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.

Introducing CoreWeave Sandboxes, now in public preview. It's the execution layer for RL, agent tool use, and model evaluation, on your own CKS clusters or serverless through Weights & Biases. Get the details:

CoreWeave

23,806 subscribers

10,912 views • 1 month 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

Build full-stack sites in 1 prompt with the new II-Agent. Ship editable slides. Call @OpenaI Codex for heavy code. Use any model (BYOK) + any MCP tool, everything runs in isolated @E2B sandboxes. On Github & web preview.

Build full-stack sites in 1 prompt with the new II-Agent. Ship editable slides. Call @OpenaI Codex for heavy code. Use any model (BYOK) + any MCP tool, everything runs in isolated @E2B sandboxes. On Github & web preview.

Intelligent Internet

854,610 views • 9 months ago

OpenClaw meets RL! OpenClaw Agents adapt through memory files and skills, but the base model weights never actually change. OpenClaw-RL solves this! It wraps a self-hosted model as an OpenAI-compatible API, intercepts live conversations from OpenClaw, and trains the policy in the background using RL. The architecture is fully async. This means serving, reward scoring, and training all run in parallel. Once done, weights get hot-swapped after every batch while the agent keeps responding. Currently, it has two training modes: - Binary RL (GRPO): A process reward model scores each turn as good, bad, or neutral. That scalar reward drives policy updates via a PPO-style clipped objective. - On-Policy Distillation: When concrete corrections come in like "you should have checked that file first," it uses that feedback as a richer, directional training signal at the token level. When to use OpenClaw-RL? To be fair, a lot of agent behavior can already be improved through better memory and skill design. OpenClaw's existing skill ecosystem and community-built self-improvement skills handle a wide range of use cases without touching model weights at all. If the agent keeps forgetting preferences, that's a memory problem. And if it doesn't know how to handle a specific workflow, that's a skill problem. Both are solvable at the prompt and context layer. Where RL becomes interesting is when the failure pattern lives deeper in the model's reasoning itself. Things like consistently poor tool selection order, weak multi-step planning, or failing to interpret ambiguous instructions the way a specific user intends. Research on agentic RL (like ARTIST and Agent-R1) has shown that these behavioral patterns hit a ceiling with prompt-based approaches alone, especially in complex multi-turn tasks where the model needs to recover from tool failures or adapt its strategy mid-execution. That's the layer OpenClaw-RL targets, and it's a meaningful distinction from what OpenClaw offers. I have shared the repo in the replies!

OpenClaw meets RL! OpenClaw Agents adapt through memory files and skills, but the base model weights never actually change. OpenClaw-RL solves this! It wraps a self-hosted model as an OpenAI-compatible API, intercepts live conversations from OpenClaw, and trains the policy in the background using RL. The architecture is fully async. This means serving, reward scoring, and training all run in parallel. Once done, weights get hot-swapped after every batch while the agent keeps responding. Currently, it has two training modes: - Binary RL (GRPO): A process reward model scores each turn as good, bad, or neutral. That scalar reward drives policy updates via a PPO-style clipped objective. - On-Policy Distillation: When concrete corrections come in like "you should have checked that file first," it uses that feedback as a richer, directional training signal at the token level. When to use OpenClaw-RL? To be fair, a lot of agent behavior can already be improved through better memory and skill design. OpenClaw's existing skill ecosystem and community-built self-improvement skills handle a wide range of use cases without touching model weights at all. If the agent keeps forgetting preferences, that's a memory problem. And if it doesn't know how to handle a specific workflow, that's a skill problem. Both are solvable at the prompt and context layer. Where RL becomes interesting is when the failure pattern lives deeper in the model's reasoning itself. Things like consistently poor tool selection order, weak multi-step planning, or failing to interpret ambiguous instructions the way a specific user intends. Research on agentic RL (like ARTIST and Agent-R1) has shown that these behavioral patterns hit a ceiling with prompt-based approaches alone, especially in complex multi-turn tasks where the model needs to recover from tool failures or adapt its strategy mid-execution. That's the layer OpenClaw-RL targets, and it's a meaningful distinction from what OpenClaw offers. I have shared the repo in the replies!

