1.00
    3. Baseline RAG0.98
    4. *★★0.58
    5. ★1.00
    6. AIE1.00
    7. ★0.99
    8. Llamalndex-RAG backed by GPT 4.00.98
    9. ★1.00
    10. and Chroma vector database1.00
    11. ★1.00
    12. ★1.00
    13. ★1.00
    14. LLMDirect1.00
    15. GraphRAG1.00
    16. OpenAl over Neo4j function call0.98
    17. Grounded but1.00
    18. Neo4j, LangChain,1.00
    19. The care plan associated with Andrea71.00
    20. incomplete1.00
    21. and OpenAI0.99
    22. Jenkins714 emphysema likely included the0.99
    23. The Care Plan for Andrea7 Jenkins714's0.99
    24. Not1.00
    25. Complete,1.00
    26. emphysema included medication1.00
    27. connected1.00
    28. to real0.98
    29. data1.00
    30. accurate1.00
    31. answer1.00
    32. exercises.1.00
    33. management, smoking cessation1.00
    34. counseling,andpulmonary rehabilitation0.99
    35. Engineering the future of Al1.00
    36. Engineer1.00
    37. J260.58

  • 6:20 #17 done20 line(s)

    shot 17·sharpness 2576.3

    1. Short term memory0.99
    2. · Current Context0.95
    3. • Compression, Relevancy0.96
    4. *★*0.57
    5. • Integrate Tool Results0.99
    6. AIE1.00
    7. Long term memory1.00
    8. ★1.00
    9. ★1.00
    10. ★1.00
    11. · Episodic0.95
    12. · Semantic / Structural0.95
    13. • Procedural / Instructional0.97
    14. Reasoning memory1.00
    15. · Tool Call Traces0.98
    16. • Reasoning Traces0.96
    17. • Previous Decisions0.99
    18. Engineering the future of Al1.00
    19. Engineer1.00
    20. 2260.63

  • 6:37 #18 done21 line(s)

    shot 18·sharpness 482.8

    1. rm memory0.91
    2. Conv0.98
    3. istory and session state with au0.97
    4. AlEngineer1.00
    5. EUROPE1.00
    6. Conversation storage1.00
    7. Sessions and messages persisted as graph nodes with1.00
    8. metadata1.00
    9. Multi-stage entity extraction0.99
    10. Pipeline combining spaCy, GLiNER, and LLM extractors with0.99
    11. configurable merge strategies1.00
    12. EU/ACC0.99
    13. Entity resolution1.00
    14. Multi-strategy dedup: exact, fuzzy, and semantic matching with1.00
    15. type-aware resolution1.00
    16. session = await memory.add_session("user_123")0.99
    17. await memory.add_message(session.id, role="user",0.95
    18. content-"Review Jessica's account")0.98
    19. elo'w0.72
    20. AlEngineer1.00
    21. EUROPE1.00

  • 6:45 #19 done39 line(s)

    shot 19·sharpness 4290.4

    1. Short-term memory0.97
    2. Conversation history and session state with automatic entity extraction1.00
    3. Neo4j agent memory entity0.98
    4. ★1.00
    5. Conversation storage1.00
    6. extraction pipeline1.00
    7. AIE1.00
    8. ★1.00
    9. Sessions and messages persisted as graph nodes with1.00
    10. Stage 10.96
    11. Stage 20.98
    12. Stage 30.94
    13. ★1.00
    14. metadata1.00
    15. Raw Text1.00
    16. spaCy1.00
    17. GLiNER20.99
    18. LLM Fallback0.99
    19. ★1.00
    20. ★1.00
    21. ★1.00
    22. Multi-stage entity extraction1.00
    23. Pipeline combining spaCy, GLiNER, and LLM extractors with0.99
    24. Merge Strategy1.00
    25. configurable merge strategies0.99
    26. Union | confidence | cascade0.96
    27. Deduplication1.00
    28. Entity resolution1.00
    29. Fuzzy + semantic matching0.99
    30. Multi-strategy dedup: exact, fuzzy, and semantic matching with1.00
    31. Enrichment1.00
    32. type-aware resolution1.00
    33. Extracted Entities1.00
    34. session = await memory.add_session("user_123")0.99
    35. await memory.add_message(session.id, role="user",1.00
    36. content="Review Jessica's account")1.00
    37. Google DeepMind0.99
    38. Engineer1.00
    39. 2260.53

