Slides: Training Agentic Reasoners — Will Brown, Prime Intellect
Source Video
Training Agentic Reasoners — Will Brown, Prime Intellect
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:
aWS
eee)
@®Graphite WW Windsurf 4 MoneobB
Mdaily £3 augment code WorkOS
OCR text:
training agentic reasoners
will brown
@willecbb
research lead @ prime intellect
pr world’s fair 2025
‘i a
, | ; a Microsoft ary?
OCR text:
the big labs are all doing it
Continuing to scale reinforcement
learning AND - #
v7) |
Throughout the development of OpenAl 03, we've observed that
large-scale rewnforcement learning exhityts the same “more oF @) | = Ss
compute © better performance™ trend observed in GPT-series
pretrainng. By retracing the scaling path—ths time in RL-we've N = xX a ‘ ¥ x
pushed an additional order of magnitude in both training 9:24:02
compute and inference-time reasoning, yet still see clear ¢" ‘nn
performance gams. validating that the models’ performance
continues to improve the more they're allowed to think. At equal Is RL +LLMs enough for AGI? - :
latency and cost with OpenAl o1, 03 delrvers higher performance .
in ChatGPT— and we've validated that if we fet it think longer, its Sholto Douglas & Trenton Bricken
Performance keeps chmbing. 123K views + 13 days ago
FY
t4
- , rai)
— PLES ION
OCR text:
agents are also a thing ee
—— + i i ae See QE) SZ
et ee ee O.0}
3 7 Ce mren gn ts sae et se ("Ge -
7 @ sravegors witbe
' | _ wn a a
_ | | : ie meee °
ES
, i a Microsoft Qe?
~ >.
OCR text:
they’re kinda the same thing actually
uve = om I hae
= | | |
7 “Sg a - eae _ ,
: - [a 4 =
rr
| aws
: |
OCR text:
.
SFT warmup + small models = fun on just a few GPUs
as ‘ St ee |
results = vf_env.evaluatet S cares)
clientectient, ,
eodel=aodel_nane, He
sanpting_args:saapling_args,
Nus_samples=num_sanptes . i
) ,
hub _sodel_id="Qwen?. $-78-Math-Python-SFT™, _ -
}
trainer > SFTTrainert ;
mode l=nodel, : :
ergsrargs,
train dateset=dataset @ type: ignore Py
) . r
trainer. trarn()
“ aws
| a
x
wf me
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.