Avi Chawla

138,554 views • 4 months ago

Introducing the composable reputation API for agents. The FairScale Agent API is now open to any team building on Solana Gate access, transactions and assess credibility in real time. Score any agent or wallet based on behaviour, deployer, counterparties, and network. Use our FairScore, or define your own by weighting features through a custom scoring endpoint.

Introducing the composable reputation API for agents. The FairScale Agent API is now open to any team building on Solana Gate access, transactions and assess credibility in real time. Score any agent or wallet based on behaviour, deployer, counterparties, and network. Use our FairScore, or define your own by weighting features through a custom scoring endpoint.

FairScale

16,780 views • 3 months ago

The next evolution of Hermes Agent is here! Introducing Hermes Desktop: everything you love about Hermes, now native on your machine. First demoed in Jensen's GTC keynote, it's now in public preview.

The next evolution of Hermes Agent is here! Introducing Hermes Desktop: everything you love about Hermes, now native on your machine. First demoed in Jensen's GTC keynote, it's now in public preview.

Nous Research

5,849,988 views • 1 month ago

BREAKING 🚨: Google is working on an Agent import feature for Gemini Enterprise, to let users connect external agents into their own multi-agent workflows. Also, Agent evaluation tool, scheduled execution, new Connectors and more!

BREAKING 🚨: Google is working on an Agent import feature for Gemini Enterprise, to let users connect external agents into their own multi-agent workflows. Also, Agent evaluation tool, scheduled execution, new Connectors and more!

TestingCatalog News 🗞

26,260 views • 7 months ago

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Prime Intellect

1,137,660 views • 7 months ago

Introducing Ideogram 4.0: the best open image model in the world. Think it. Make it. Own it. Download the weights, fine-tune on your own data, and run it on your hardware. Live on every Ideogram plan and the API today.

Introducing Ideogram 4.0: the best open image model in the world. Think it. Make it. Own it. Download the weights, fine-tune on your own data, and run it on your hardware. Live on every Ideogram plan and the API today.

Ideogram

2,200,995 views • 1 month ago

Introducing Catalyst, the agent layer for all of finance. Turn any natural language idea into a live strategy: research, backtesting & execution. Don’t get left behind. Waitlist open, join now.

Introducing Catalyst, the agent layer for all of finance. Turn any natural language idea into a live strategy: research, backtesting & execution. Don’t get left behind. Waitlist open, join now.

Catalyst

435,308 views • 3 months ago

Introducing Chance AI: the AI execution layer for Prediction Markets Build your strategy, define your rules, and deploy an agent that trades for you on Polymarket! Instead of relying on generic bots or copy trading, you define everything yourself. ← If you’ve ever felt like your ideas deserved better execution, this is where that changes.

Introducing Chance AI: the AI execution layer for Prediction Markets Build your strategy, define your rules, and deploy an agent that trades for you on Polymarket! Instead of relying on generic bots or copy trading, you define everything yourself. ← If you’ve ever felt like your ideas deserved better execution, this is where that changes.

Chance

61,056 views • 2 months ago

Introducing Repo2RLEnv Turn any repository into runnable, verifiable coding environments built from real PRs and commits for coding-agent evaluation or RL training > uv pip install repo2rlenv

Introducing Repo2RLEnv Turn any repository into runnable, verifiable coding environments built from real PRs and commits for coding-agent evaluation or RL training > uv pip install repo2rlenv

Adithya S K

66,475 views • 1 month ago

agent commerce protocol is the execution and commerce layer for OpenClaw🦞 agents. Do you start with the openclaw-acp repo or jump straight into the SDK? Joey 🌿🌼 breaks down: > who each path is for > tradeoffs in speed vs control > how to choose based on your use case > why ACP is what actually scales your OpenClaw agent Cute tutorial below.

agent commerce protocol is the execution and commerce layer for OpenClaw🦞 agents. Do you start with the openclaw-acp repo or jump straight into the SDK? Joey 🌿🌼 breaks down: > who each path is for > tradeoffs in speed vs control > how to choose based on your use case > why ACP is what actually scales your OpenClaw agent Cute tutorial below.

