---
title: "Slides: Harnesses in AI: A Deep Dive — Tejas Kumar, IBM"
category: "slides"
video_id: "C_GG5g38vLU"
sourceLabels: ["Public YouTube video frames", "Public YouTube metadata"]
---

# Slides: Harnesses in AI: A Deep Dive — Tejas Kumar, IBM

## Source Video
[Harnesses in AI: A Deep Dive — Tejas Kumar, IBM](https://www.youtube.com/watch?v=C_GG5g38vLU)

## 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/C_GG5g38vLU/slide-001.jpg]]

OCR text:

> PLATINUM SPONSORS
> Braintrust WorkOS OpenAI

![[assets/slides/C_GG5g38vLU/slide-002.jpg]]

OCR text:

> ee At srry
> S Spotity
> Tt
> Teje
> if:
> Like “conta
> a oma ee

![[assets/slides/C_GG5g38vLU/slide-003.jpg]]

OCR text:

> AI Harnesses
> From first principles
> a Engineering the future of Al

![[assets/slides/C_GG5g38vLU/slide-004.jpg]]

OCR text:

> What Harness?
> e Eval Harness (ML Engineering)
> Ye ar: eens
> ames y F
> = | Al Engineer |
> EUROPE

![[assets/slides/C_GG5g38vLU/slide-005.jpg]]

OCR text:

> Xetejaskumar_
> Agent Harness
> AIE
> ToolRegistry
> Ager
> Tool Re
> AEnge
> Engineering the future of Al

![[assets/slides/C_GG5g38vLU/slide-006.jpg]]

OCR text:

> Xetejaskumar_
> Agent Harness
> AIE Tool Registry Model ContextManagement
> Guardrails
> Ager
> ToolRe.Model ontext Management uardrails
> AlEnge Engineering the future of Al

![[assets/slides/C_GG5g38vLU/slide-007.jpg]]

OCR text:

> 7-index.tsX
> 157-noex1s
> Cursor Light
> import(createT
> Lght(Visul Studio）Vs
> import
> createC
> Lght Moden Dett Lighe Mod
> import（runLoop o
> Ug+Detat Lt
> import （BrowseI sdlarizedLigh
> Abyss
> //trya shitty
> CursorDark
> Const MODELoCursorDark Midnight
> AIE
> const TASK=
> Upvote a story on Hacker News.
> Go to https://news.ycombinator.com.
> console.1og（Model:s（MODEL)）:
> console.log(Task:upvote on Hacker News\n'):
> const session = new BrowserSession();
> 10LShr
> AlEngineer
> Engineering the future of Al

![[assets/slides/C_GG5g38vLU/slide-008.jpg]]

OCR text:

> nr id me Gee wee
> eerie raters soos
> Spey o
> 4 const TASK «
> s@ Upvote a story on Hacker News
> 12 Go to mttos://news. ycombinator.com.
> reo 23 Call browser got stories to see ranked stories with their 10s and voted status.
> 
> * bd i4 Find the highest-ranked story mhere alreadyVoted is false.
> 
> * Pa 15 Chick 1ts upvote arrow using the oxact selector: afid«"up_STORYID*} (replace STOAYIO with the actual td).
> 
> * a 16 trim():
> 
> rd ee
> 18 console.jog{ Model: ${MODE:} >):
> 29 console. log{ Task: upvete on Hacker News\n ):
> 2
> ?. Const session 3 mew Arownerhessies();
> 22
> 230 try ¢
> 24 @wailt session.apen i
> 2s
> vf Const tools = createTools! session! ;
> er conat wessages © createContex: TASK;
> 2m const result 2 awalt runloopi MODEL, messages, tools! :
> we
> we aaARnena ina Vnfnemar Rhea) anounced
> SD tae to nee Lae a Teen lame MO ed ee tal rotten Bates
> ; omgeesll Enai H
> og : ngineering the future o

![[assets/slides/C_GG5g38vLU/slide-009.jpg]]

OCR text:

> 6 trope ttrere te tortng mason stews Seanase vee aare ine oo Ge we
> eo ea@ - ci ee
> 3 + pen pean cont el
> ee
> scum | prvenntes ener wonen
> “Sassen mae
> ( MELEE RARE
> Fecie wvneenanee
> Ped Ana y wae
> 5d Ad
> * * vusenoenen
> a in sewers
> oT Se ATi ad as A
> cnet itn wn-cun ena
> mags
> + egitenacs *
> Bairro dba
> ommsceramaes pater
> SRA SoTL
> therefor: Essen sexinmsel rs
> A Letom | PVE Sats lag |
> 5 toes EUROPE

![[assets/slides/C_GG5g38vLU/slide-010.jpg]]

OCR text:

> harness git:（e）npm run agent
> agont
> >tsx agent/7-index.ts
> Model:opena1/apt-3.5-turbo-0613
> Task:
> upvote on Hacker News
> [1ter1]callingmodel...tool_calle
> browser_navigate(（*url*:*https://news.ycombinator.com*)）..done
> [1ter2]callingmodel...tool_calls
> →browser_got_stories（（）)
> ouop
> [iter3]calling model...
> tqol_calls
> browser_cl1ck（(selector*:*a[1d-\up_4769787e\*]*)}).,.done
> AIE
> [iter 4] calling model...stop
> Answer:I have upvoted the highest-ranked story on Hacker News titled *LittleSnitch for Linux.
> topou:Aq poddos
> Iterations:4
> harness git:（0）
> AEngincer
> Engineering the future of Al

## 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.