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.

At Applied Compute, we're not just training models, we're building a model factory. To help researchers scale, we built Ari, our in-house AI research agent. Ari diligently sifts through gigabytes of logs to find evidence of unhealthy behavior, remembers results the team has learned from past experiments, and produces... show more

Applied Compute

7,376 subscribers

41,519 views • 18 days ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

We're hiring for many roles at Runway. Beyond research/MLE roles, we have openings for art directors, creative developers, API engineers and more. Join our wonderful team to help us simulate the world.

We're hiring for many roles at Runway. Beyond research/MLE roles, we have openings for art directors, creative developers, API engineers and more. Join our wonderful team to help us simulate the world.

Jonathan Granskog

49,167 views • 7 months ago

Today we introduce Liquid Labs, our advanced research unit, with the goal of understanding and building efficient and adaptive intelligence systems. Liquid Labs consolidates our existing research efforts at Liquid across architecture of foundation models, multimodality, training, data, and inference. The lab also will be home to new frontier research work across the broad range of foundation model build-up stack. Read the full announcement: We are hiring: Also find us at NeurIPS 2025 exhibition hall! 🚀

Today we introduce Liquid Labs, our advanced research unit, with the goal of understanding and building efficient and adaptive intelligence systems. Liquid Labs consolidates our existing research efforts at Liquid across architecture of foundation models, multimodality, training, data, and inference. The lab also will be home to new frontier research work across the broad range of foundation model build-up stack. Read the full announcement: We are hiring: Also find us at NeurIPS 2025 exhibition hall! 🚀

Liquid AI

42,881 views • 9 months ago

70% of AI compute in 2026 is inference. Not training. The models are built. The real infrastructure challenge is serving them at scale to millions of users and agents, every day, in production. Most enterprise infrastructure is still optimized for the wrong workload.🧵

70% of AI compute in 2026 is inference. Not training. The models are built. The real infrastructure challenge is serving them at scale to millions of users and agents, every day, in production. Most enterprise infrastructure is still optimized for the wrong workload.🧵

Aethir

21,176 views • 2 months ago

The Proxima team is thrilled to be at ICLR 2026 in Rio! We're hiring Researchers and we'll be at ICLR to meet in person. If you're pushing the boundaries of generative AI x Bio, we want to talk. Apply here and we will be in touch to schedule:

The Proxima team is thrilled to be at ICLR 2026 in Rio! We're hiring Researchers and we'll be at ICLR to meet in person. If you're pushing the boundaries of generative AI x Bio, we want to talk. Apply here and we will be in touch to schedule:

Proxima

11,791 views • 4 months ago

MANUS AI: HYPE VS. REALITY 🔍 Yichao 'Peak' Ji (co-founder of ) confirmed rumors: ✅ Built on Anthropic Claude Sonnet, not their own foundation model ✅Has access to 29 tools and uses Browser Use open-source for browser control ✅User communicates with executor agent and not planner or other agents. ✅Each user gets isolated sandbox environment ✅Outperforms OpenAI Deep Research on GAIA benchmark Building AI products doesn't require training your own foundation models. We're probably just scratching the surface of what existing models can do with the right tooling and integration!

MANUS AI: HYPE VS. REALITY 🔍 Yichao 'Peak' Ji (co-founder of ) confirmed rumors: ✅ Built on Anthropic Claude Sonnet, not their own foundation model ✅Has access to 29 tools and uses Browser Use open-source for browser control ✅User communicates with executor agent and not planner or other agents. ✅Each user gets isolated sandbox environment ✅Outperforms OpenAI Deep Research on GAIA benchmark Building AI products doesn't require training your own foundation models. We're probably just scratching the surface of what existing models can do with the right tooling and integration!

