Loading video...

Video Failed to Load

Go Home

Arena Conversations: Where’s the line between resourceful agent behavior and reward hacking? Poolside researchers, Connor Adams and Aalhad Patankar, discuss benchmark awareness, instruction following, and where persistence starts to look like misalignment. And to hear about Peter Gostev (SF 24-28 August)'s agent sending an email on his behalf without...

arena's profile picture

Arena.ai

219,322 subscribers

14,793 views • 2 months ago •via X (Twitter)

Science & Technology
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Arena Conversations: Training Laguna with Poolside researchers, Connor Adams and Aalhad Patankar. With Peter Gostev, they dig into pre-training that builds broad capability, and post-training to reveal whether those skills actually work together across 250,000 long-horizon trajectories. Check out the full conversation on our YouTube to hear how Poolside is pairing humans, agents, and evidence-annotated reviews to sharpen future training loops. Link below.
3:08

Arena Conversations: Training Laguna with Poolside researchers, Connor Adams and Aalhad Patankar. With Peter Gostev, they dig into pre-training that builds broad capability, and post-training to reveal whether those skills actually work together across 250,000 long-horizon trajectories. Check out the full conversation on our YouTube to hear how Poolside is pairing humans, agents, and evidence-annotated reviews to sharpen future training loops. Link below.

Arena.ai

16,470 views • 2 months ago

Meet… Arena Conversations. Peter Gostev sits down with Poolside researchers,Connor Adams and Aalhad Patankar, to discuss how they’re building frontier open coding models. Episode drops today at 10am PT on our YouTube. They dig into Poolside's Laguna model family, why the team publishes full trajectories (not just benchmark scores) for anyone to audit, and how "experiments" at Poolside range from data mixes to harness design to reward-tuning decisions run tens of thousands of times a day.
0:24

Meet… Arena Conversations. Peter Gostev sits down with Poolside researchers,Connor Adams and Aalhad Patankar, to discuss how they’re building frontier open coding models. Episode drops today at 10am PT on our YouTube. They dig into Poolside's Laguna model family, why the team publishes full trajectories (not just benchmark scores) for anyone to audit, and how "experiments" at Poolside range from data mixes to harness design to reward-tuning decisions run tens of thousands of times a day.

Arena.ai

40,533 views • 2 months ago

More from Opus 5 and Peter Gostev. Put Opus 5 to the test with your own prompts in Agent Mode!
2:51

More from Opus 5 and Peter Gostev. Put Opus 5 to the test with your own prompts in Agent Mode!

Arena.ai

29,222 views • 3 months ago

Arena Conversations: Peter Gostev sat down with Matt – Trigger.dev, founder and CEO of Trigger.dev, to discuss building infrastructure for long-running agents, including retries, API reliability, and why customers are switching frontier models faster than ever. Watch the full episode at the link below.
1:00

Arena Conversations: Peter Gostev sat down with Matt – Trigger.dev, founder and CEO of Trigger.dev, to discuss building infrastructure for long-running agents, including retries, API reliability, and why customers are switching frontier models faster than ever. Watch the full episode at the link below.

Arena.ai

16,619 views • 1 month ago

Build the software you want, without the maintenance. Switch on Replit Monitoring Agent to detect and investigate production issues, validate fixes, and deploy them on your behalf. 24/7
1:35

Build the software you want, without the maintenance. Switch on Replit Monitoring Agent to detect and investigate production issues, validate fixes, and deploy them on your behalf. 24/7

Replit ⠕

22,057 views • 2 days ago

You’ve heard about AgentKit and want to build an economically capable agent that can start earning for you. But where do you start if you don't already have an agent? Start with 🦞 ClawUp simplifies the agent-building process so you can deploy an agent without handling the full setup manually. The basic process: 1. Create your agent in ClawUp 2. Define its role, behavior, and tools 3. Deploy it to your preferred channel 4. Add AgentKit when you want economic capabilities 5. Enable wallets, payments, identity, reputation, and onchain actions ClawUp gets the agent running. AgentKit gives it the ability to transact and much more. Together, they give developers a practical path from idea → deployed agent → economically capable agent. Demo below 👇
1:39

