account known to cause systemic risk
Cape Town
Joined August 2010
- Tweets1K
- Following5.7K
- Followers621
- Likes5.2K
liquidsnake retweeted
Introducing Sterling: Inference from verified providers.
The model you’re paying for might not be the one answering you. Every API model is served by a provider. Some serve a smaller version. Some break tool calls.
Sterling tests all providers and routes to the verified ones.
liquidsnake retweeted
We post-trained @physical_int's π0.5 on our ego data and found it outperformed Apple’s EgoDex on LIBERO-10.
Equal amounts of data were sampled from both datasets with matched tasks, environments, and objects.
Our results suggest that viewpoint diversity and richer action supervision can overcome a fundamental limitation of robot learning: dependence on fixed camera positions.
liquidsnake retweeted
Sortex: our data quality platform built to search and filter the world's largest egocentric dataset.
Across 50+ environments and 20k+ tasks, it uses an in-house model and evaluation harness to score each clip's usability, selecting for specific training objectives by:
→ Narrating tasks and steps
→ Checking hand and action visibility
→ Comparing against the rest of the corpus
We also tested Sortex on popular open datasets and found they differ widely in quality and coverage.
liquidsnake retweeted
Everyone has a story of missed investment riches.
Agents promise to fix that. After building systems across prediction markets, crypto, perps and stocks, we think most won’t... yet.
"10 hard problems in agentic investing" (written w/ @SpinFN)
Thread + bitkraft.vc/insights/10-hard…
liquidsnake retweeted
sim2real gap is one of the biggest bottlenecks in robotics. we trained akari in-house, the first of the major model breakthroughs that will close it
turns out a data moat and a cracked research team can compound pretty fast
Introducing Akari: a frontier learned rendering model for robotics simulation.
Akari generates photorealistic rendering for any physics engine to close the sim-to-real gap.
Trained in-house, it is the fastest and highest-fidelity learned renderer in any sim platform: 0.046 LPIPS at 0.5s/frame, consistent across 1K+ frames.
liquidsnake retweeted
Introducing Akari: a frontier learned rendering model for robotics simulation.
Akari generates photorealistic rendering for any physics engine to close the sim-to-real gap.
Trained in-house, it is the fastest and highest-fidelity learned renderer in any sim platform: 0.046 LPIPS at 0.5s/frame, consistent across 1K+ frames.
liquidsnake retweeted
There are very few teams out there with the resources and research talent to tackle learned sim. We are adamant that solving the sim2real gap will require neural systems to model the increasingly complex laws of the universe— and Akari is the first step towards that bet.
Learned sim is the way. Introducing Akari.
Introducing Akari: a frontier learned rendering model for robotics simulation.
Akari generates photorealistic rendering for any physics engine to close the sim-to-real gap.
Trained in-house, it is the fastest and highest-fidelity learned renderer in any sim platform: 0.046 LPIPS at 0.5s/frame, consistent across 1K+ frames.
liquidsnake retweeted
Announcing the Midcentury Research Residency.
We're bringing together a small cohort of researchers to tackle the hardest problems in physical AI, backed by:
→ $1 million in compute
→ The world's largest robotics dataset spanning 2M+ hours
→ SOTA action supervision infrastructure
Our first cohort includes an author of NVIDIA's Cosmos 3, a CoRL award winner, and researchers from CMU, UCL, and Brown, with publications at NeurIPS and ICRA.
Their research focuses on fundamental questions in robot learning and world models. We've already published MC-PhysBench, with more work coming soon.
Applications for Cohort 2 are now open. Apply here: midcentury.xyz/residency
liquidsnake retweeted
some serious talent density on the research team
- co-author of NVIDIA's Cosmos 3
- multiple papers at NeurIPS, ICRA, CoRL, best paper awards, etc.
- core contributors to Stanford's RoboNet and OpenAI's GDPVal
been ahead of the curve on human-to-robot transfer, WAMs, and RL in sim. their early work here will be consensus soon
Announcing the Midcentury Research Residency.
We're bringing together a small cohort of researchers to tackle the hardest problems in physical AI, backed by:
→ $1 million in compute
→ The world's largest robotics dataset spanning 2M+ hours
→ SOTA action supervision infrastructure
Our first cohort includes an author of NVIDIA's Cosmos 3, a CoRL award winner, and researchers from CMU, UCL, and Brown, with publications at NeurIPS and ICRA.
Their research focuses on fundamental questions in robot learning and world models. We've already published MC-PhysBench, with more work coming soon.
Applications for Cohort 2 are now open. Apply here: midcentury.xyz/residency
liquidsnake retweeted
We tested leading video world models on their ability to predict simple physics and found that most fail within a few frames.
Today we're releasing MC-PhysBench, a new kind of benchmark that measures how long a model's output stays physically valid.
15+ models. 9 scenes. 6,000+ videos scored. 7 failure checks.
liquidsnake retweeted
we're so early on world models as simulators
we scored 17 video world models on a ball rolling down a ramp and every one breaks physics in under a second. a no-model baseline beats them in 91 of 103 results
training on pixels alone only gets you so far. modeling physics accurately takes physics-rich data
We tested leading video world models on their ability to predict simple physics and found that most fail within a few frames.
Today we're releasing MC-PhysBench, a new kind of benchmark that measures how long a model's output stays physically valid.
