Videos RVxym6mmIns

How Forward Deployed Engineering is done at Cognition — Jia Wu

index_state ready data_status ok

AI Engineer· published 2026-07-28· 0:17:38· en-US· indexed 2026-08-10 19:38

Open on YouTube

51 shot(s).

keyframes kept every frame deduplicated

What was stored

cues: 210; whisperx 210
chunks: 31; from 210 cues
keyframes: 38; kept of 51 captured
frames with text: 38; 641 lines read
chapters: 10; from the source metadata
keyframe bytes: 6.6 MB; word timings on 210 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage	state	model	started	took
`fetch`	done	—	2026-08-09 22:09	0s
`stt`	done	—	2026-08-09 05:42	20s
`chunk`	done	—	2026-08-09 05:43	0s
`text_embed`	done	—	2026-08-10 19:38	1s
`keyframe`	done	—	2026-08-09 05:43	2m 17s
`ocr`	done	—	2026-08-09 05:45	15s
`frame_embed`	done	—	2026-08-10 19:38	6s

Frames, and what the machine read

0:02 #0 done2 line(s)

shot 0·sharpness 454.1
1. AlEngineer0.95
2. World's Fair0.97
0:03 #1 done2 line(s)

shot 1·sharpness 665.2
1. AIEngineer0.95
2. World's Fair0.99
0:10 #2 done24 line(s)

shot 2·sharpness 2746.2
1. LAB & PLATINUM SPONSORS0.99
2. Amazon AGI Lab0.98
3. ANTHROP\C1.00
4. Google DeepMind1.00
5. MINIMAX0.94
6. OpenAI0.92
7. Akamai1.00
8. arize0.93
9. aws1.00
10. Braintrust bright data0.98
11. B1.00
12. Browserbase1.00
13. docker1.00
14. :neo4j0.92
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
0:19 #3 done1 line(s)

shot 3·sharpness 337.6
1. World's Fair0.96
0:29 #4 done12 line(s)

shot 4·sharpness 1871.1
1. AlEngineer0.99
2. World'sFair1.00
3. Devin, Deployed0.99
4. The Forward Deployed motion that makes Al Engineering real1.00
5. PRESENTED BY1.00
6. Microsoft1.00
7. Jia Wu0.96
8. Deployed Engineering Lead1.00
9. cognition.ai1.00
10. PROPRIETARY & CONFIDENTIAL0.99
11. Workaar0.68
12. Engineering the future of Al1.00
0:36 #5 done7 line(s)

shot 5·sharpness 1128.2
1. AlEngineer0.99
2. World's Fair0.97
3. PRESENTED BY1.00
4. Microsoft1.00
5. Remember Devin?1.00
6. PROPRIETARY & CONFIDENTIAL0.99
7. Engineering the future of Al1.00
0:46 #6 done1 line(s)

shot 6·sharpness 325.0
1. World's Fair0.99
0:49 #7 done10 line(s)

shot 7·sharpness 1363.0
1. AlEngineer0.99
2. World'sFair1.00
3. Oras on March 12, 2024 | prev | next [-]0.99
4. While it is a progress, its far away from being useful to be a software engineer.0.99
5. PRESENTED BY1.00
6. Microsoft1.00
7. Remember Devin?1.00
8. PROPRIETARY &CONFDENTIAL0.97
9. 20.87
10. Engineering the future of Al0.99
0:53 #8 done15 line(s)

shot 8·sharpness 2096.3
1. AlEngineer0.95
2. World'sFair0.97
3. Oras on March 12, 2024 | prev | next [-]0.99
4. While it is a progress, its far away from being useful to be a software engineer.1.00
5. PRESENTED BY0.97
6. Microsoft1.00
7. Remember Devin?1.00
8. ▲ HarHarVeryFunny on March 13, 2024 | root | parent | next [-]0.98
9. A 13% success rate is both wildly impressive and also WAY below the level where I would personally1.00
10. find something like this useful. I can't even see reaching for a tool that I knew would fail 90% of the1.00
11. time, unless I was desperate and out of ideas.0.99
12. PROPRIETARY &CONFIDENTIAL0.97
13. 21.00
14. TRACK 8·JUNE 30, 20260.96
15. Forward Deployed Engineering0.99
1:04 #9 done13 line(s)