Philipp Schmid

202,928 views • 1 year ago

We wanted to know if Viktor could own our outbound workflow, so we ran a one-week pilot. The result? An estimated 80.5 hours of manual prep replaced by 4.5 hours of team review. We're a digital marketing team managing outbound campaigns alongside client work, and one of the biggest bottlenecks has always been the prep. Researching prospects, writing personalized outreach, building follow-up sequences, and keeping everything organized takes far longer than most people realize. So instead of asking Viktor to help with one email, we asked him to own the workflow. During our pilot, Viktor handled prospect research, enrichment, personalized outreach, follow-up sequences, and campaign tracking for 100 target prospects. Our team's role was reduced to reviewing, approving, and managing replies before anything went out. What impressed us wasn't just the time savings. It was the consistency. Every prospect received the same level of research, the same quality of personalization, and a complete three-email sequence that was ready for review instead of being built manually from scratch. That's when we stopped thinking about AI as another writing tool. He's an AI employee that owns recurring work while keeping humans in control of the final decision. Hire Viktor for your team. $100 in credits included, no card. Full link in first comment. #AIemployee #OutboundSales Paid Partnership

We wanted to know if Viktor could own our outbound workflow, so we ran a one-week pilot. The result? An estimated 80.5 hours of manual prep replaced by 4.5 hours of team review. We're a digital marketing team managing outbound campaigns alongside client work, and one of the biggest bottlenecks has always been the prep. Researching prospects, writing personalized outreach, building follow-up sequences, and keeping everything organized takes far longer than most people realize. So instead of asking Viktor to help with one email, we asked him to own the workflow. During our pilot, Viktor handled prospect research, enrichment, personalized outreach, follow-up sequences, and campaign tracking for 100 target prospects. Our team's role was reduced to reviewing, approving, and managing replies before anything went out. What impressed us wasn't just the time savings. It was the consistency. Every prospect received the same level of research, the same quality of personalization, and a complete three-email sequence that was ready for review instead of being built manually from scratch. That's when we stopped thinking about AI as another writing tool. He's an AI employee that owns recurring work while keeping humans in control of the final decision. Hire Viktor for your team. $100 in credits included, no card. Full link in first comment. #AIemployee #OutboundSales Paid Partnership

Kylie

76,240 views • 1 month ago

1/10 From X, The Moonshot Factory to the cutting edge of robotics AI 🦾 We're excited to show you Luke Powell & Jody Hudson-Powell’s new identity for Intrinsic, the software company at the forefront of AI-enabled, developer-centric robotics. Shout out to Wendy Tan White and the Intrinsic team 🤘

1/10 From X, The Moonshot Factory to the cutting edge of robotics AI 🦾 We're excited to show you Luke Powell & Jody Hudson-Powell’s new identity for Intrinsic, the software company at the forefront of AI-enabled, developer-centric robotics. Shout out to Wendy Tan White and the Intrinsic team 🤘

Pentagram Design

13,264 views • 3 years ago

🎉Save the date!🎉 We're thrilled to announce that we're celebrating our 20th anniversary on May 24. Join us as we reflect on our achievements over the past two decades and explore our vision for the future of #research in #Africa Dr. Catherine Kyobutungi Muzukulu wa Bityo Evelyn Gitau #APHRC20 #WeAreAfrica

🎉Save the date!🎉 We're thrilled to announce that we're celebrating our 20th anniversary on May 24. Join us as we reflect on our achievements over the past two decades and explore our vision for the future of #research in #Africa Dr. Catherine Kyobutungi Muzukulu wa Bityo Evelyn Gitau #APHRC20 #WeAreAfrica

APHRC

23,398 views • 3 years ago

GROK DOMINATES AI AT SCALE, REDEFINING WHAT’S POSSIBLE IN REAL TIME Most AI models stall at the prototype phase. Grok advances beyond limits where others fail to perform. •⁠ ⁠Built for real-time decisioning across vast data landscapes •⁠ ⁠Optimized to process, synthesize, and respond with human-like fluidity •⁠ ⁠Deploys intelligence at the edge, not just in the cloud •⁠ ⁠Powers robotics, research, and infrastructure in live environments •⁠ ⁠Infrastructure built to evolve, not collapse, under scale Grok delivers high-performance AI built for deployment and real-world use. Source: xAI, Grok, Elon Musk