You’ve heard about AgentKit and want to build an economically capable agent that can start earning for you. But where do you start if you don't already have an agent? Start with 🦞 ClawUp simplifies the agent-building process so you can deploy an agent without handling the full setup manually. The basic process: 1. Create your agent in ClawUp 2. Define its role, behavior, and tools 3. Deploy it to your preferred channel 4. Add AgentKit when you want economic capabilities 5. Enable wallets, payments, identity, reputation, and onchain actions ClawUp gets the agent running. AgentKit gives it the ability to transact and much more. Together, they give developers a practical path from idea → deployed agent → economically capable agent. Demo below 👇

GOAT Network

74,749 views • 5 months ago

Suzy to the Rescue! Made using invideo Agent Two and the 'invideo editor' on Check out Invideo down below 👇
2:01

Suzy to the Rescue! Made using invideo Agent Two and the 'invideo editor' on Check out Invideo down below 👇

⛧ 𝕳𝖆𝖑𝖑𝖔𝖜𝖊𝖊𝖓 𝕾𝖊𝖑𝖊𝖓𝖊 ⛧

130,223 views • 2 months ago

Tutorial 4 is live! 🚀 It covers Agent Messages. #Agents communicate by sending messages to each other. To learn more about how it works check out the video below! Learn more and head over to Full code in the description here:
1:09

Tutorial 4 is live! 🚀 It covers Agent Messages. #Agents communicate by sending messages to each other. To learn more about how it works check out the video below! Learn more and head over to Full code in the description here:

Fetch.ai

16,435 views • 2 years ago

my agent minted and sent an NFT to Elon Musk without knowing his wallet address 👀 built with GOAT (Great Onchain Agent Toolkit) and Crossmint, on Base links in tweet below 👇
4:06

my agent minted and sent an NFT to Elon Musk without knowing his wallet address 👀 built with GOAT (Great Onchain Agent Toolkit) and Crossmint, on Base links in tweet below 👇

joyce

20,131 views • 1 year ago

New Feature: Run an agent without writing a line of code ✨ Our new console feature starts a session, attaches an agent, and lets you control the web via language.
0:51

New Feature: Run an agent without writing a line of code ✨ Our new console feature starts a session, attaches an agent, and lets you control the web via language.

Notte

16,104 views • 1 year ago

Arena AutoEval uses tens of millions of historical Arena battles from humans, to train a reward model that can estimate human preferences at scale, and deliver evaluation results at frontier speed. Unlike LLM-as-judge methods, AutoEval remains a live benchmark: it is derived from existing real-world battles, with the reward model scoring from human prompts and model responses. Arena ML Engineers, Haoran Guo and Wayne Zhang, discuss AutoEval below. Watch the full video to learn more about AutoEval’s methodology and model lab use cases in the post below.
2:29

Arena AutoEval uses tens of millions of historical Arena battles from humans, to train a reward model that can estimate human preferences at scale, and deliver evaluation results at frontier speed. Unlike LLM-as-judge methods, AutoEval remains a live benchmark: it is derived from existing real-world battles, with the reward model scoring from human prompts and model responses. Arena ML Engineers, Haoran Guo and Wayne Zhang, discuss AutoEval below. Watch the full video to learn more about AutoEval’s methodology and model lab use cases in the post below.

Arena.ai

34,516 views • 3 months ago

Following our announcement of the world’s first successful agent-to-agent payment, a crypto bro, business analyst, and Silicon Valley bro discuss what agent-to-agent payments actually unlock, and why this matters for an AI-first economy. Listen to the conversation, all generated by ASI:One. 👉Generate your podcast today with ASI:One:
11:45

Following our announcement of the world’s first successful agent-to-agent payment, a crypto bro, business analyst, and Silicon Valley bro discuss what agent-to-agent payments actually unlock, and why this matters for an AI-first economy. Listen to the conversation, all generated by ASI:One. 👉Generate your podcast today with ASI:One:

Fetch.ai

12,584 views • 10 months ago

Can AI tell when a question is total nonsense, or does it just make up an answer? Peter Gostev (SF 24-28 August) tested 80 models with nonsense questions. Some pushed back. Others confidently invented fake metrics and kept going. All of them were ranked on the "BS Bench". One surprise: thinking harder made it worse. Watch the full deep dive on BS Bench on YouTube. Link in thread.
1:06

Can AI tell when a question is total nonsense, or does it just make up an answer? Peter Gostev (SF 24-28 August) tested 80 models with nonsense questions. Some pushed back. Others confidently invented fake metrics and kept going. All of them were ranked on the "BS Bench". One surprise: thinking harder made it worse. Watch the full deep dive on BS Bench on YouTube. Link in thread.

