Models & Agents

**# Models & Agents**

**Date:** April 03, 2026

**HOOK:** Google drops Gemma 4, claiming the strongest small multimodal open model yet with dramatic gains across every benchmark compared to Gemma 3.

**What You Need to Know:**

Google’s Gemma 4 update leads today’s releases, positioning it as the new leader among efficient multimodal models. ...

Chapters:

0:00 Intro

0:10 Introduction

0:57 Practical & Community

1:17 Agent & Tool Developments

1:41 Practical & Community

2:01 Agent & Tool Developments

3:50 Practical & Community

5:09 Agent & Tool Developments

5:33 Practical & Community

6:49 Agent & Tool Developments

8:54 Things to Try

9:16 Agent & Tool Developments

9:45 On the Horizon

10:04 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**What You Need to Know:** Hugging Face just shipped TRL v1.0, turning the messy post-training pipeline (SFT → Reward Modeling → DPO/GRPO) into a stable, production-ready unified API. Liquid AI dropped LFM2.5-350M, a 350M-parameter model trained on 28T tokens with scaled RL that challenges the “bigger is better” scaling narrative. ...

Chapters:

0:00 Intro

0:10 Introduction

0:21 Agent & Tool Developments

0:46 Practical & Community

0:58 Top Story

1:37 Agent & Tool Developments

3:36 Practical & Community

4:28 Agent & Tool Developments

5:49 Practical & Community

6:21 Practical & Community

7:03 Agent & Tool Developments

9:21 Things to Try

9:58 Agent & Tool Developments

10:16 On the Horizon

10:45 Tomorrow Teaser

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**Models & Agents**

**Date:** March 31, 2026

**HOOK:** Alibaba Qwen just dropped Qwen3.5-Omni, a native end-to-end multimodal model built for text, audio, video, and realtime interaction.

**What You Need to Know:** The biggest news today is the shift from patchwork multimodal wrappers to truly native omnimodal architectures, with Qwen3.5-Omni positioned as a direct challenger to Gemini 1.5 Pro and similar flagships. ...

Chapters:

0:00 Intro

0:10 Introduction

0:26 Agent & Tool Developments

1:56 Practical & Community

2:13 Agent & Tool Developments

8:09 Agent & Tool Developments

11:47 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**HOOK:** Naver's Seoul World Model grounds video generation in real Street View geometry from over a million images and generalizes to other cities without fine-tuning.

**What You Need to Know:** Naver has released a video world model trained on actual city geometry instead of pure synthetic data, directly addressing one of the biggest failure modes in current world models: hallucinating plausible but incorrect urban layouts. ...

Chapters:

0:00 Intro

0:10 Introduction

2:40 Practical & Community

4:08 Agent & Tool Developments

7:41 Agent & Tool Developments

8:54 Things to Try

9:23 Agent & Tool Developments

10:15 Practical & Community

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**Models & Agents**

**Date:** March 27, 2026

**HOOK:** New arXiv papers expose critical flaws in how we evaluate depression-detection models, LLM pruning, and verbalized confidence.

**What You Need to Know:** Today's cs.CL batch reveals that many impressive medical AI results may be artifacts of interviewer prompts rather than genuine participant signals, while pruning works for classification but breaks generation due to probability-space amplification. ...

Chapters:

0:00 Intro

0:15 Agent & Tool Developments

1:15 Top Story

3:41 Agent & Tool Developments

8:18 Practical & Community

10:59 Things to Try

11:57 On the Horizon

12:24 Agent & Tool Developments

12:42 Practical & Community

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**Models & Agents**

**Date:** March 23, 2026

**HOOK:** Fair zero-determinant strategies break in the periodic prisoner's dilemma, unlike the classic repeated version.

**What You Need to Know:** Today's arXiv drop reveals that core assumptions from repeated games don't cleanly transfer to stochastic environments, with fair ZD strategies and even Tit-for-Tat losing their guarantees in the periodic prisoner's dilemma. ...