GROK DOMINATES AI AT SCALE, REDEFINING WHAT’S POSSIBLE IN REAL TIME Most AI models stall at the prototype phase. Grok advances beyond limits where others fail to perform. •⁠ ⁠Built for real-time decisioning across vast data landscapes •⁠ ⁠Optimized to process, synthesize, and respond with human-like fluidity •⁠ ⁠Deploys intelligence at the edge, not just in the cloud •⁠ ⁠Powers robotics, research, and infrastructure in live environments •⁠ ⁠Infrastructure built to evolve, not collapse, under scale Grok delivers high-performance AI built for deployment and real-world use. Source: xAI, Grok, Elon Musk

Mario Nawfal

28,837 views • 7 months ago

I am thrilled to join Rhoda AI as an advisor, where I am helping harness the abilities of large-scale pre-training and video models for robotics, putting many of my lab's research learnings of the past few years into practice! I will be in-person at the Mountain View office for part of the summer - reach out if you want to chat :)

I am thrilled to join Rhoda AI as an advisor, where I am helping harness the abilities of large-scale pre-training and video models for robotics, putting many of my lab's research learnings of the past few years into practice! I will be in-person at the Mountain View office for part of the summer - reach out if you want to chat :)

Vincent Sitzmann

55,283 views • 2 months ago

Got some exciting news to share! Arcium has acquired the tech and team from our largest Web2 competitor! and Solana just accellerated by 100x, the global supercomputer is evolving at an insane speed! Inpher has been leading in the confidential AI space for many years, raising over $25M from J.P. Morgan and others. The Inpher team developed the most advanced protocols for confidential AI and federated learning over the span of eight years, publishing some of the most foundational research in the MPC space for efficient ML, real number arithmetic, confidential tensor operations and oblivious algorithms. Inpher has achieved prestigious projects in the defense, finance and healthcare sectors and is now part of Arcium. Our new Arcium team members are amazing engineers and researchers, who have spent many years solving some of the most complex problems in secure Multi-Party computation and confidential computing, performing foundational research at Harvard, Berkeley, EPFL, and other prestigious institutions. I'm truly proud that we have onboarded this tremendous team – the smartest people I have ever had the pleasure to work with – into our big Arcium vision of building the hyper-encrypted internet with our global supercomputer. I am especially excited about not just integrating the most efficient confididential AI technology into Arcium but also fully open sourcing all of the acquired patents. The internet will never be the same!

Got some exciting news to share! Arcium has acquired the tech and team from our largest Web2 competitor! and Solana just accellerated by 100x, the global supercomputer is evolving at an insane speed! Inpher has been leading in the confidential AI space for many years, raising over $25M from J.P. Morgan and others. The Inpher team developed the most advanced protocols for confidential AI and federated learning over the span of eight years, publishing some of the most foundational research in the MPC space for efficient ML, real number arithmetic, confidential tensor operations and oblivious algorithms. Inpher has achieved prestigious projects in the defense, finance and healthcare sectors and is now part of Arcium. Our new Arcium team members are amazing engineers and researchers, who have spent many years solving some of the most complex problems in secure Multi-Party computation and confidential computing, performing foundational research at Harvard, Berkeley, EPFL, and other prestigious institutions. I'm truly proud that we have onboarded this tremendous team – the smartest people I have ever had the pleasure to work with – into our big Arcium vision of building the hyper-encrypted internet with our global supercomputer. I am especially excited about not just integrating the most efficient confididential AI technology into Arcium but also fully open sourcing all of the acquired patents. The internet will never be the same!

Yannik Schrade

156,421 views • 1 year ago

Increasingly, HTML Artifacts are becoming a core part of how I work with AI agents. Long-horizon agent sessions need a better way to surface insights about what work it has done. This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it. You also don't want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making. I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research. I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more. I believe Andrej Karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.

