appughar

CEO & co-founder, Mobius Labs (exited to Dropbox). Making Dash multimodal. Trying to make computers see.

[appughar @appughar]

@dennismuellr @paulg Isn’t litigation from standard VC financings rare on both sides of the Atlantic? The notary process seems to solve a non-problem while adding friction. As conveyed by the notary, some capital increases were rejected by the commercial court because the PoA was not notarized.

[appughar @appughar]

@dennismuellr @paulg Even with a PoA, the notary accepts only notarized and apostilled versions. This pushed some our us-based angels to consider exiting in later rounds, forcing extensive firefighting. Nearly all stakeholders see the notary process as adding negative value, yet it persists.

[Mobius Labs @Mobius_Labs]

If you’re at @EurIPSConf this week and looking for your next ML role, ping us. Few of us from @Dropbox including @mobicham @appughar will be there Happy to chat about roles in multimodal, search, and video-AI.

[appughar @appughar]

After an incredible 7-year roller coaster of building @Mobius_Labs — a journey filled with both amazing excitement and moments of grief — I’m super thrilled to join @Dropbox along with my fellow Mobius team to bring multimodality to Dropbox Dash.

[Mobius Labs @Mobius_Labs]

@Dropbox is acquiring our IP, and the Mobius core tech team is joining them. A huge milestone for us!

https://t.co/pglCNRmneK

[appughar @appughar]

A big part of my PhD was on manifold optimization https://t.co/CAx0d1UwKS. Back then, training data were only in the 1000s. I never thought I’d hear about the Stiefel manifold again in the era of stochastic gradients—where massive datasets often make simpler, heuristic methods work just fine. Exciting to see these ideas resurface!

[Thinking Machines @thinkymachines]

Efficient training of neural networks is difficult. Our second Connectionism post introduces Modular Manifolds, a theoretical step toward more stable and performant training by co-designing neural net optimizers with manifold constraints on weight matrices.

https://t.co/PGG4zy3u23

We explore a fundamental understanding of the geometry of neural network optimization.

[Mobius Labs @Mobius_Labs]

GemLite now supports @vllm_project vllm V1, which brings up to 1.25x faster inference speed vs V0!

https://t.co/hgP3fpLXjv

[mobicham @mobicham]

Well optimized Triton kernels can perform very well end-2-end, even competing with highly optimized kernels like Marlin.

[appughar @appughar]

@dieworkwear Not arguing against industrialisation or homogenisation. But comparing clothing to visual arts isn’t really apples to apples. Clothing demand is limited to a few hundred items, mostly utilitarian (e.g., daily office wear). Visual art allows far more freedom for unique expressions

[appughar @appughar]

Personally I am ambivalent of my feed being full of studio ghibli inspired pictures, But hope such tooling will be used to tell stories that can connect. At this moment, it looks like a fad though.

[appughar @appughar]

What fascinates me about Miyazaki’s work is how much empathy it stirs. In Totoro, when Satsuki runs searching for her sister—worried, confused, a bit angry—it felt like I was sharing the experience with her. That, to me, is the core aesthetics behind the shot.

[appughar @appughar]

That said, with reasoning models and looping through it, it is not unthinkable ( imho quite plausible ) machine intelligence will have self awareness and a sense of empathy derived from their past. What is not clear is how these sense diverges from our experiences.

[mobicham @mobicham]

GemLite now supports bfloat16 on both Hopper and non-Hopper gpus 🫡
https://t.co/qLOSDoryVp

[appughar @appughar]

@hssandhawalia @LeRobotHF @huggingface Bravo! 👏

[appughar @appughar]

@Thom_Wolf Thanks for writing this! W.r.t. benchmarks, shouldn’t we start attacking unsolved problems across disciplines as targets? From Hilbert’s problems to applied challenges in areas like fusion and cancer research—these are, by definition, out-of-distribution samples.

[appughar @appughar]

@ramz Which, in turn, originated from Tamil (nāraṅkāy), I think. https://t.co/gPsluxsjrn , https://t.co/AFxZvvgmMT

[appughar @appughar]

Gemlite and HQQ comes to @vllm_project 🤘

[Mobius Labs @Mobius_Labs]

New releases for hqq ( https://t.co/fHvaC4RGxb…) and gemlite ( https://t.co/KW2KscaaKj… )!

The most exciting update is that we're bringing support to @vllm_project

[appughar @appughar]

My 2 cents on R1 https://t.co/ZCD2Vf29fb: Open source and innovation unlocks opportunities for European startups. Speed & innovation now matter most. Deploying models for a few 100€s was once unimaginable. Now we can. Eg: https://t.co/qmmBbQAiwy.

https://t.co/MA2kWNFYJ3