Lucas

ml research, ex-quant 2x, clanker inference @hf0

[Lucas @quantbagel]

Robot action models shouldn't need 256 vision tokens per frame.

Pi0.5 spends 400M parameters on SigLIP just to see. We replaced it with a 4.4M encoder that
outputs 5 tokens — and action quality barely changes.

91x smaller. 51x fewer tokens. 7.3x faster inference.

[Arnaud Denis-Remillard @dr8_unix]

@sean_pixel @interlatent @allen_ai We got the same model fully optimized on our inference engine. Running at less than 95ms full e2e with MolmoAct2-Think. Running anywhere in the US.

[Lucas @quantbagel]

On a china trip rn, feeling like the fear of the sz robots is somewhat overstated

[Lucas @quantbagel]

We just labelled 1 million hours of visual data with a VLM at 40% the cost of other open source. If any data factory shops are interested we can get started fast

[David Liu @davidliuxyz]

why hasn't anyone made these autonomous robots yet

[hardmaru @hardmaru]

For over a decade, we’ve accepted that end-to-end backprop is the only way to train deep networks. But holding the entire network in memory all at once is why AI training is hitting a resource wall.

We found a new way to break the network into blocks and train them independently. The trick? Treating the network’s forward pass like a diffusion model denoising a signal.

This reinterpretation slashes the memory needed to train deep models. In our #ICLR2026 paper (https://t.co/PK5h0mqQSo), we matched end-to-end performance across ViTs, DiTs, and LLMs. We did this while training just one isolated block at a time.

[PsudoMike 🇨🇦 @PsudoMike]

Wealthsimple just launched a USD chequing account that works on both sides of the border. One account, two currencies, no conversion racket. Canadian banks had years to build this. They built fees instead. Now Wealthsimple has it. Good luck.

[Lucas @quantbagel]

Immigration is all you need

[Lucas @quantbagel]

If you aren’t fighting for control over the lightcone ngmi

[Lucas @quantbagel]

Landed in Canada and I haven’t had peace from drake in 12 hours

[Lucas @quantbagel]

Coolest thing I have ever been a part of

Literally Neil Armstrong for the Noetic Dyson sphere

[Prophetic @PropheticAI]

Meet @quantbagel.

We technologically altered his perception and recall of dreams.

He was a participant in our placebo/sham controlled study.

User interview #3

[Jason Gao @jasonzgao]

becoming a scholar was undoubtedly the best thing that happened to me in college

if you've ever had the desire to venture into startups, you'd be doing yourself a disservice by not applying :)

[Han Guo @HanGuo97]

LLM training is built on fast MatMuls. But many surrounding ops still run as memory-bound kernels.

CODA reparameterizes them to hide in the matmul’s shadow, fused into its epilogue before results leave the chip.

Bonus: LLMs can write fast CODA kernels too (approaching SoLs).

[Lucas @quantbagel]

One good girl is worth thousands of Anthropic secondaries

[Simon Eskildsen @Sirupsen]

turbopuffer crossed $100M run-rate in March. 19mo after $1M. Profitable & <$1M raised.

Cursor・Anthropic・Notion・Cognition・Harvey・Bridgewater・Ramp・Linear・Legora・Superhuman・Atlassian・Granola

We’d be nowhere without them. We work like hell to exceed their expectations.

[Daniel @danielgothits]

these types of properties are gonna look insanely underpriced in a few years when everything is all about drone delivery, EVTOLs, autonomous airport pickup and dropoff, starlink, robotic/AI farm equipment and civil unrest keeps rising in cities

[Lucas @quantbagel]

if any upcoming spaceX ballers wanna invest in what oai was originally supposed to be (robots), i’m on some pretty wild sht right now

[Lucas @quantbagel]

Reflex is actively hiring world-class engineers/kernelDevs for robot inference, even if you have zero prior experience in physical AI. Smart humans figure it out fast.

Please send ~3 bullet points demonstrating evidence of exceptional ability

[Lucas @quantbagel]

Reflex is actively hiring world-class engineers/kernelDevs for robot inference, even if you have zero prior experience in physical AI. Smart humans figure it out fast.

Please send ~3 bullet points demonstrating evidence of exceptional ability