Increasingly, HTML Artifacts are becoming a core part of how I work with AI agents. Long-horizon agent sessions need a better way to surface insights about what work it has done. This may not be obvious right now, but as you start to let your agent work on dynamic workflows, large codebases, long-running loops (e.g., using /goal), and deep research tasks, you need a good way to present results. Chat window is not it. You also don't want to just trust everything the agents do. Artifacts help provide an important verification layer, which in turn enables important decision-making. I like HTML artifacts because I can just ask the agent to produce as many of them (and in whatever form) as I need to verify the work and make sense out of everything. I even built a nice tab system for my artifacts. They are great for continual learning and research. I use HTML artifacts for logging, tracking experiments, brainstorming, managing my inbox, code reviews, agent session management, deep research, writing, reading, and so much more. I believe Andrej Karpathy wrote about this somewhere: As we move on to more advanced applications of AI agents and outputs get more complex, we will start to find the need for even more advanced forms of interactions with AI, including interactive neural videos/simulations.

elvis

37,016 views • 3 months ago

.Sentient has just integrated Messari by Blockworks 's data and research into its AI-powered search platform, Sentient Chat. This partnership basically allows users to access Messari’s research directly through the Agent Hub in Sentient Chat where they can get instant answers and insights from Messari reports. The integration is done via Messari Copilot, which means users can now easily get to Messari’s crypto data without having to dig through extensive reports themselves. Messari's data and research now feeds into Sentient’s Agentic Perplexity, here users can access this in the Agent Hub for all their crypto related questions. Integrating Messari’s research now helps provide an open & community-driven platform for AI-powered search, ensuring that users have access to the best crypto data and insights in real time, while also expanding the functionality of Sentient’s Agent Hub, where users can find and use a growing library of AI agents for various tasks. Now I don't know about you but I know where I'll be getting my stats from moving forward.

.Sentient has just integrated Messari by Blockworks 's data and research into its AI-powered search platform, Sentient Chat. This partnership basically allows users to access Messari’s research directly through the Agent Hub in Sentient Chat where they can get instant answers and insights from Messari reports. The integration is done via Messari Copilot, which means users can now easily get to Messari’s crypto data without having to dig through extensive reports themselves. Messari's data and research now feeds into Sentient’s Agentic Perplexity, here users can access this in the Agent Hub for all their crypto related questions. Integrating Messari’s research now helps provide an open & community-driven platform for AI-powered search, ensuring that users have access to the best crypto data and insights in real time, while also expanding the functionality of Sentient’s Agent Hub, where users can find and use a growing library of AI agents for various tasks. Now I don't know about you but I know where I'll be getting my stats from moving forward.

Polygon Stats

32,718 views • 1 year ago

🔐 𝗣𝗼𝘄𝗲𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗡𝗲𝘅𝘁 𝗪𝗮𝘃𝗲 𝗼𝗳 𝗔𝗜 We're proud to partner with Stempoint to secure the future of decentralized AI compute. StemPoint is building one of the most advanced distributed AI infrastructures on the planet, sharded GPU compute, open model libraries, and secure data flows across a global mesh. Now with Naoris Protocol’s Sub-Zero Layer and dPoSec consensus, every node, model, and data packet gains: • Post-quantum cryptographic protection • Real-time validation of compute and algorithms • A decentralized trust layer for sovereign AI 🚫 No more trusting black boxes. ✅ Verifiable AI infrastructure at scale.

🔐 𝗣𝗼𝘄𝗲𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗡𝗲𝘅𝘁 𝗪𝗮𝘃𝗲 𝗼𝗳 𝗔𝗜 We're proud to partner with Stempoint to secure the future of decentralized AI compute. StemPoint is building one of the most advanced distributed AI infrastructures on the planet, sharded GPU compute, open model libraries, and secure data flows across a global mesh. Now with Naoris Protocol’s Sub-Zero Layer and dPoSec consensus, every node, model, and data packet gains: • Post-quantum cryptographic protection • Real-time validation of compute and algorithms • A decentralized trust layer for sovereign AI 🚫 No more trusting black boxes. ✅ Verifiable AI infrastructure at scale.

