Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

PerfectBit (PerfectBit.ai) creates high-quality training data that's correct by construction. They verify against physics simulators, scientific databases, formal proof systems. LLMs, robotics, AI for Science, and more. Congrats on the launch, Peter Vajda and Seiji Yamamoto!

Y Combinator

1,652,524 subscribers

12,495 views • 4 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

15 Comments

Ekta4 months ago

@PerfectbitAI @vaphab @sjy82155065 I love the demo videos you all have. Do you like professionally do launch video production?

Aditya || GTM guy4 months ago

@PerfectbitAI @vaphab @sjy82155065 the formal proof verification angle is the part that actually matters. LLMs generating physics data without constraint checking is how you get confidently wrong training sets

Felix Manufaktur4 months ago

@PerfectbitAI @vaphab @sjy82155065 Coffee ☕️ and this. AI training data verified against physics simulators, not web scrapes. We need this rigor for additive manufacturing and robotics.

Sridhar A1 month ago

@PerfectbitAI @vaphab @sjy82155065 training data that's verified against real physics and proof systems, not just scraped, is exactly what serious ai needs. congrats on the launch

Marcus4 months ago

@PerfectbitAI @vaphab @sjy82155065 how is the training data generated to begin with?

Parth Maheshwari4 months ago

@PerfectbitAI @vaphab @sjy82155065 🚀🚀

Gabriel Jarrosson4 months ago

@PerfectbitAI @vaphab @sjy82155065 Congrats! 🚀

Afroz Mohiuddin4 months ago

@PerfectbitAI @vaphab @sjy82155065 Super cool! Congratulations folks!

Arun jishnu4 months ago

@PerfectbitAI @vaphab @sjy82155065 Nice ad.

Julien Active4 months ago

@PerfectbitAI @vaphab @sjy82155065 Who is building verified training datasets for mobile-first vision models? This gets close. Please take my money.

Arseniy Shishaev (YC P26)4 months ago

@PerfectbitAI @vaphab @sjy82155065 Awesome video and product, congrats on the launch!!

Yanjun Chen4 months ago

@PerfectbitAI @vaphab @sjy82155065 Love the demo video and how it accurately pictures how it feels like to handle messy uncleaned data! Congratulations on the launch!

JP.AI4 months ago

@PerfectbitAI @vaphab @sjy82155065 Congrats👍 10 .ai 🪲PERFECT 10 FROM BINARY ORIGINS🪲

FlutterFlow4 months ago

@PerfectbitAI @vaphab @sjy82155065 congrats <3

motifyra4 months ago

@PerfectbitAI @vaphab @sjy82155065 Love the demo

Related Videos

Liva AI (@thelivaAI) provides high-quality human voice and video data for companies building realistic AI models. Congrats on the launch, aoi and Ashley!

Liva AI (@thelivaAI) provides high-quality human voice and video data for companies building realistic AI models. Congrats on the launch, aoi and Ashley!

Y Combinator

43,106 views • 1 year ago

Fixpoint (fixpoint) automates hiring vetted experts—doctors, lawyers, PhDs, coders, etc—for AI training data and evals. They handle sourcing and vetting so customers can focus on creating quality data and benchmarking AI. Congrats on the launch, Dylan Mikus and Jakub!

Fixpoint (fixpoint) automates hiring vetted experts—doctors, lawyers, PhDs, coders, etc—for AI training data and evals. They handle sourcing and vetting so customers can focus on creating quality data and benchmarking AI. Congrats on the launch, Dylan Mikus and Jakub!

Y Combinator

12,073 views • 11 months ago

Origami Robotics is building high-DOF robotic hands with in-joint motors and a co-designed data-collection glove to eliminate the embodiment gap by collecting high-quality, real-world data at scale. Congrats on the launch, Quanting Xie and Ryan Xie!

Origami Robotics is building high-DOF robotic hands with in-joint motors and a co-designed data-collection glove to eliminate the embodiment gap by collecting high-quality, real-world data at scale. Congrats on the launch, Quanting Xie and Ryan Xie!

Y Combinator

89,704 views • 6 months ago

.Cajal is massively scaling formal verification to accelerate scientific discovery. They deploy superhuman AI mathematicians to high-impact applied domains, starting with quantum computing and finance. Congrats on the launch, Luke & Pedro Nobre!

