---
title: "Slides: Why Agent Engineering — swyx"
category: "slides"
video_id: "5N33E9tC400"
sourceLabels: ["Public YouTube video frames", "Public YouTube metadata"]
---

# Slides: Why Agent Engineering — swyx

## Source Video
[Why Agent Engineering — swyx](https://www.youtube.com/watch?v=5N33E9tC400)

## Relationship To World's Fair 2026
These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.

## Related Scheduled Sessions
- No individual scheduled session mapping has been assigned yet; treat this as an event livestream deck.

## Extracted Slides
![[assets/slides/5N33E9tC400/slide-001.jpg]]

OCR text:

> re 7 ee a ' ee tf M o¢ atl * QP ato ;
> oe AO A see IRI) VO |” \ OV Ola
> i Se ne we
> a Sr)
> nd ql a a a0: re
> “ P oF , a & “we a a
> PT ae ‘y* ey ss
> Be o a PO ~¥e- 6
> ” ' ba a “ ‘ Pp be
> ae 8 N Ps oe O . & 7 i “—§
> oe + : . o
> a 4 ad 4. rr . a
> , . ; ae J + a
> -_ - oo , a. y =
> ————— ——————
> a _

![[assets/slides/5N33E9tC400/slide-002.jpg]]

OCR text:

> Wivee AYOIULOCOI MO) mare
> Rise of APbngineer SUTURE coe ECO hema CUO OL CT
> Tne Gas
> -
> ~ Saat PSatn ire ae
> 
> ee — le Fes Zee te yin =
> eT = mr Tih ey Teel
> 
> a Disrespecter
> 
> fare 3 Seat tee ro | ; =
> | Caegece ] Cond

![[assets/slides/5N33E9tC400/slide-003.jpg]]

OCR text:

> The Evolution of the AI Engineer (old)
> AIE Data/Research corstrainel API Prooluet/user constrnined
> M.Researcher/ MLErg AIEngineer Fullstack Engineer
> data
> modesandResoanchEngineers go rg to bulld etopAPl toolThs dagramha aiso beer griticized for the plecement ofevas and dita we certanly gree evals mortant e APl linespermeale-A/Engeenscan goleft tofinetunehost
> akspretrainscale data and general benchmar esalt: out A Engineers should certainly viewproduct-speohicdstaand ealsas theirob
> AGENTENGINEERING
> HTTPS://AI.ENGINEER

![[assets/slides/5N33E9tC400/slide-004.jpg]]

OCR text:

> The Evolution of the AI Engineer (ideal)
> MLE AIE SWE
> first time
> compute

![[assets/slides/5N33E9tC400/slide-005.jpg]]

OCR text:

> LatentSpace
> AIE
> AnncuncingAI EngineerSummitNYC:Allin on
> Agent Engineering+Leadership
> why we are
> all-in on agents
> AGENTENGINEERING
> HTTPS:I/AI.ENGINEER

![[assets/slides/5N33E9tC400/slide-006.jpg]]

OCR text:

> | ; : ; 5 “
> : ry
> cS ; -_ 7 - 7
> oe Se ls a. a :
> , 7 : rn Cor
> Ay 7 . be
> 4 : _
> o an 3 r") rs 7
> ee a ;
> . : : e ° rn , [ a
> o- Senter are — - i ee i ene
> an —_—_—_—
> ma |
> =
> |
> eR ri ~ |

![[assets/slides/5N33E9tC400/slide-007.jpg]]

OCR text:

> a bot . e 7 he a
> eS 7 -~ * ‘ : . Py . b
> : = me | a iS ra Ce 4 a
> . en Cor ee
> sl “ oS i ci |
> rg . 7 j > ad oe
> ad ; vee at
> Fs . . ol ° ry 7
> -e | a
> a _ a re ee pee
> | , a
> J ad
> a ; ;
> 7 in = nn ~~
> i? 1
> | ee 5 eae

![[assets/slides/5N33E9tC400/slide-008.jpg]]

OCR text:

> “2025 is the year of
> agents”
> AGENT ENGINEERING
> HTTPS://AI.ENGINEER

![[assets/slides/5N33E9tC400/slide-009.jpg]]

OCR text:

