Slides: 120k players in a week: Lessons from the first viral CLIP app: Joseph Nelson
Source Video
120k players in a week: Lessons from the first viral CLIP app: Joseph Nelson
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
OCR text:
an 86a
a Oey ie
A Ss ee
aa a
Mi
OCR text:
AIE
AUTOgPt
smol?
OCR text:
The Premise: Convince an Al you’re the best artist
1) GPT: Generate a prompt for users to draw
eo User: Draws prompt in MS Paint interface
6 CLIP: Judge vector similarity of text
prompt and user image
@ ???: 120k players in one week, 7 requests
per second
OCR text:
Prompt: A Raccoon Driving a Tractor
Draw a raccoon driving a tractor Draw araccoon driving atractor | Oraw araccoon driving a tractor
Global Ranking: 462 out of 10.187 wubewssions Gtobel Ranking 335 out of 10.187 tudmesont Gobel Ranking: 3 out of 10.187 submessons
: cs ™ 7
\ fo} = ©
4 | Ly nneron?
revnonn - QR revscumn GQ"
OCR text:
Prompt: A Bumblebee that Loves Capitalism
Draw a bumblebee that loves capitaliam Draw a bumblebee that loves capitalism Draw a bumblebee that loves capitalism
Cadet Resharg. 216 ot of 14758 woemvom Globel Renting $51 ot of 14.7S8 wornnwons Ghebet Rentong 288 ont of 14.758 womeveons
eo wy aR
&
, ¢ ww, }
wy
oe: ‘mm
Tevacan - QD Tavacan - QD Tevacan - GD
6
OCR text:
CLIP: Trained on 400M image/text pairs, OpenAl, 2021
text_embedding: how CLIP Paint.wtf Scoring
maps the paint.wtf prompt into
its feature space CLIP’s Interpretation
x, of the Paint.wtf Prompt
Image_embedding: how CLIP cuPtissarpraention of
maps the user submission into saiaaeteiaiiaaiienaiiniias-
its feature space /e
Winning paint.wtf: minimizing Cosine Similarity S|
distance between the prompt i
and the user drawing
7
OCR text:
AIE
SUN
AUTOgPt
smoo
OCR text:
SILAR.
C
statpy
apy>no
iapertc2
ortisterence
iaport spevisiesassv
anotator-sv.Bcsknoteter()
etrender(resvtt,
Labge)t
piet(resolt)
ieshovl
"ootpot".
anotater-annetate(
sere-iavpe.
detectionsniv.De
(v2.vaitKeyi1)
AT
iaference.Streanc
source·2,
modet-"rock-
wesain_thresd-Trut,
ovtput_chanoeorder
osredicties.render
LARColS
OCR text:
"
‘ ALENGN
7 MANN VY
OCR text:
tent rd c-cd Pn
e ao .
.
ec rs seth a
te torr treat
LO
Oe cc ce ce crs ee ee
De ee ee eer a
SO re
o cee os
Eee cc
Ce Pr cr
rs
ea aaa Me Mle oe a
ee ee re re rr ae
a are
a cr
ney Soy Ca
rc a cc cy oe
Ce a ea 2 ee re 2)
a a re rs ne
Pa
re
=
OCR text:
Ch ie way ee oO 8 of -
e a ,
a .
Se ee a Se 2 Pe Pe
| mn ee a Te ne 0
Sa a a
CS Se SO
SCE PM eee Oe eR
ee oe or ee ce a ce oe ot
Ce Cn ee ae Per Corea ot
A Pi a I ae ee
Se ee rs cre ae ee Sa ee
ee Oe See PC
i a ee Se ee 2 ce oe a oor er Paar
ee eC
Se SO
Se an SS OT Pe 2 CS
Se a a ce ce Ce oe os
re Pe Ce ar Cn eT ee Se er Cee ae or
SOCDDIPD Me 22 a Ltts@aete OL roreaieteta at os tarrertze at
“st ce ee Se ee ee See cree)
ee a ee
Se a EE RS PT PO ROY Oe TO
a Se Se Se
re ow
Ca a OS I
ee cr
Sa eS
Cy
OCR text:
rae] = = |
7 @FINIE 4 supabase
OCR text:
ove. pu
arrare le
32
cw2. im
cv2.Wa
inference,
source
model
; use_me
a CUR aot
OCR text:
oes ry fa) : a
D paint.wtf @
Convince an Ai thet you're the beet artiet.
Best of All Time
toverer’ tenons
" wre men ly inp. evan i ethan ont tare o> Oped OF
Sent .
9p} Ssimoln
== AUTO GPL,
OCR text:
. |. : : as > er soe —_
ci o oar] cr ee 2) .
| Oraw a gorilla gardening with grapes suséIT
oorttrtwti‘a‘SOOOOOOS
70
O°.
; or
+
a
s
*.
as
QQ! y
eT
OCR text:
Seog | oe -
ath arn
Pid
OCR text:
i a
he 47,
» T/O oa
iN
AlEngineer | | |e
ST ~
cf wm ON
MiG
[xine] “ eee Fr dan “3 a lh grcpes Lo
Aree,
OCR text:
LIVECODING
AlEngineer
SUMMIT
Let's be 1000x Engineers Today
A!F
RUTO
OCR text:
Lessons from Building Paint.wtf with CLIP Lee
CLIP Can Read
Gomanenne sot tet ios atone ance merino sore "
a ae A faccoo}
2Y __
ne 0 AR Wry A
roan - |
10
OCR text:
Lessons from Building Paint.wtf with CLIP ; Se
CLIP Can Read
ee teawr teen tenene sheila lala
ae — Reo A facc Ooh
eae ARivivry Xs
= #
. TRACTOR.
vt J ter scam - GD ravacan - ED
ed
te
OCR text:
Lessons from Building Paint.wtf with CLIP
Roboflow Inference Makes Life Easy
Serving done right. an aad
Built on the lessons of serving 100M of API ; interop - . -
calls and thousands of hours of video. ee
e Maximize throughput on your target ee
hardware (GPU, CPU, edge) ie ao
e Use ready-to-go foundation models ini
e Pullin over 50k pretrained models from a aed
Roboflow Universe community are eas
14
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.