.Cajal is massively scaling formal verification to accelerate scientific discovery. They deploy superhuman AI mathematicians to high-impact applied domains, starting with quantum computing and finance. Congrats on the launch, Luke & Pedro Nobre!

Y Combinator

31,136 views • 7 months ago

Today we're launching Hebbian Robotics (YC S26) Hebbian Robotics . We build APIs for evaluating data quality, without training a robotics model. Data vendors use our API to search and analyze Physical AI data at scale, getting on-demand quality signals and metrics without managing infrastructure. High-quality data is the bottleneck to the next frontier of intelligence in robotics. Buyers want data they can trust. Data providers are focused on operational excellence. They don't have the bandwidth to understand the quality of their data at scale. We're fixing this. Vendors use our APIs to understand the impact of their decisions on their data quality: - Buyers demanding per-collector hour limits? Prove compliance while we stay ahead of the workarounds. - Changed your collection SOPs? Monitor for quality drift and regression. - Suspect duplicate samples? Verify with a single API call. Before Hebbian, I trained AI models for industrial cleaning robots, and Kingston Kuan spent years building large-scale infrastructure at Jane Street.

Today we're launching Hebbian Robotics (YC S26) Hebbian Robotics . We build APIs for evaluating data quality, without training a robotics model. Data vendors use our API to search and analyze Physical AI data at scale, getting on-demand quality signals and metrics without managing infrastructure. High-quality data is the bottleneck to the next frontier of intelligence in robotics. Buyers want data they can trust. Data providers are focused on operational excellence. They don't have the bandwidth to understand the quality of their data at scale. We're fixing this. Vendors use our APIs to understand the impact of their decisions on their data quality: - Buyers demanding per-collector hour limits? Prove compliance while we stay ahead of the workarounds. - Changed your collection SOPs? Monitor for quality drift and regression. - Suspect duplicate samples? Verify with a single API call. Before Hebbian, I trained AI models for industrial cleaning robots, and Kingston Kuan spent years building large-scale infrastructure at Jane Street.

Brandon Ong

166,669 views • 1 month ago

Traverse (Traverse (YC W26)) is a data research lab solving the hardest problem in AI training: subjective, taste-dependent work. They generate long-horizon training data for frontier models by observing and capturing what specialists actually do on the job. Congrats on the launch, Lance Yan and Zachary Yu!

Traverse (Traverse (YC W26)) is a data research lab solving the hardest problem in AI training: subjective, taste-dependent work. They generate long-horizon training data for frontier models by observing and capturing what specialists actually do on the job. Congrats on the launch, Lance Yan and Zachary Yu!

Y Combinator

78,149 views • 6 months ago

LLMs are great for human in the loop applications, but fail at deterministic developer tasks. Interfaze (YC P26) is a new AI model that outperforms general LLMs on high accuracy tasks like: OCR, Object Detection, Web scraping, Speech-to-text, Classification and more. Congrats on the launch, Yoeven and Harsha!

LLMs are great for human in the loop applications, but fail at deterministic developer tasks. Interfaze (YC P26) is a new AI model that outperforms general LLMs on high accuracy tasks like: OCR, Object Detection, Web scraping, Speech-to-text, Classification and more. Congrats on the launch, Yoeven and Harsha!

Y Combinator

71,358 views • 5 months ago

“Math is code. And code is math.” Carina Hong CEO Axiom on why that idea matters for AI: “For centuries, mathematicians reasoned in natural language. But because of the Curry–Howard correspondence, a proof can also be a program.” “You can translate natural-language math into formal code using systems like Lean. Lean is both a theorem-proving language and a programming language.” “That means you can generate solutions as code and then formally verify them. You can even use the same system to verify programs themselves.” “This creates a flywheel of generation and verification, where AI can both write solutions and prove that they’re correct.”

“Math is code. And code is math.” Carina Hong CEO Axiom on why that idea matters for AI: “For centuries, mathematicians reasoned in natural language. But because of the Curry–Howard correspondence, a proof can also be a program.” “You can translate natural-language math into formal code using systems like Lean. Lean is both a theorem-proving language and a programming language.” “That means you can generate solutions as code and then formally verify them. You can even use the same system to verify programs themselves.” “This creates a flywheel of generation and verification, where AI can both write solutions and prove that they’re correct.”