shot 9·sharpness 1401.7
1. AlEngineer0.97
2. World'sFair1.00
3. Remember1.00
4. Devin?1.00
5. It's actually1.00
6. PRESENTED BY1.00
7. good now0.94
8. Microsoft1.00
9. $Cognition0.89
10. PROPRIETARY & CONFIDENTIAL0.98
11. WorksFar0.77
12. TRACK 8· JUNE 30, 20260.96
13. Forward Deployed Engineering1.00
1:39 #10 done11 line(s)

shot 10·sharpness 1461.8
1. AlEngineer0.97
2. Devin1.00
3. World'sFair1.00
4. Devin CLI0.98
5. Devin Cloud1.00
6. Devin Desktop1.00
7. (formerly Windsurf)1.00
8. PROPRIETARY & CONFIDENTIAL0.98
9. Workdfa ar0.73
10. TRACK 8• JUNE 30, 20260.96
11. Forward Deployed Engineering1.00
1:50 #11 done50 line(s)

shot 11·sharpness 2321.6
1. AlEngineer0.94
2. The Cognition Platform1.00
3. World's Fair0.97
4. Built to meet you where you are and support your entire SDLC.1.00
5. DeepWiki1.00
6. UNDERSTAND1.00
7. LOCAL · SYNC / ASYNC0.91
8. CLOUD - AI SOFTWARE ENGINEER0.99
9. QUALITY GATE1.00
10. Auto-generated1.00
11. documentation1.00
12. Devin CLI0.98
13. Devin Desktop1.00
14. Devin Cloud1.00
15. Devin Review1.00
16. Exploring unfamilia code, onboarding,0.97
17. BEST FOR1.00
18. Terminal-based agent0.99
19. IDE + Agent Manager0.98
20. Autonomous software engineer0.99
21. PR review tool1.00
22. answering architectural questions.1.00
23. Ask Devin1.00
24. ASSIST/SCOPE1.00
25. Prompt-building0.97
26. assistance1.00
27. Shaping specs, drafting tickets,0.99
28. scoping work before it starts.1.00
29. BEST FOR1.00
30. Repetitive tasks, scripting, etc1.00
31. benefits from a tight feedback loop.1.00
32. Creative, exploratory work that1.00
33. tasks that can run in parallel.1.00
34. Well-scoped, repetitive, or multi-hour0.99
35. BEST FOR1.00
36. across every PR that lands.1.00
37. Quality control and consistent review0.99
38. BEST FOR1.00
39. Enterprise security & governance1.00
40. Persistent & compounding knowledge0.99
41. Unified consumption0.99
42. Flexible1.00
43. Auditable0.99
44. Shared1.00
45. One pool of ACUS· The right surface for the job0.98
46. PROPRIETARY & CONFIDENTIAL0.99
47. 51.00
48. Workdfa Far0.78
49. TRACK 8· JUNE 30, 20260.96
50. Forward Deployed Engineering1.00
2:08 #12 done20 line(s)

shot 12·sharpness 2182.6
1. AlEngineer0.97
2. Impact on Software Engineering Efficiency1.00
3. World's Fair0.98
4. Merged PR volume0.96
5. Engineering headcount (R&D)1.00
6. 8X0.96
7. 7.7X0.98
8. 5.3X1.00
9. 7X0.86
10. 3X0.99
11. 2X0.97
12. DEC1.00
13. JAN 20261.00
14. FEB1.00
15. MAR1.00
16. At Cognition over past 6 months: 7x code shipped, ~same size team1.00
17. PROPRIETARY & CONFIDENTIAL0.99
18. Workfs Far0.73
19. TRACK 8· JUNE 30, 20260.95
20. Forward Deployed Engineering1.00
2:39 #13 done15 line(s)

shot 13·sharpness 1425.4
1. AlEngineer0.97
2. Impact across the Enterprise1.00
3. World'sFair1.00
4. Oct1.00
5. Nov1.00
6. Dec1.00
7. Jan0.99
8. Feb1.00
9. Mar1.00
10. Apr1.00
11. May1.00
12. Jun0.99
13. PROPRIETARY & CONFIDENTIAL0.99
14. TRACK 8• JUNE 30, 20260.96
15. Forward Deployed Engineering0.99
2:43 #14 done7 line(s)

