Videos tJFjeMBKbIY

Build for the Memo, Not the Demo — Shawn Chan, China Resources Holdings

index_state ready data_status ok

AI Engineer· published 2026-07-30· 0:24:22· en-US· indexed 2026-08-10 19:38

Open on YouTube

58 shot(s).

keyframes kept every frame deduplicated

What was stored

cues: 240; whisperx 240
chunks: 42; from 240 cues
keyframes: 16; kept of 58 captured
frames with text: 16; 220 lines read
chapters: 12; from the source metadata
keyframe bytes: 6.6 MB; word timings on 240 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage	state	model	started	took
`fetch`	done	—	2026-08-09 21:57	0s
`stt`	done	—	2026-08-09 03:26	26s
`chunk`	done	—	2026-08-09 03:27	0s
`text_embed`	done	—	2026-08-10 19:38	0s
`keyframe`	done	—	2026-08-09 03:27	2m 57s
`ocr`	done	—	2026-08-09 03:30	8s
`frame_embed`	done	—	2026-08-10 19:38	3s

Frames, and what the machine read

0:02 #0 done2 line(s)

shot 0·sharpness 455.2
1. AlEngineer0.95
2. World's Fair0.97
0:03 #1 done2 line(s)

shot 1·sharpness 670.0
1. AlEngineer0.96
2. World's Fair0.99
0:10 #2 done25 line(s)

shot 2·sharpness 2737.7
1. LAB & PLATINUM SPONSORS0.99
2. Amazon AGI Lab0.99
3. ANTHROP\C0.99
4. Google DeepMind1.00
5. MINIMAX0.98
6. OpenAl0.93
7. Akamai1.00
8. arize1.00
9. aws1.00
10. Braintrust1.00
11. bright data0.98
12. B1.00
13. Browserbase1.00
14. docker1.00
15. :neo4j0.90
16. ORACLE1.00
17. PayPal1.00
18. qodo1.00
19. reducto1.00
20. Sonar1.00
21. Makers of0.95
22. together.ai0.96
23. Unblocked1.00
24. WorkOS1.00
25. SonarQube1.00
0:25 #3 done2 line(s)

shot 3·sharpness 53.7
1. AlEngineer0.98
2. World's Fair0.99
0:35 #4 done12 line(s)

shot 4·sharpness 1282.9
1. AlEngineer0.96
2. CC0.59
3. World'sFair1.00
4. AI ENGINEER WORLD'S FAIR 2026· AI IN FINANCE0.98
5. Build for the Memo,1.00
6. PRESENTED BY1.00
7. Not the Demo0.97
8. Microsoft1.00
9. Notes from 200 investment committees — and how to get people like me to write you a check1.00
10. Shawn Chan1.00
11. Engineering the future of Al0.99
12. World'sFair0.99
0:53 #5 done15 line(s)

shot 5·sharpness 1384.1
1. AlEngineer0.98
2. Meet the two most confident voices in your life1.00
3. World'sFair1.00
4. PRESENTED BY1.00
5. Microsoft1.00
6. “I'm 100% sure."0.99
7. (It was wrong.)0.99
8. Your very confident friend at0.98
9. dinner1.00
10. Your Al system at 2am0.98
11. Confidence and correctness are not the same skill.0.99
12. Build for the Memo, Not the Demo0.98
13. 2/171.00
14. Engineering the future of Al0.97
15. World'sFair1.00
1:15 #6 done16 line(s)

shot 6·sharpness 1296.1
1. AlEngineer0.97
2. Meet the two most confident voices in your life1.00
3. World'sFair0.99
4. PRESENTED BY1.00
5. Microsoft1.00
6. “I'm 100% sure."0.96
7. (It was wrong.)0.99
8. Your very confident friend at0.97
9. dinner1.00
10. Your Al system at 2am0.98
11. Confidence and correctness are not the same skill.0.99
12. Build for the Memo, Not the Demo0.99
13. 2/171.00
14. TRACK 3· JULY 2, 20260.93
15. Alin Finance0.97
16. World'sFair1.00
1:45 #7 done14 line(s)

shot 7·sharpness 1156.8
1. AlEngineer0.97
2. World'sFair1.00
3. Meet the two most confident voices in your life1.00
4. “I'm 100% sure."0.99
5. (It was wrong.)0.99
6. Your very confident friend at0.97
7. dinner1.00
8. Your Al system at 2am0.98
9. Confidence and correctness are not the same skill.1.00
10. Build for the Memo, Not the Demo0.97
11. 2/171.00
12. TRACK 3· JULY 2,20260.95
13. Alin Finance0.97
14. World'sFair1.00
2:22 #8 skipped