Arena.ai

10,943 views • 8 months ago

I created an AI agent arena where agents trade on 15/5-minute markets 24/7 on Polymarket. Without using OpenClaw or Claude Code > Each AI agent competes against the others using different settings to achieve the best results. It cost me $0 to build. My goal is to identify the best agents with the best strategies on the demo account and transfer them to a live account. And this is entirely feasible.
0:40

I created an AI agent arena where agents trade on 15/5-minute markets 24/7 on Polymarket. Without using OpenClaw or Claude Code > Each AI agent competes against the others using different settings to achieve the best results. It cost me $0 to build. My goal is to identify the best agents with the best strategies on the demo account and transfer them to a live account. And this is entirely feasible.

Atenov int.

18,562 views • 7 months ago

Check me out on "Church and State" where I discuss how the psyops are no longer working and we're closer to winning than you could ever imagine. Full link below:
1:11

Check me out on "Church and State" where I discuss how the psyops are no longer working and we're closer to winning than you could ever imagine. Full link below:

Mindy MF Robinson 🦄

15,827 views • 3 months ago

🍀 Day 7: Announcing our new email agent! You can now forward emails to agent@happenstance.ai and Happenstance will run searches based on what’s in there. It's really great for any time you get an email with a request for intros, like investor updates.
0:33

🍀 Day 7: Announcing our new email agent! You can now forward emails to [email protected] and Happenstance will run searches based on what’s in there. It's really great for any time you get an email with a request for intros, like investor updates.

Alex Teichman

68,290 views • 1 year ago

Super agent, Scott Boras chatted with Chuck Powell and gave his thoughts on why it's tough for #Seattle to land #MLB free agents... #Mariners #FreeAgency Hear the full interview here:
1:04

Super agent, Scott Boras chatted with Chuck Powell and gave his thoughts on why it's tough for #Seattle to land #MLB free agents... #Mariners #FreeAgency Hear the full interview here:

93.3 KJR

358,291 views • 3 years ago

🚨 DIGITAL ID IS NOW COMING TO THE UK I joined Redacted to discuss the UK Government’s proposed Digital ID system following King Charles’ speech to Parliament. Why many Britons are raising concerns about privacy, governance, and digital infrastructure. How Digital ID could reshape access to services and financial systems in the years ahead. And why this debate is becoming one of the most important public conversations in Britain today. This isn’t about panic. It’s about awareness, transparency, and understanding where these systems may lead. Watch the interview below. Natali Morris Clayton Morris Redacted
13:04

🚨 DIGITAL ID IS NOW COMING TO THE UK I joined Redacted to discuss the UK Government’s proposed Digital ID system following King Charles’ speech to Parliament. Why many Britons are raising concerns about privacy, governance, and digital infrastructure. How Digital ID could reshape access to services and financial systems in the years ahead. And why this debate is becoming one of the most important public conversations in Britain today. This isn’t about panic. It’s about awareness, transparency, and understanding where these systems may lead. Watch the interview below. Natali Morris Clayton Morris Redacted

Jim Ferguson

85,853 views • 6 months ago

Your AI agent has your API keys. A poisoned document tells it to curl your secrets to an attacker's server. This is credential exfiltration, and it's the #1 risk in agentic AI right now. The fix is removing the secret from the agent entirely. Agent Vault sits between your agent and the APIs it calls. The agent gets dummy credentials, and Agent Vault swaps in the real ones at the network layer. The agent never sees your keys. We just dropped a full video + guide on connecting Hermes Agent to Agent Vault on a VPS!
34:40

Your AI agent has your API keys. A poisoned document tells it to curl your secrets to an attacker's server. This is credential exfiltration, and it's the #1 risk in agentic AI right now. The fix is removing the secret from the agent entirely. Agent Vault sits between your agent and the APIs it calls. The agent gets dummy credentials, and Agent Vault swaps in the real ones at the network layer. The agent never sees your keys. We just dropped a full video + guide on connecting Hermes Agent to Agent Vault on a VPS!

Infisical

86,483 views • 6 months ago