Forward Future

24,729 views • 7 months ago

Palace (Palace (YC S25)) is a unified platform for investment monitoring. Their platform helps private market investors save time and act on portfolio company updates more effectively by better organizing the reporting data they collect. Congrats on the launch, leeds and Derek Ye!

Palace (Palace (YC S25)) is a unified platform for investment monitoring. Their platform helps private market investors save time and act on portfolio company updates more effectively by better organizing the reporting data they collect. Congrats on the launch, leeds and Derek Ye!

Y Combinator

19,608 views • 1 year ago

Could You Guess What Comes After the OptimAI Edge Node? We’re excited to share another glimpse into the future, Community-Powered Data Validation, a unique feature within the OptimAI Core Node for desktop. What is Data Validation? It’s the essential process of ensuring AI systems learn from accurate, unbiased, and high-quality data. Through human intelligence, users verify, correct, and refine information - making it reliable for training advanced AI agents. This process is enabled by the OptimAI DeHIN (Decentralized Human Intelligence Network), which brings collective intelligence into the heart of AI development. Why is it important? Because truly responsible and powerful AI starts with trusted data. Community validation improves fairness, transparency, and performance - ensuring AI benefits everyone. And as a participant, you’re not just supporting better AI; you’re earning rewards and shaping its future. Your desktop is about to become much more than a device. It’s a gateway to reinforcing the quality of data that powers Agentic AI. Validate data. Strengthen AI. Get rewarded. The future of AI is collective — and it needs you. #BUIDL with us! 👉OptimAI Lite Node: 👉OptimAI Edge Node:

Could You Guess What Comes After the OptimAI Edge Node? We’re excited to share another glimpse into the future, Community-Powered Data Validation, a unique feature within the OptimAI Core Node for desktop. What is Data Validation? It’s the essential process of ensuring AI systems learn from accurate, unbiased, and high-quality data. Through human intelligence, users verify, correct, and refine information - making it reliable for training advanced AI agents. This process is enabled by the OptimAI DeHIN (Decentralized Human Intelligence Network), which brings collective intelligence into the heart of AI development. Why is it important? Because truly responsible and powerful AI starts with trusted data. Community validation improves fairness, transparency, and performance - ensuring AI benefits everyone. And as a participant, you’re not just supporting better AI; you’re earning rewards and shaping its future. Your desktop is about to become much more than a device. It’s a gateway to reinforcing the quality of data that powers Agentic AI. Validate data. Strengthen AI. Get rewarded. The future of AI is collective — and it needs you. #BUIDL with us! 👉OptimAI Lite Node: 👉OptimAI Edge Node:

OptimAI Network

60,810 views • 1 year ago

AI agents are the web's biggest users. They run on TinyFish, so we see what they reach for: data that's licensed, not published. Today, with crunchbase, Similarweb, and 13 more companies, we're launching the TinyFish Data Partners Alliance. The best data, curated by category, for agents.

AI agents are the web's biggest users. They run on TinyFish, so we see what they reach for: data that's licensed, not published. Today, with crunchbase, Similarweb, and 13 more companies, we're launching the TinyFish Data Partners Alliance. The best data, curated by category, for agents.

TinyFish

330,273 views • 2 days ago

Been looking into RICE AI ( ◉ - ◉ ) lately, and it’s one of the few projects actually tackling a real challenge in robotics and AGI: the lack of high-quality training data. Robots can’t get truly intelligent without it. They building a decentralized way to collect and share robotics data, using a DePIN network to train physical AI foundation models. The project was even hand-picked by CZ as a winner of the Dubai BNB Chain Demo Day, and it’s backed by Floki while also being part of the NVIDIA Inception Program. They also launched FLOKI Minibot M1 - a small AI companion that doubles as a data node. Every interaction with it feeds the RICE AI Foundry, the first decentralized platform for Robotics Foundation Models, and contributors earn $RICE tokens in return. And it’s working: – 800 minibots sold in 24h – 33,000% oversubscribed Binance IDO – 500+ robots already deployed worldwide Feels like a project with real long-term potential. Worth keeping an eye on. RICE AI ( ◉ - ◉ )

