Gural Nuriyev

Autonomous AI research @theresidency, @cornell

[Gural Nuriyev @GuralNuriyev]

More research discussions coming soon. Time to lock the fk in for now.

[the residency @theresidency]

the gap between a lean team and a frontier lab is closing fast

recently, chapter 9 resident @GuralNuriyev hosted the frontier research club for a night on recursive self-improvement together with @marcelluzs & @shoobiedoo313

gural is giving lean AI teams an army of research agents running parallel experiments, so they can ship frontier work in-house, continuously, without the overhead of a big lab

[Gural Nuriyev @GuralNuriyev]

Decision models should be personal

[Gural Nuriyev @GuralNuriyev]

@SakanaAILabs Thank you. Good paper. Breaking the problem down is a critical skill to optimize

[Gural Nuriyev @GuralNuriyev]

@iaindunning To get to 2), we need Aha! moments from models.

To get Aha moments from current models, we need to engineer creativity by feeding existing ideas from other domains for combination.

To keep the scale manageable we logic prune for the most sane ideas

[Gural Nuriyev @GuralNuriyev]

@BarrAlexandra @officialanvith

[Gural Nuriyev @GuralNuriyev]

Come on bro, I am training a classifier rn, not as zero skill as it seems 😭

[Marcellus Amadeus @marcelluzs]

Scoring, routing and ranking can be classification too. Classification might represent most of AI/ML use cases. Thats why zero-shot classification models like Jev make sense, but note that zero technical breakthrough is required for that

[Gural Nuriyev @GuralNuriyev]

@marcelluzs I almost convinced them I was a LATAM VC

[Gural Nuriyev @GuralNuriyev]

@simonleon_xr @officialanvith lock in

[Gural Nuriyev @GuralNuriyev]

@marcelluzs We talked a lot bro

[Gural Nuriyev @GuralNuriyev]

Wild that I am not questioning the results. Just because the founders are that good.

[engineer cat 🐈 @MLCatttt]

Introducing Kardashev-0.7, the world’s first trained swarm made of 32 distinct models

Trained with RL for Population Scaling (RLPS), 32 models organically develop specialization & complementary capabilities, delivering frontier performance at:
- 0.007x ~ 0.02x of the inference cost
- 0.03x of the required memory

Civilization advances through different minds specializing and working together. We’re bringing that principle into AI

@BanburyRoadAI, we’re scaling intelligence by model count, toward civilizations of models that learn to build on one another

[Gural Nuriyev @GuralNuriyev]

@rahiltrips On-policy behavior

[Gural Nuriyev @GuralNuriyev]

@nicoleegong Get well, the world needs its events girlie

[Gural Nuriyev @GuralNuriyev]

@scale_at_gmi @theresidency @Kristopherfloyd @gmi_cloud @orbitremorse

[Gural Nuriyev @GuralNuriyev]

Holy vibes. Thank you everyone for coming.

@Kristopherfloyd @nicoleegong for making this possible.

God tier shoutout to @theresidency crew @TalhaZahids @orbitremorse @marcelluzs @shoobiedoo313 @rahiltrips @andyboyd1806 @ayaamghi @officialanvith for logistics

[Gural Nuriyev @GuralNuriyev]

Featured in San Francisco goes hard

[Gural Nuriyev @GuralNuriyev]

@orbitremorse You dont need sales, you need money

[Luigi Dintrono @Luigi_Dintrono]

Closing a box ✅ with @trainrobotsfast

[Gural Nuriyev @GuralNuriyev]

Nah nah nah

[Tobi Oyebade @orbitremorse]

I identify as a neolab

[Gural Nuriyev @GuralNuriyev]

@orbitremorse Pics aren't free, Tobi

[Gural Nuriyev @GuralNuriyev]

@brunoreisnet sales game > hbs