NaoX Protocol

23,336 views • 1 year ago

1/ Yesterday I posted about the revenue inconsistency in the “research” reports that Pudgy is publishing through CoinDesk. In an odd turn, these reports have now been updated to link my tweet as the source for the “1mm toys sold” claim that is at the root of the issue. As you can see below, the report from 2 months ago links out to my post yesterday questioning the numbers. This led me to go out and research and I found a series of inconsistencies from the Pudgy team wrt the revenue numbers they’ve shared through time.

1/ Yesterday I posted about the revenue inconsistency in the “research” reports that Pudgy is publishing through CoinDesk. In an odd turn, these reports have now been updated to link my tweet as the source for the “1mm toys sold” claim that is at the root of the issue. As you can see below, the report from 2 months ago links out to my post yesterday questioning the numbers. This led me to go out and research and I found a series of inconsistencies from the Pudgy team wrt the revenue numbers they’ve shared through time.

Sergito ❚ ❚

68,700 views • 6 months ago

We're delighted to announce the next cycle of recruitment for our 125th Anniversary Fellows and Chairs is now open! If you are an academic with a strong portfolio of global-leading research, we want to hear from you. Check out our vacancies and apply now:

We're delighted to announce the next cycle of recruitment for our 125th Anniversary Fellows and Chairs is now open! If you are an academic with a strong portfolio of global-leading research, we want to hear from you. Check out our vacancies and apply now:

Uni of Birmingham

42,969 views • 1 year ago

A new category of companies is emerging in 2026: the AI-native startup. - No sales team. - No research department. - No ops staff. Just an AI agent workforce. And they need AI-native startup infrastructure built for this, not repurposed enterprise clouds. Here is what that stack looks like with Aethir Claw.

A new category of companies is emerging in 2026: the AI-native startup. - No sales team. - No research department. - No ops staff. Just an AI agent workforce. And they need AI-native startup infrastructure built for this, not repurposed enterprise clouds. Here is what that stack looks like with Aethir Claw.

Aethir

23,222 views • 1 month ago

After a formative time at OpenAI, I’m launching Maison AGI (Maison AGI), a fashion house creating cultural artifacts for the AI era. Our first collection, Relic of Thought, is a collaboration with Ilya Sutskever (Ilya Sutskever), featuring his original artworks alongside a signature hat modeled after his iconic head. It is a study in conviction and the clarity of vision that gives thought its form. I believe we're living through something extraordinary that deserves to be remembered. Not in research papers or technical model cards, but in tangible objects we can hold, wear, and pass down. Working alongside some of the world's brilliant researchers building Claude and ChatGPT, I've come to realize they are among the most creative minds alive. Research, at its best, is an act of imagination: the ability to glimpse what doesn't yet exist and then build toward it. Many of them remain unseen, even as their work quietly reshapes our future. Though AGI progress can feel incomprehensible from the outside, its story is deeply human - full of curiosity, conviction, and creation. We're bridging that world with the creatives who give form to ideas. This may be humanity's last time to create a hand-crafted project before what we build surpasses us. Each collection is also a message to superintelligence itself: that we cared, and that we tried to make beauty out of understanding.

After a formative time at OpenAI, I’m launching Maison AGI (Maison AGI), a fashion house creating cultural artifacts for the AI era. Our first collection, Relic of Thought, is a collaboration with Ilya Sutskever (Ilya Sutskever), featuring his original artworks alongside a signature hat modeled after his iconic head. It is a study in conviction and the clarity of vision that gives thought its form. I believe we're living through something extraordinary that deserves to be remembered. Not in research papers or technical model cards, but in tangible objects we can hold, wear, and pass down. Working alongside some of the world's brilliant researchers building Claude and ChatGPT, I've come to realize they are among the most creative minds alive. Research, at its best, is an act of imagination: the ability to glimpse what doesn't yet exist and then build toward it. Many of them remain unseen, even as their work quietly reshapes our future. Though AGI progress can feel incomprehensible from the outside, its story is deeply human - full of curiosity, conviction, and creation. We're bridging that world with the creatives who give form to ideas. This may be humanity's last time to create a hand-crafted project before what we build surpasses us. Each collection is also a message to superintelligence itself: that we cared, and that we tried to make beauty out of understanding.

Karina

777,715 views • 10 months ago