  • 7:33 #20 done34 line(s)

    shot 20·sharpness 3713.1

    1. Reasoning memory1.00
    2. Decision traces, tool usage audits, and provenance — the layer that makes Al explainable0.99
    3. *★*0.63
    4. ★1.00
    5. Tool call traces0.97
    6. Decision provenance0.97
    7. AIE1.00
    8. ★0.98
    9. 10.99
    10. 21.00
    11. Every tool invocation,1.00
    12. Why did the agent choose this path?1.00
    13. ★1.00
    14. ★1.00
    15. ★1.00
    16. ★1.00
    17. parameters, results —0.98
    18. complete audit trail1.00
    19. Causal chain fully recorded1.00
    20. Learning from1.00
    21. Compliance &1.00
    22. experience1.00
    23. 31.00
    24. debugging1.00
    25. Agent checks if it solved1.00
    26. When something goes1.00
    27. something similar before,0.98
    28. wrong, you can trace exactly1.00
    29. reuses successful patterns1.00
    30. what happened and why1.00
    31. AlEngineer0.97
    32. Engineer1.00
    33. J260.85
    34. EUROPE1.00

  • 7:53 #21 skipped

    shot 21·duplicate of #20

  • 8:12 #22 done19 line(s)

    shot 22·sharpness 425.7

    1. ng memory1.00
    2. Decis.1.00
    3. us, tool usage audits, and proven0.97
    4. AlEngineer1.00
    5. EUROPE1.00
    6. Tool call traces1.00
    7. 11.00
    8. Every tool invocation,0.99
    9. parameters, results -0.94
    10. complete audit trail1.00
    11. EU/ACC0.99
    12. Learning from1.00
    13. experience1.00
    14. 31.00
    15. Agent checks if it solved0.99
    16. something similar before,0.99
    17. reuses successful patterns1.00
    18. AlEngineer0.99
    19. EUROPE1.00

  • 8:45 #23 done30 line(s)

    shot 23·sharpness 4067.6

    1. Why graphs for agent memory?1.00
    2. Relationships Are First-Class1.00
    3. Multi-Hop Traversal1.00
    4. 品0.97
    5. **★0.60
    6. ★1.00
    7. Connections between entities, decisions, and1.00
    8. Follow chains: Customer → Account →0.99
    9. AIE1.00
    10. ★0.99
    11. events are the data — not an afterthought0.99
    12. Transaction → Decision → Policy → Employee0.97
    13. ★1.00
    14. ★1.00
    15. ★★0.74
    16. Structural Similarity1.00
    17. Explainable Decisions1.00
    18. Graph embeddings (FastRP) find similar1.00
    19. Full provenance chain: trace exactly1.00
    20. situations by network topology, not just text1.00
    21. how and why every decision was made1.00
    22. Cross-Session Knowledge1.00
    23. Production Ready0.99
    24. Information learned in one conversation0.99
    25. Neo4j: ACID compliance, enterprise scale,0.99
    26. is available in all future interactions1.00
    27. proven technology1.00
    28. Engineering the future of Al1.00
    29. Engineer1.00
    30. 2860.61