shot 14·sharpness 1746.4
1. AlEngineer0.99
2. World'sFair1.00
3. Forward Deployed Engineering1.00
4. at Cognition1.00
5. PROPRIETARY & CONFIDENTIAL0.99
6. TRACK 8· JUNE 30, 20260.96
7. Forward Deployed Engineering1.00
3:05 #15 done9 line(s)

shot 15·sharpness 1612.1
1. AlEngineer0.97
2. World'sFair1.00
3. Product1.00
4. Problem1.00
5. PROPRIETARY & CONFIDENTIAL0.98
6. 101.00
7. Workdfa ar0.74
8. TRACK 8· JUNE 30, 20260.96
9. Forward Deployed Engineering0.99
3:30 #16 done8 line(s)

shot 16·sharpness 1672.9
1. AlEngineer0.97
2. World's Fair0.96
3. PMF1.00
4. Product ∩ Problems1.00
5. PROPRIETARY & CONFIDENTIAL0.99
6. n10.65
7. TRACK 8• JUNE 30, 20260.97
8. Forward Deployed Engineering0.99
3:36 #17 done9 line(s)

shot 17·sharpness 1799.0
1. AlEngineer0.99
2. World'sFair1.00
3. FDE at Cognition1.00
4. Product ∩ Problems1.00
5. PROPRIETARY & CONFIDENTIAL0.98
6. 121.00
7. Workdfa Far0.76
8. TRACK 8·• JUNE 30, 20260.94
9. Forward Deployed Engineering0.99
3:50 #18 done9 line(s)

shot 18·sharpness 1957.5
1. AlEngineer0.98
2. World'sFair0.99
3. FDE is about deeply understanding0.99
4. the problem at hand1.00
5. PROPRIETARY &CONFIDENTIAL0.99
6. 131.00
7. Workfa ar0.77
8. TRACK 8· JUNE 30, 20260.96
9. Forward Deployed Engineering1.00
3:52 #19 done28 line(s)

shot 19·sharpness 2291.4
1. AlEngineer0.98
2. Cognition is building across the1.00
3. World'sFair1.00
4. [entire SDLC]1.00
5. Understanding1.00
6. Design1.00
7. Coding1.00
8. Testing1.00
9. Review & Deploy0.98
10. Maintenance1.00
11. C0.99
12. DeepWiki0.98
13. AskDevin +0.99
14. Skills1.00
15. Session1.00
16. Devin1.00
17. Feedback Loop0.99
18. Devin's1.00
19. Reviews1.00
20. Devin1.00
21. Event-Driven1.00
22. Automations1.00
23. APIs0.89
24. PROPRIETARY &CONFIDENTIAL0.99
25. 140.77
26. Workdfs Far0.76
27. TRACK 8· JUNE 30, 20260.96
28. Forward Deployed Engineering1.00
4:10 #20 skipped

shot 20·duplicate of #19
4:46 #21 skipped

shot 21·duplicate of #19
5:00 #22 skipped

shot 22·duplicate of #19
5:34 #23 done14 line(s)

shot 23·sharpness 2086.3
1. FDEs map Devin's Capabilities to1.00
2. AlEngineer0.99
3. World'sFair1.00
4. Customer Problems0.99
5. The SDLC is complex ...0.98
6. Deploying Devin without clear direction leads to wasted tokens0.99
7. PRESENTED BY1.00
8. • Which initiatives are the highest leverage for the business?0.99
9. Microsoft1.00
10. PROPRIETARY &CONFIDENTIAL0.98
11. 151.00
12. Workda0.73
13. TRACK 8· JUNE 30, 20260.96
14. Forward Deployed Engineering1.00

