Saksham Consul

Training @bfl_ml | Ex @Nvidia, @Stanford | Builder

[Saksham Consul @TheNoise2Signal]

100%

[Anjney Midha @AnjneyMidha]

we are still so so early on unlocking the true potential of diffusion models

[Saksham Consul @TheNoise2Signal]

Bringing the bowtie back in fashion

[Black Forest Labs @bfl_ai]

Our co-founder and CEO, @robrombach, sat down with President Trump, President von der Leyen, President Macron, and other world leaders at the G7 to stress the vital role of open innovation in AI. With openness under pressure around the world, Robin urged governments and industry to make open and responsible development the norm, not the exception. Check out his speech below!

https://t.co/REOofUXPaM

[Saksham Consul @TheNoise2Signal]

In case someone else is stuck, hope this helps: https://t.co/KgQ0Vullyr

[Saksham Consul @TheNoise2Signal]

Small victories of the day: Got my DGX sparks on my k8s cluster complete with the ever dreaded network operator running.

[Saksham Consul @TheNoise2Signal]

@Kalshi_Finance Google’s Nvidia??

[Matthew Prince 🌥 @eastdakota]

Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa

[Saksham Consul @TheNoise2Signal]

@EHuanglu Which model are you using?

[Saksham Consul @TheNoise2Signal]

Excited for what's next to come!

[Black Forest Labs @bfl_ai]

"The fact that someone like Martin Scorsese — one of the greatest, most impressive filmmakers to exist — is using our technology and curious about exploring it...it's such a great proof point that this works.”

- our CEO @robrombach in an interview with @brooksbarnes for the @nytimes. They discussed why Martin Scorsese joined BFL as an advisor.

At Black Forest Labs, we're building visual intelligence: AI models that can understand and reason in the physical and digital worlds. Scorsese is helping shape how our models serve creators who care deeply about their craft, whether they're storytellers, designers, engineers, or roboticists.

Link to article in the thread.

[Saksham Consul @TheNoise2Signal]

@StockSavvyShay Oof another neolab https://t.co/OOJDCxbU3i

[Saksham Consul @TheNoise2Signal]

Gotta rename ourselves now

[Ronan @Ronanchamberss]

he said Black Mountain Labs lol

[Saksham Consul @TheNoise2Signal]

@karan_bangia made it with the help of claude. SVG skills

[Saksham Consul @TheNoise2Signal]

How does frontier training use 2,048 GPUs?

Because there are five dimensions you can split work across - and at scale, you use all of them at once.

Hope this helps demystify some of the model training frameworks out there:

[Saksham Consul @TheNoise2Signal]

@AnjneyMidha I miss the dip of Oct'25. The good ol' days lmao

[Saksham Consul @TheNoise2Signal]

@immortaldip @PandaAshwinee Of the 5 parallelism methods, DP has the highest ROI, and TP,CP come to play with with HBM constraints. PP is the least ROI imo, given the eng complexity, but frontier labs still use them, as evidenced by Deepseek Dualpipe schedule. EP is MOE specific. Different trade-offs

[Saksham Consul @TheNoise2Signal]

@PandaAshwinee in what way lol?

[Saksham Consul @TheNoise2Signal]

More details can be found:
https://t.co/d8MCWISujS

[Saksham Consul @TheNoise2Signal]

https://t.co/C6EStjbbtw

[Saksham Consul @TheNoise2Signal]

Will be at #MLSys2026! Would love to catch up with familiar faces and meet new ones 👋 especially keen to geek out on highly scalable, fault-tolerant training. If you're going, drop a reply or DM. Let's connect!

[Saksham Consul @TheNoise2Signal]

The systems depth is there if you look
for it.

Full paper: [https://t.co/4df9TaevCq]

[Saksham Consul @TheNoise2Signal]

Decoupled DiLoCo is one of the most practically important training papers in a while. Spent a lot of time working through the systems implications. A few things I had to figure out myself that might save you time 🧵

[Saksham Consul @TheNoise2Signal]

10/ Overall: Decoupled DiLoCo is one of the most practically important training papers in a while.

The core insight - prioritize availability
and partition tolerance over consistency —-
will matter more as training scales to
geo-distributed clusters.