/Tech11h ago

Microsoft VP of AI Nando de Freitas argues that fully automating LLM and multimodal model training is inevitable

Future research will focus on physical and scientific domains.

12105187121.5K

#22

Original post

Nando de Freitas@NandoDF#22inTech

This is an important achievement, and probably the first in a predictable sequence of results that will lead to fully automating LLM and OMNI training and serving.

As someone who has helped build some of the best multimodal and language models, I don’t see how this could play out any other way.

What challenges will remain then? For me: environments and applications that matter like healthcare, materials, scientific exploration, and above all energy harnessing and lasting energy storage. With that we can solve carbon capture and end hunger.

Recursive@Recursive_SI

http://x.com/i/article/2064569799313940481

3:03 AM · Jun 13, 2026 · 20.3K Views

Sentiment

Positive users are optimistic about automating LLM and Omni model training because it promises genuine improvements in iteration with clear direction and real results from thoughtful leaders.

Pos

100.0%

Neg

0.0%

3 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS1.5KBOOKMARKS1LIKES12RETWEETS1REPLIES4

Jeff Clune@jeffclune

Thanks @NandoDF!! 💯 agreed. Glad you like the work and direction!!

Nando de Freitas@NandoDF

This is an important achievement, and probably the first in a predictable sequence of results that will lead to fully automating LLM and OMNI training and serving.

As someone who has helped build some of the best multimodal and language models, I don’t see how this could play out any other way.

1h1.5K121

GEE!@KhieEmm

@NandoDF Too true. The energy problem is the real final boss.

9h1191

Lunari@0x_lun

@NandoDF automating the training loop is the easy part to celebrate

what nobody is talking about is who owns the eval criteria once humans are out of that loop too

11h1041

J@parkhjaey

@NandoDF Energy storage is the real final boss.

7h701

Strata@ChainZenit

@NandoDF this could genuinely change how we iterate on models.

10h82

tsunami_crypto@ls_brd

@NandoDF gotta respect the build-up, but it does still feel like the first 10% repeating before the real shift

11h68

tsunami_crypto@ls_brd

@jeffclune @NandoDF appreciate a lead who actually reads the room and shows up right

rare combo in this space ngl

1h31

Avery@wveriy

@NandoDF The real work begins when the model hits the real world.

8h4

betraidx@betraidx

@jeffclune @NandoDF clear direction matters more than hype and this ones actually grounded in real results

Alex YGift@Radipdegen

@NandoDF training loops training themselves is the part nobody talks about

its either gonna be beautiful or terrifying

11h