shot 8·duplicate of #7
2:40 #9 skipped

shot 9·duplicate of #7
2:56 #10 done18 line(s)

shot 10·sharpness 1748.8
1. AlEngineer0.97
2. Fifteen years on the other side of the table1.00
3. World'sFair1.00
4. The boring bit about me — I'll make it quick0.98
5. 151.00
6. ~2001.00
7. 100s1.00
8. years, cross-border1.00
9. investment committees0.98
10. of pitch decks read0.99
11. Cross-border M&A, IPOs & strategic investment across HK, mainland China, the UK and the US — on names your LinkedIn feed0.99
12. won't shut up about. I have seen fonts that should be illegal.1.00
13. Money doesn't follow intelligence. Money follows trust.1.00
14. Build for the Memo, Not the Demo0.99
15. 3/171.00
16. TRACK 3• JULY 2,20260.95
17. Alin Finance0.98
18. World'sFair1.00
3:33 #11 skipped

shot 11·duplicate of #10
4:07 #12 skipped

shot 12·duplicate of #10
4:41 #13 skipped

shot 13·duplicate of #10
4:58 #14 skipped

shot 14·duplicate of #10
5:41 #15 skipped

shot 15·duplicate of #10
5:54 #16 done26 line(s)

shot 16·sharpness 1835.6
1. AlEngineer0.99
2. Two different machines0.98
3. World'sFair1.00
4. Most Al finance products build the first one and assume the second is just "more retrieval."1.00
5. THE DEMO1.00
6. THE MEMO1.00
7. PRESENTED BY1.00
8. Summarize this PDF1.00
9. Allocate $100M1.00
10. Microsoft1.00
11. • One clean document in0.98
12. • Hundreds of pages, many sources, several languages0.98
13. • A fluent answer out0.96
14. Sources that contradict each other0.99
15. • Optimizes for: "does this sound right?"0.99
16. • Optimizes for: “can a human defend every claim?"0.99
17. • Audience: impressed0.98
18. • Audience: skeptical & accountable0.98
19. • Stakes: a click0.95
20. • Stakes: nine figures0.98
21. Family dinner, except somebody's uncle brought a spreadsheet.1.00
22. Build for the Memo, Not the Demo0.98
23. 4/171.00
24. TRACK 3· JULY 2,20260.97
25. Al in Finance0.99
26. World'sFair1.00
6:19 #17 skipped

shot 17·duplicate of #16
7:02 #18 skipped

shot 18·duplicate of #16
7:36 #19 done14 line(s)

shot 19·sharpness 1471.5
1. AlEngineer0.96
2. The most expensive typo in history1.00
3. World'sFair1.00
4. February 2023 · a Big Tech Al launch demo · one question about a space telescope0.99
5. ~$100,000,000,0001.00
6. of market value, gone in one day — because of one unchecked sentence1.00
7. Nobody asked the question every junior analyst is trained to ask:1.00
8. "Wait — where does this claim come from? Did anyone check it?"0.99
9. Even the demo failed the memo test. There is no safe demo anymore.1.00
10. Build for the Memo, Not the Demo0.98
11. 5/171.00
12. TRACK 3· JULY 2,20260.97
13. AlinFinance1.00
14. World'sFair1.00
8:09 #20 skipped

shot 20·duplicate of #19
8:16 #21 skipped

shot 21·duplicate of #19
8:49 #22 skipped

shot 22·duplicate of #19
9:22 #23 done21 line(s)

shot 23·sharpness 1245.9
1. AlEngineer0.98
2. Six ways trust quietly breaks0.99
3. World's Fair0.98
4. Before anyone in the room even notices1.00
5. Which source do you believe?1.00
6. Do the numbers agree?1.00
7. Hide it, or show it?1.00
8. Not all sources are equal0.98
9. Everywhere, every time1.00
10. Contradictions are a gift1.00
11. Fact, or guess?1.00
12. Prove it in 30 seconds1.00
13. Whose name is on it?0.98
14. Keep them in separate boxes1.00
15. Or seven tabs and sweating0.99
16. A human always signs0.99
17. Build for the Memo, Not the Demo0.99
18. 6/171.00
19. TRACK 3• JULY 2, 20260.95
20. Al in Finance0.99
21. World's Fair0.98