Transcript

210 cues· 3,365 words· 18,598 chars

chunk 0 · 0:12–0:55 · 149 words · 793 chars
0:12 It's nice to be here.
0:13 I appreciate you all.
0:14 My name is Jia.
0:15 I'm a deployed engineering lead at Cognition.
0:17 And today, hopefully what you'll take away from this is that how we deploy Devon in the field is very much a function of how we view deployed engineering at Cognition.
0:26 So how the forward deployed motion makes AI engineering actually real.
0:31 Before I start, how many people have heard of Devon or know of Devon?
0:36 Very cool.
0:36 And I'm not talking about the Devon of today.
0:38 I'm talking about the Devon back in 2024, when we first released, and it was like, oh, sweet bench, 13%.
chunk 1 · 0:43–1:26 · 150 words · 768 chars
0:43 We're so back.
0:44 And as engineers, we were like, we're so cooked.
0:46 But I mean, after a week, everyone's like, oh, this is actually not that useful.
0:51 I would only use this if I was desperate and out of ideas.
0:54 That's 2024.
0:56 And let it be said that we have a sense of humor, because we took this and we ran with it.
1:00 I'm sure you've seen all these ads around SF.
1:03 We're actually good now.
1:05 So the reason why we're good is I'll talk a little bit about the product surface area a little bit, just to give you folks a little bit of context for who might not have used Devon before, who might not have exposure to something like Cognition.
chunk 2 · 1:18–2:00 · 144 words · 781 chars
1:18 So if you've used Cloud Code, we also expose a CLI.
1:21 If you've used something like Cursor, Windsurf, we also expose an interface such as an IDE.
1:27 Actually, how many people know of Windsurf or have used Windsurf in the past?
1:31 Sweet.
1:33 So I come over from the Windsurf side after the Windsurf acquisition, great times.
1:38 And then what we were actually known for specifically is for Devon Cloud or the Devon Cloud agent.
1:44 And I'm not gonna bore you to death talking about like all the components and features and everything that comprise of the actual product.
chunk 3 · 1:50–2:25 · 130 words · 681 chars
1:50 I'm not here to sell you on that.
1:51 What I am here to sell you on is we are one of the premier software engineering functions across the enterprise.
1:56 And we're actually able to deliver impact on a global scale.
2:00 What do I mean by that?
2:01 We can take a pause and take a look at this figure.
2:04 So internally at Cognition, over the last six months, for better or for worse, we might've been behind on hiring, but using our agent, we were able to ship almost an order of magnitude more good quality, robust PRs across the organization.
2:18 So it's a step function increase in the amount of engineering leverage that we can have by deploying our own agent.
chunk 4 · 2:24–3:01 · 125 words · 709 chars
2:24 You don't have to take our word for this.
2:26 If we can take a look at the specifics of how we're actually being consumed, how we're being utilized across the enterprise, it's a parabolic growth of how companies are adopting our agent, deploying our agent, and using it in multiple different use cases and multiple different scenarios.
2:41 So what does that mean?
2:43 How does this actually happen?
2:44 Well, it can only happen with the four deployed engineers at Cognition.
2:48 And I'm gonna frame up the problem from a couple of buckets.
2:51 So there's two circles in front of you on the screen.
2:53 One of them can represent the domain of a product.
chunk 5 · 2:57–3:41 · 142 words · 850 chars
2:57 As a business, as a software engineering organization, obviously you have a product.
3:02 you obviously also on the other side, on the left-hand side, right-hand side for you guys, you specifically have like a bucket of problems that you're looking to solve, right?
3:10 You have a product, you have problems that you're trying to solve, and the intersection of this two or whatever you would call it is the product market fit, right?
3:18 Hopefully you have a pretty good overlap in the sense that whatever it is that your company does, you can actually bring value to customers and hopefully whatever problems the customer has, you can solve with your company stuff.
chunk 6 · 3:31–4:14 · 143 words · 834 chars
3:31 So the forward deployment motion at Cognition essentially aims to maximize the overlap between the products that we typically build and the problems that we are experiencing across the enterprise.
3:42 So what does that mean, right?
3:43 The first fundamental concept that I would like to convey is that forward deployed engineers at Cog deeply understand the problem space at hand.
3:52 And specifically, if we think of the problem of software engineering, and I'm just gonna mask the features at the bottom, we don't really care about those, but if we think about when you go ahead to take some sort of code base, some sort of implementation, and you need to build features, you need to maintain that software, you need to review, deploy, maintain that software, all of these steps, all of these functions have a lot of business value behind them.
chunk 7 · 4:15–4:59 · 164 words · 881 chars
4:15 You can only do...

Chapters

0:00 Introduction: deployed engineering at Cognition
2:05 Step function to parabolic productivity
3:35 Where products meet the customer's problems
5:16 Pointing agents at the highest leverage work
7:07 How deployment challenges recur
8:34 What a deployed engineer actually is
10:25 Measuring outcomes, not token usage
12:56 The 82% result, measured before and after
14:38 Developers plus Devin, autonomously
15:53 Core values: correctness and customer success