I am super excited to share a new AI tool, Refine. Refine thoroughly studies research papers like a referee and finds issues with correctness, clarity, and consistency. In my own papers, it regularly catches problems that my coauthors and I missed. 1/

The model is the network Fortytwo’s Swarm Inference is a new AI architecture built on a distributed network of specialized models For every query: 1. Relevant nodes generate individual responses 2. Those same nodes evaluate and rank each other’s answers 3. The highest-ranked

Building systems for 1 hour a day completely rewires your company.

The interface of the future is human. We’ve raised a $40M Series B from CRV, Scale, Sequoia, and YC to teach machines the art of being human, so that using a computer feels like talking to a friend or a coworker. And today, I’m excited for y’all to meet the PALs: a new

Finally unveiling Pickle 1, our AI-powered AR Glass. We’ve been getting a few recurring questions from investors and people following our journey, so here’s how I think about them: 1. “Why AR Glass? Aren’t you building a memory system?” We believe AR Glass is inevitable to

I’m Leo, founder of Dealism Introducing Dealism - World’s First Vibe Selling AI Agent We believe everyone should be able to sell with confidence and ease. It learns your style, applies proven sales strategies, and helps you turn every conversation into a winning moment.

Banger paper. Finally getting to read the rest of it 😅 There's an award for whoever finds all the secrets first in the new arxiv version that's coming soon.

Kyle🤖🚀🦭 It would have to be massively better than torch for me to use it as a full replacement. So far, what's there is better, but it's missing some stuff and the overall perf w/ compile is worse than torch eager. But I think I can use its kernel gen to output custom fused ops and bind

Led a stakeholder session on this. Case in point, with the advent and growth of frontier ai models, efficiency is only seen through the lens of what is considered “unique” output by the observer but even with a human touch, there is sometimes lack of emotional , sentimental

I love the advent of AI from the beginning to now and beyond. I fear what it can take from humanity.

VC’s should be looking to fund Reinforced learning Environments for the next phase of growth. Off the back of annotating data for a research lab, i think frontier ai models have grown beyond training based on human instructions and relentless oversight. With what Gemini 3.0 and

With what i have just witnessed this morning, the bar to entry is not going to tend to zero in the next 12months, it’s literally going to be zero. And it’s only the beginning.

God buries the gold in the filth and this is not cruelty it is selection. He asks - how badly do you want it. badly enough to kneel in what others step over. badly enough to touch what the clean men will not name. and the poor stay poor because they want clean gold, gold that

This is so structurally engineered!

This can be highly debated but maybe, we might be approaching transformer models from the angle of building to infinity while we are most evidently feeding in limits. Maybe AGI won’t come from just more GPUs but from finally reading nature’s blueprint for how to grow a mind.

We do not have enough structured DATA ! And yet we are at the point of zero to one to 1 and 1 equals 1. Imagine when the need for the desired input of one tends to zero! Really, the world has been changing right in front of us and currently , it’s at an accelerated pace.

Google is training llm models at scale!

I don’t think folks understand that local/on-prem servers,GPU runs, most times than never get more expensive than cloud instances. Hybrid solves this. And helps your GPU’s(particularly focused on this) live longer.

Genetic mapping at the speed of thought! Wild, just wild. This could also help discover species beyond the data bank we have on existential life forms.

The desire for Network Engineers will be profound!