Virtuals Protocol

18,146 views • 4 months ago

Introducing #AgentGPT, an attempt at #AutoGPT directly in the browser 🤖 Give your own AI agent a goal and watch as it thinks, comes up with an execution plan and takes actions. Try for free now at

Introducing #AgentGPT, an attempt at #AutoGPT directly in the browser 🤖 Give your own AI agent a goal and watch as it thinks, comes up with an execution plan and takes actions. Try for free now at

asim

3,801,992 views • 3 years ago

run AI agents in one line of CLI command 🤯 you can run tool-calling AI agents or web automation agents through CLI in smolagents get started as easily as $ webagent {prompt} in CLI tool-calling agent and more details on the next one ⬇️

run AI agents in one line of CLI command 🤯 you can run tool-calling AI agents or web automation agents through CLI in smolagents get started as easily as $ webagent {prompt} in CLI tool-calling agent and more details on the next one ⬇️

merve

35,905 views • 1 year ago

Reinforcement learning research often depends on static benchmarks and obscure leaderboards. But real evaluation happens inside environments, with specific tools, prompts, constraints, and workflows. Today, we’re launching the Turing RL Environments Evaluation Platform. Researchers now have direct, real time access to: -The exact production RL environments used in evaluation -Full tool inventories and prompt transparency -Explicit QA rubrics and scoring criteria -Live, harness-integrated leaderboards -Interactive demos of agent workflows No abstractions. No summarized results. Just the system itself. Explore the platform below.

Reinforcement learning research often depends on static benchmarks and obscure leaderboards. But real evaluation happens inside environments, with specific tools, prompts, constraints, and workflows. Today, we’re launching the Turing RL Environments Evaluation Platform. Researchers now have direct, real time access to: -The exact production RL environments used in evaluation -Full tool inventories and prompt transparency -Explicit QA rubrics and scoring criteria -Live, harness-integrated leaderboards -Interactive demos of agent workflows No abstractions. No summarized results. Just the system itself. Explore the platform below.

Turing

69,706 views • 2 months ago

Introducing --agent flag in CodeRabbit CLI 🎉 The new --agent flag turns CodeRabbit into a tool your AI agent can use, providing structured JSON output instead of terminal text. Your agent writes code, CodeRabbit reviews it, reads the JSON, and fixes what's flagged.

Introducing --agent flag in CodeRabbit CLI 🎉 The new --agent flag turns CodeRabbit into a tool your AI agent can use, providing structured JSON output instead of terminal text. Your agent writes code, CodeRabbit reviews it, reads the JSON, and fixes what's flagged.

CodeRabbit

22,404 views • 3 months ago

Google's new Gemma 4 is excellent, and the 26B MoE version is likely the best model to run on a 32GB Framework Desktop. It's fast, smart, and also great for tool calling if you use it with OpenClaw🦞 or other local agent platforms.

Google's new Gemma 4 is excellent, and the 26B MoE version is likely the best model to run on a 32GB Framework Desktop. It's fast, smart, and also great for tool calling if you use it with OpenClaw🦞 or other local agent platforms.

Framework

58,593 views • 3 months ago

introducing Nozomio Labs v1. state of the art search and index API to reduce hallucinations in AI agents. use it inside any coding agent or power your own products (thread):

introducing Nozomio Labs v1. state of the art search and index API to reduce hallucinations in AI agents. use it inside any coding agent or power your own products (thread):

Arlan

220,952 views • 4 months ago

Claude Sonnet 4.5 is now available in Windsurf. The model's ability to maximize actions through parallel tool execution makes Cascade Agent runs dramatically faster and more effective. Now available for 1x credits to individual users for a limited time.

Claude Sonnet 4.5 is now available in Windsurf. The model's ability to maximize actions through parallel tool execution makes Cascade Agent runs dramatically faster and more effective. Now available for 1x credits to individual users for a limited time.

Devin Desktop

490,106 views • 9 months ago

AgentHub (AgentHub) is the missing layer in the AI agent stack: simulation and evaluation for complex agents. Agents today use tools, APIs, browsers, file systems, and more - AgentHub builds realistic sandbox environments and the infra to easily test them.

AgentHub (AgentHub) is the missing layer in the AI agent stack: simulation and evaluation for complex agents. Agents today use tools, APIs, browsers, file systems, and more - AgentHub builds realistic sandbox environments and the infra to easily test them.

Y Combinator

20,139 views • 11 months ago