> https: latentspace p adept March 2024
> Alessio [00.21.47]: No, no, no, no | think like- You think so? No, no. I think like the,
> whatis possible today and like vehat 1s worth investing in, you know? And | think
> hke, [mean, people look at you and say, well, these quys are bu:iding agents. They
> needed 400 million to do it. So a lot of VCs are maybe like, oh, | would rather invest
> In something that is tacking on Al to an existing thing, which is like easier to get the
> market and kind of get some of the flywheel going. But I'm also surprised a lat of
> funders just don’t want to do agents. It's not even the funding Sometimes we look
> around and it's kke, why is nobody doing EE for x? wow
> David [00:22:17]: That's good to k never knew that before My sense
> from my limited perspective is thi Int Company poppsng up every day.
> Swyx [00:22:24]: So maybe I'm- They are. They are. But like [have advised people
> to take agents off of them title because it's so diluted.
> David (00:22:31): It's now so diluted.
> Swyx [00:22.32]: Yeah. So then it doesn’t stand for anything. Yeah.
> David (00:22'35}: That's a really good point

![[assets/slides/5N33E9tC400/slide-010.jpg]]

OCR text:

> TS :
> ATURE KEL DS Oy ea ek LEM Driven ad)
> Oriented Systems
> Sear a ee ame ery ace eee ae ee Ce
> Processes AVUTG era a ros
> Completion
> 
> et ae ar ee ee ee re ce a
> CS Ca cS aeee a or OAc ee Oe ee ed
> 
> | eee aoe rrr eas en So ast a BPM See

![[assets/slides/5N33E9tC400/slide-011.jpg]]

OCR text:

> a eg ee J
> re a mee er
> a —_ SF ee aa iio y aaa : : "
> ani ne
> Pe P a oe 7 - a a _ a ; . co ig
> il - me 7
> a aaa a. . ; Ce
> » va
> , a ‘
> , 2 7 __® ; ;
> > iT = SS SS
> é ra J 3 . * | ;
> i a i
> | po HT I
> ee Net LN re

![[assets/slides/5N33E9tC400/slide-012.jpg]]

OCR text:

> SidesCarnial Aagust19,2024
> why agents now?
> AIE Longinference/reasoning Better/More data MMLU/GPQASaturation Bettercapabilities Bettertool use 100%StructuredOutput,BFCLv3 MCP,Sandboxes.BrowsersSearch Better tools(AX) Memory/RAGframeworks/DBs
> AGENTENGINEERING
> HTTPS://AI.ENGINEER

![[assets/slides/5N33E9tC400/slide-013.jpg]]

OCR text:

> 5 | 7 i F : ~
> J 7 - an 7 ~ om at : - |
> a are on |
> iin an |
> Peary - ~* a ~ * cs . Loe
> - ta oe
> ~ sO ied ee To
> | aod o - a Ps * :
> : wo. a ..
> a os Oe a oe
> — om : rt
> } am, | Na 4
> ih RN . a
> ne rT | naa. 1
> oN 7 rT F ' |

![[assets/slides/5N33E9tC400/slide-014.jpg]]

OCR text:

> RL Finetuning
> Preference fine-tuning
> 10:35AM - 10:55AM
> Agent Engineering
> Room: TimesCenter Theater
> What RL Means for Agents
> See More Details
> Will Brown
> Researcher Morgan Stanley

![[assets/slides/5N33E9tC400/slide-015.jpg]]

OCR text:

> Agent PMF - ChatGPT.com
> AIE CNBC RATCHUST
> 三 MARETS EUSNESS NVESTINS TICH POLITICS
> OpenAltops4oomillionusersdespite DeepSeek'semergence TICN
> ABX NIOS The artficiat in er.up33% inless han hee Lighicap.told CNiC. bsOpenAnschiefopersing
> AGENTENGINEERING
> HTTPS:I/AI.ENGINEER

![[assets/slides/5N33E9tC400/slide-016.jpg]]

OCR text:

> a
> | Agent PME ChatGP'T.com
> Latent Space MMSE Ee BAR awe tae
> 7 oud t surprise you t ChatGPT had zero growth for the past year” ~ J. :
> ChatGPT stopped fe
> 7 growing for a year LZ oy
> sm se with GPT4/4 Turbo wae
> “ 7 a? - eee TOLD Ses ce EL SC
> | | oe | : AGENT ENGINEERING
> a? t A1.ENGINEER
> ee

![[assets/slides/5N33E9tC400/slide-017.jpg]]

OCR text:

> i a]
> ~ Lae)
> °
> aad Tal eres ~~ o™
> nN 5 . F
> al .
> - n #
> aq! a ‘ , .
> 
> , an : .
> ee __ : ri . . A
> SE aS nna ‘ sas Ene aeneene ee
> 
> a
> = + a a a
> | h j
> A i - , t ri |
> : ; - ;
> i ath
> el er lal el —— ee a iu ni a fm |

![[assets/slides/5N33E9tC400/slide-018.jpg]]

OCR text:

> The Evolution of the AI Engineer (ideal)
> MLE AIE SWE
> fixed time compute inference time compute reasoning
> models agents software

## Slide-Derived Subjects To Review
Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.