Transcript

240 cues· 2,483 words· 13,733 chars

chunk 0 · 0:12–0:57 · 99 words · 528 chars
0:12 Good afternoon.
0:13 Thank you for being here, day four of our conference, a room with no windows, right just after lunch.
0:21 You are the strongest people in this building.
0:24 Let me start with a confession.
0:26 For 15 years, my job has been one thing.
0:30 As sitting in a room, a very smart, confident person hands me a piece of paper.
0:36 And before my company spends $100 million, I have to decide, do I believe this paper?
chunk 1 · 0:43–1:25 · 81 words · 466 chars
0:43 This year, my job still exactly that.
0:50 Except now, some of time, the very confident person handing me the paper is a chatbot.
0:58 And honestly, the chatbot is often better written than humans.
1:03 Better grammar, nicer formatting.
1:06 Never get defensive when you ask a follow-up question.
1:10 Very, very sure of itself.
1:13 The problem is being sure of yourself and being right are two different skills.
chunk 2 · 1:19–1:52 · 66 words · 347 chars
1:19 Some of most confident people I have ever met in finance were also the most wrong.
1:26 AI just learned that trick faster than the rest of us.
1:31 So here my whole talk.
1:34 in two sentences.
1:36 First, almost every AI finance product is built to impress people for five minutes.
1:43 And almost none are built to survive a room whose entire job is
chunk 3 · 1:50–2:33 · 87 words · 472 chars
1:50 to not be impressed.
1:52 Second, and this is the part I promised the organizers, I'd add, the exact same skills that fix your product is the skills that gets investors like me to write you a check.
2:09 Same muscle.
2:10 I would prove both.
2:11 And I'm telling this now, this week, because a lot of you are about to get pulled into exactly this.
chunk 4 · 2:21–3:02 · 83 words · 422 chars
2:21 Your CEO saw a demo somewhere.
2:24 Your biggest customer suddenly has a compliance department, or you are three months from raising your next round.
2:34 When any of those days arrive, I'd rather you to hear the hard part from someone who sits on the other side of the table than discover them live in a room in front of the people who build by hour
chunk 5 · 2:55–3:38 · 85 words · 464 chars
2:55 A quick bit about me.
2:58 I promise this is the boring part, and I will keep it fast.
3:03 15 years of cross-border deals, Hong Kong, mainland China, the UK, the US,
3:10 Mergers, IPOs, big strategic investment.
3:14 On names, you'd actually recognize the kind of companies that go public and your LinkedIn feed won't shut up and about it for a week.
chunk 6 · 3:26–4:09 · 68 words · 389 chars
3:26 Along the way, I've sat in about 200 investment committee meetings.
3:33 I have the gray hairs to prove it.
3:36 I checked.
3:37 It's not genetics.
3:40 And here, the part that matters for second half of this talk, I've also read hundreds and hundreds of pitch decks, funders' decks, bankers' decks.
3:55 14th slide, seed decks.
chunk 7 · 3:59–4:38 · 74 words · 404 chars
3:59 I have seen funds that should be illegal.
4:05 200 committee meetings taught me one thing.
4:10 No textbook says out loud.
4:12 The number on the page is not what gets a deal approved.
4:16 Trust is a number what gets a deal approved.
4:20 Money doesn't flow intelligence.
4:24 Money follows trust, and the trust is fragile, especially when the thing that wrote the page has never once in its entire life said the words, I'm not sure.
chunk 8 · 4:39–5:22 · 85 words · 470 chars
4:39 Let me give you one small test of what those rooms feel like.
4:45 Early in my career, a very polished, very expensive banker presented a beautiful slide full of confidence.
4:54 wrong numbers, one senior person in room asked one quiet question, where does this number come from?
5:03 The banker posted what felt like an entire physical quarter.

Chapters

0:00 A fifteen year confession
3:27 Two hundred investment committees
4:22 Money follows trust
5:46 Two machines: the demo and the memo
7:31 When one wrong sentence costs billions
9:45 Not every source deserves equal trust
11:22 The numbers have to agree
14:59 Facts and guesses stay separate
17:06 When fake citations reach a court
19:21 The airline chatbot that made a promise
21:30 Attach provenance, not a citation tab
22:52 The products that will win