15+ models. 9 scenes. 6,000+ videos scored. 7 failure checks.
liquidsnake retweeted
Introducing MC-EgoHands, the first all-in-one harness for turning raw ego video into training-ready data.
→ Frontier hand pose (<7mm) across 4 public benchmarks
→ Human grasps retargeted to Panda/LIBERO and Yam in one pass
→ Occluded frames flagged as interpolated or missing
→ Structured episodes in @huggingface @LeRobotHF v3
liquidsnake retweeted
How do you scale 2 million hours of ego data without sacrificing quality?
Earlier this week, we revealed the world’s largest egocentric dataset. Our frontier pipeline has also turned it into the most labeled, annotated, and human QA’ed robotics ego dataset in the world.
We’ve focused on some of the hardest problems in computer vision and ego data to ensure every clip contains SOTA action supervision directly helpful for training dexterous manipulation.
liquidsnake retweeted
We at @Delphi_Ventures could not be more excited to back @dgmonsoon and @ckxyz_ of @MidcenturyAI
After first meeting David and Chetan over a call and being impressed, my partner @ZeMariaMacedo and I flew to New York City to spend time with them in person. We were blown away by their clock speed, clarity of thought and just how obsessed they were with what they wanted to build. As an investor you just know when founders are impressive and David and Chetan blew us away early on
Chetan and David came out of Stanford’s AI ecosystem, working across labs led by Fei-Fei Li, Andrew Ng, Michael Bernstein and others, including on early robotics efforts like RoboTurk and RoboNet. They later built AI products at Replit and Salesforce, then became repeat founders together through YC.
Today they’re announcing a $15M seed round and are already working with frontier labs across two products:
1/ Real world training data
A massive, densely labeled dataset of people actually doing physical work
- 2M+ hours of first-person data
- 50+ environments and 20,000+ tasks
- Hands visible 90% of the time - Manufacturing, construction, warehouses, agriculture and more
- Enriched with 3D hand/body motion, object tracking and detailed task labels The point isn’t just more video. It’s capturing what people’s hands are doing, which objects they’re interacting with and how a task gets completed. This is whats important to Robotics labs.
2/ Matrix: Their simulation platform combines classical simulation, learned physics and real world data.
Robotics teams can test across thousands of scenarios in parallel, replay failures and turn those failures into new training experience. You can’t efficiently run every experiment on a physical robot. Matrix is being built to scale that testing and learning The end game is getting robots out of controlled demos and into useful work. Robotics needs both the real world examples to learn from and somewhere to practice at scale.
David and Chetan are building both. If we want robots working reliably in factories, warehouses and eventually our homes, this infrastructure matters. That’s the ambition we’re backing.
We couldn't be more excited to back David and Chetan
Introducing Midcentury.
We’re building the data and simulation infra for physical AI.
Today, we’re coming out of stealth with a $15M Series Seed to scale robotics beyond polished demos.
We’re already supporting frontier labs with:
→ The world’s largest egocentric dataset: 2M+ hours, 50+ environments, 20,000+ tasks
→ Matrix: a frontier simulation platform scaled with our real-world data to evaluate and post-train policies at scale
liquidsnake retweeted
We're ushering the era of ego data commoditization. Scale and quality of training corpus is the only thing that matters. If you're not aligned, you will get left behind
Introducing Midcentury.
We’re building the data and simulation infra for physical AI.
Today, we’re coming out of stealth with a $15M Series Seed to scale robotics beyond polished demos.
We’re already supporting frontier labs with:
→ The world’s largest egocentric dataset: 2M+ hours, 50+ environments, 20,000+ tasks
→ Matrix: a frontier simulation platform scaled with our real-world data to evaluate and post-train policies at scale
liquidsnake retweeted
Introducing Midcentury.
We’re building the data and simulation infra for physical AI.
Today, we’re coming out of stealth with a $15M Series Seed to scale robotics beyond polished demos.
We’re already supporting frontier labs with:
→ The world’s largest egocentric dataset: 2M+ hours, 50+ environments, 20,000+ tasks
→ Matrix: a frontier simulation platform scaled with our real-world data to evaluate and post-train policies at scale
liquidsnake retweeted
The last major keyboard moment was in 2007,
when Apple put the keyboard on glass.
Nearly 20 years later, we're introducing:
Acti @openacti1, the Agentic Keyboard.
Not another AI keyboard that fixes grammar.
Not another voice keyboard that types faster.
An invisible agent in every text field.
It doesn't predict your next word.
It takes your next action. 👇
liquidsnake retweeted
I always dreamed to make proper hard sci-fi.
It been around my head for years - to bring life to new universe with predicting-the-future approach. And, of course, add something new to the genre as I did with Tarkov.
Now my life has become twice as difficult. I decided to rush in with the thought: what if more time passes and I no longer can or want to do anything new?
Frago is being developed by separate company that I created, full of great ppl around the globe.
Escape from Tarkov is my life and I'm continue working on it too with my BSG family no matter what.
Tarkov for life.
Frago is the new adventure that I want to share with you.
Thank you for your support and long live hardcore!
Welcome to Fragmentary Order.
Fragmentary Order is a hardcore sci-fi FPS set in 2251.
Warfare is systemized.
Your position in it is yours to build.
Visit fragmentaryorder.com for more info.
youtu.be/-HXp0KTNBm4?si=P7Fz…