Been looking into RICE AI ( ◉ - ◉ ) lately, and it’s one of the few projects actually tackling a real challenge in robotics and AGI: the lack of high-quality training data. Robots can’t get truly intelligent without it. They building a decentralized way to collect and share robotics data, using a DePIN network to train physical AI foundation models. The project was even hand-picked by CZ as a winner of the Dubai BNB Chain Demo Day, and it’s backed by Floki while also being part of the NVIDIA Inception Program. They also launched FLOKI Minibot M1 - a small AI companion that doubles as a data node. Every interaction with it feeds the RICE AI Foundry, the first decentralized platform for Robotics Foundation Models, and contributors earn $RICE tokens in return. And it’s working: – 800 minibots sold in 24h – 33,000% oversubscribed Binance IDO – 500+ robots already deployed worldwide Feels like a project with real long-term potential. Worth keeping an eye on. RICE AI ( ◉ - ◉ )

Christiaan

12,696 views • 1 year ago

BOND is the AI Chief of Staff for CEOs and busy execs. They centralize scattered data from Slack, Jira, Notion, and more into one place— and ping you in real-time on blockers and wins. Cut meetings, save 10+hrs/week, and move 1000x faster. Congrats on the launch, Chloe Samaha & René Sultan!

BOND is the AI Chief of Staff for CEOs and busy execs. They centralize scattered data from Slack, Jira, Notion, and more into one place— and ping you in real-time on blockers and wins. Cut meetings, save 10+hrs/week, and move 1000x faster. Congrats on the launch, Chloe Samaha & René Sultan!

Y Combinator

278,469 views • 1 year ago

What is the best data for training humanoid & robotics foundation models? Pete Florence Pete Florence (CEO Eric Nehrlich, ex-Google DeepMind) dropped his live data tier list in this 7-minute clip on TBPN: - S-tier: Real-world robot experience (especially glove/sensor high-dexterity data) - A/B-tier: Internet/YouTube videos. Surprisingly powerful for transfer learning (the “web data” moment for physical AI) - B-tier: Text/common crawl (Reddit, books, etc.). Useful priors, but not enough alone - C-tier: Motion Capture. Great for whole-body motion, weak on finger dexterity - C or lower: Simulation / synthetic / world models. High potential, still waiting for strong real-world proof Generalist has collected 270,000+ hours of real-world manipulation data (scaling ~10k hours/week). And Pete stressed one key point: “The quality of data is incredibly important.” It’s not just about volume. It’s developing intuition for what actually drives performance through hands-on work. As Physical AI scales, curated real-world and high-quality internet video looks like a winning combo. h/t YUJI | AI Robotics

What is the best data for training humanoid & robotics foundation models? Pete Florence Pete Florence (CEO Eric Nehrlich, ex-Google DeepMind) dropped his live data tier list in this 7-minute clip on TBPN: - S-tier: Real-world robot experience (especially glove/sensor high-dexterity data) - A/B-tier: Internet/YouTube videos. Surprisingly powerful for transfer learning (the “web data” moment for physical AI) - B-tier: Text/common crawl (Reddit, books, etc.). Useful priors, but not enough alone - C-tier: Motion Capture. Great for whole-body motion, weak on finger dexterity - C or lower: Simulation / synthetic / world models. High potential, still waiting for strong real-world proof Generalist has collected 270,000+ hours of real-world manipulation data (scaling ~10k hours/week). And Pete stressed one key point: “The quality of data is incredibly important.” It’s not just about volume. It’s developing intuition for what actually drives performance through hands-on work. As Physical AI scales, curated real-world and high-quality internet video looks like a winning combo. h/t YUJI | AI Robotics

Michelle Sun

15,351 views • 3 months ago

Robotics is still data starved. Collecting high-quality robot demonstrations remains brutally slow and expensive. Introducing COBALT: A cloud-native teleoperation platform designed for large-scale robot learning. We are democratizing data collection by leveraging the hardware everyone already owns: the smartphone All you need is to download an app (today)! Read on for more!

Robotics is still data starved. Collecting high-quality robot demonstrations remains brutally slow and expensive. Introducing COBALT: A cloud-native teleoperation platform designed for large-scale robot learning. We are democratizing data collection by leveraging the hardware everyone already owns: the smartphone All you need is to download an app (today)! Read on for more!