Transcript

152 cues· 2,753 words· 15,597 chars

  1. 0:14 Hello and welcome everybody to connecting the dots with context graphs.
  2. 0:20 My name is Steven Chen.
  3. 0:22 I run the developer relations team at Neo4j and you are in store for the power hour of context and graphs and all this technology.
  4. 0:31 So I'm the first speaker.
  5. 0:33 We have some other amazing talks after me.
  6. 0:35 So I hope you enjoy all the great content which you're going to see over the next
  7. 0:40 hour or so.
  8. 0:40 So what I'm gonna talk about is a bit about how we've all been feeling with the AI revolution where we are
  9. 0:49 trapped as engineers.
  10. 0:51 We are using AI coding tools, or maybe they're using us, where our work is being reviewed.
  11. 0:57 Who here has their work reviewed by an agent when they check in their PRs?
  12. 1:02 Yes, all of you.
  13. 1:03 So we're stuck in this limbo where we have amazing tools, we have amazing capabilities, but rather than us controlling them, they are controlling us.
  14. 1:14 And we would like to get to a state
  15. 1:18 where we're in control of this.
  16. 1:20 So we have to decide, is it going to be the,
  17. 1:24 The blue pill where we're stuck inside of this mire of disparate knowledge, stuck in Slack discussions and little customer threads and different enterprise systems which are all segregated and siloed and when we ask the agents to make critical business decisions or our applications to make critical business decisions with all this spread, it can't possibly give good answers because it doesn't have the context or do we wanna,
  18. 1:54 dive in and embrace the red pill, escape from the matrix, and have a system of reasoning where we actually have all these systems connected, all of our different enterprise data sources, previous decision traces, the reasoning calls of the tools to give us a more consolidated view of our enterprise stack, and escape from the matrix.
  19. 2:22 So who's in the escape club?
  20. 2:24 Who wants to break out?
  21. 2:25 Okay, hopefully if you're in the room, you're with me.
  22. 2:29 And guess who else is with us?
  23. 2:33 Gartner has now officially made context graphs as part of the AI hype cycle.
  24. 2:39 So we have been officially recognized by the analysts of the world.
  25. 2:46 they also realize that we're all stuck in this mire.
  26. 2:49 Foundation Capital actually started this thread with their $3 trillion startup opportunity post about how context graphs are gonna move forward the industry and dramatically change how we build applications.
  27. 3:01 And what I'll do is I'll show some demos and I'll talk about how we can move from being stuck
  28. 3:06 in this matrix stuck in this world and then become the superheroes of our organization and actually build the capability of the systems using technologies like knowledge graphs.
  29. 3:18 So knowledge graphs are a very powerful tool for us to aggregate all this information, create the connections, create the relationships.
  30. 3:25 And at a fundamental level, they hold nodes, which are people or things or companies or relationships.
  31. 3:35 you have relationships between nodes where in this case, Dan, those are properties.
  32. 3:42 lives with Anne, he drives her car apparently, so we know who wears the pants in this relationship.
  33. 3:48 And we have some embeddings on top of the car, so we're embedding vector information in it.
  34. 3:53 So we can also do similarity searches and kind of combine the best of both worlds with building information, but then also combining it with LLMs.
  35. 4:03 So when we take what LLMs are really good at,
  36. 4:06 this language, this reasoning, this creativity.
  37. 4:09 When we combine that with what knowledge graphs are really effective at, so knowledge, context, and enrichment, then we can start doing things with our data, like storing all these relationships together, visualizing them so we can get to the data which matters, finding hidden patterns, and then analyzing this and getting more insights, which will help power the context graph demonstrations, which I'm going to show you all.
  38. 4:37 So here's a simple example of how graphs power retrieval because I think it's good to understand what the difference is between a baseline LLM.
  39. 4:47 So this is a healthcare case.
  40. 4:49 What was the care plan associated with Andrea Jenkins emphysema?
  41. 4:53 And when you ask the LM, it has broad knowledge.
  42. 4:55 It understands a lot of information.
  43. 4:57 It knows what emphysema is.
  44. 4:58 It knows what standard practice is.
  45. 5:00 So it gives a very generic answer, preventing damage to the lungs, yada, yada, yada.
  46. 5:05 Now, when we give it a RAG system, so we go to vector database, now it has more context.
  47. 5:11 It knows a bit about the patient and their information.
  48. 5:14 And it tells you, maybe recommend some activities like respiratory theory, deep breathing, coughing exercises.
  49. 5:20 So this is pretty generic medical advice.
  50. 5:24 Now where we want to get to,

Open at this second