Animesh Garg

104,553 views • 3 months ago

Scientific progress relies on empirical evidence, rigorous testing, and peer review, none of which are advanced through public debates with untrained individuals. Unlike politics or philosophy, where personal opinions and rhetoric can shape policy, science is governed by objective data and reproducible facts that remain true regardless of public consensus. Engaging in formal debates with non-experts creates a false equivalence, giving the misleading impression that a robustly tested scientific consensus is merely one side of an ongoing opinion piece. Furthermore, scientific communication requires a specialized understanding of methodology and data analysis; arguing with untrained individuals often forces experts to defend fundamental facts against emotional appeals or misinformation rather than advancing actual knowledge. Ultimately, science validates its claims through the laboratory and the field, making public debate an inefficient and counterproductive forum for establishing objective truth. Watch the full video:

Scientific progress relies on empirical evidence, rigorous testing, and peer review, none of which are advanced through public debates with untrained individuals. Unlike politics or philosophy, where personal opinions and rhetoric can shape policy, science is governed by objective data and reproducible facts that remain true regardless of public consensus. Engaging in formal debates with non-experts creates a false equivalence, giving the misleading impression that a robustly tested scientific consensus is merely one side of an ongoing opinion piece. Furthermore, scientific communication requires a specialized understanding of methodology and data analysis; arguing with untrained individuals often forces experts to defend fundamental facts against emotional appeals or misinformation rather than advancing actual knowledge. Ultimately, science validates its claims through the laboratory and the field, making public debate an inefficient and counterproductive forum for establishing objective truth. Watch the full video:

TheLiverDoc™

146,190 views • 4 months ago

.Polymath is training world generation models to automate the creation of RL environments. Traditionally, RL environment generation has been bottlenecked by human data. Superintelligence will never be achieved by human data alone. Polymath is building the core technology to enable automated environment generation using far less human effort than traditionally required, and eventually none. This allows for more complex and realistic worlds, and higher quality, scale, and diversity of tasks. This will be essential to unlock RL scaling. The end goal is to create large-scale, long-horizon environments from a text description alone. This will enable the creation of worlds of arbitrary complexity and scale, which is foundational for training & evaluating autonomous, superintelligent AI agents. Congrats on the launch, Dylan Ma and Naren Yenuganti!

.Polymath is training world generation models to automate the creation of RL environments. Traditionally, RL environment generation has been bottlenecked by human data. Superintelligence will never be achieved by human data alone. Polymath is building the core technology to enable automated environment generation using far less human effort than traditionally required, and eventually none. This allows for more complex and realistic worlds, and higher quality, scale, and diversity of tasks. This will be essential to unlock RL scaling. The end goal is to create large-scale, long-horizon environments from a text description alone. This will enable the creation of worlds of arbitrary complexity and scale, which is foundational for training & evaluating autonomous, superintelligent AI agents. Congrats on the launch, Dylan Ma and Naren Yenuganti!

Y Combinator

44,192 views • 7 months ago

Robotics has a data problem, and egocentric data is a compelling way to solve it because egocentric data is (1) scalable — it’s cheap to collect — and (2) easily captures the breadth and diversity of real human tasks, while (3) being true to real physics. Lightwheel recently released a hundred thousand hours of high-quality egocentric video data, spread across a wide range of human tasks from cleaning and tidying to assembly and construction to retail tasks. Here to tell us about this tremendous effort is Steve Xie, CEO and co-founder of Lightwheel. The data is also available on HuggingFace right now. Watch Episode 102 of RoboPapers, with Michael Cho - Rbt/Acc and Jiafei Duan, to learn more!

Robotics has a data problem, and egocentric data is a compelling way to solve it because egocentric data is (1) scalable — it’s cheap to collect — and (2) easily captures the breadth and diversity of real human tasks, while (3) being true to real physics. Lightwheel recently released a hundred thousand hours of high-quality egocentric video data, spread across a wide range of human tasks from cleaning and tidying to assembly and construction to retail tasks. Here to tell us about this tremendous effort is Steve Xie, CEO and co-founder of Lightwheel. The data is also available on HuggingFace right now. Watch Episode 102 of RoboPapers, with Michael Cho - Rbt/Acc and Jiafei Duan, to learn more!

RoboPapers

33,470 views • 18 days ago