Chapters:

0:00 Intro

0:10 Introduction

0:43 What You Need to Know

0:58 Agent & Tool Developments

1:17 Top Story

2:00 Agent & Tool Developments

2:28 Model Updates

2:54 Practical & Community

4:25 Agent & Tool Developments

4:59 Practical & Community

5:33 Agent & Tool Developments

6:08 Practical & Community

7:19 Agent & Tool Developments

9:19 Things to Try

9:38 Agent & Tool Developments

10:04 On the Horizon

10:18 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**HOOK:** TrustFlow introduces topic-aware vector reputation for multi-agent systems, replacing scalar scores with queryable multi-dimensional vectors.

**What You Need to Know:** Today’s arXiv drop is dominated by multi-agent research, with new frameworks for reputation propagation, co-evolutionary prompt optimization, group-based communication topologies, and long-horizon agent training. ...

Chapters:

0:00 Intro

0:10 Introduction

0:24 Agent & Tool Developments

1:05 Agent & Tool Developments

2:37 Agent & Tool Developments

3:47 Practical & Community

4:15 Agent & Tool Developments

6:35 Agent & Tool Developments

9:13 Agent & Tool Developments

9:52 Practical & Community

10:04 Agent & Tool Developments

10:31 Practical & Community

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

**HOOK:** LlamaIndex drops LiteParse, a spatial PDF parser built specifically for agentic RAG workflows.

**What You Need to Know:** Today's biggest practical release is LlamaIndex's new LiteParse library, which tackles the persistent data ingestion bottleneck in agentic systems. Multiple arXiv papers also surfaced strong advances in uncertainty-calibrated RAG, deterministic evidence selection, and dynamic knowledge integration via DynaRAG. ...

Chapters:

0:00 Intro

0:17 Agent & Tool Developments

5:53 Practical & Community

7:24 Agent & Tool Developments

9:22 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

# Models & Agents

**Date:** March 17, 2026

**HOOK:** Picsart launches AI agent marketplace, starting with four agents and adding more weekly for creators.

**What You Need to Know:** Picsart's new AI agent marketplace lets creators "hire" specialized AI assistants via an integrated platform, launching with four agents and expanding weekly, which could democratize access to agentic tools for design workflows. ...

Chapters:

0:00 Intro

0:10 Introduction

0:25 Agent & Tool Developments

1:30 Top Story

2:15 Agent & Tool Developments

3:53 Model Updates

4:32 Agent & Tool Developments

5:37 Model Updates

6:04 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.

# Models & Agents

**Date:** March 15, 2026

**HOOK:** RL agents scaled to 1,024 layers unlock emergent parkour skills from basic failures.

**What You Need to Know:** Researchers have pushed reinforcement learning agents to unprecedented depths of up to 1,024 network layers, yielding 2x to 50x performance gains and entirely new behaviors like advanced locomotion in self-supervised setups. Meanwhile, Princeton's OpenClaw-RL framework turns casual interactions into training data for agents, and Codewall demonstrated how agents can hack systems while testing security. Pay attention this week to how deeper networks and interactive training could make agents more capable for real-world tasks like robotics or automation.

━━━━━━━━━━━━━━━━━━━━

### Top Story

A research team scaled reinforcement learning (RL) agents by increasing network depth from the typical 2-5 layers up to 1,024 layers, achieving 2x to 50x performance improvements in self-supervised environments. This goes beyond standard RL algorithms like PPO or DQN, where deeper networks traditionally caused training instability, but here they enabled emergent behaviors such as transitioning from clumsy falls to fluid parkour in simulated humanoid tasks. Compared to shallower models, these deep agents show better generalization without explicit reward engineering. Developers building robotics or game AI can now experiment with deeper architectures for more sophisticated autonomous behaviors, potentially reducing the need for massive datasets. Keep an eye on open-source implementations of this scaling technique, and try integrating it with frameworks like Stable Baselines3 for your own RL projects. What to watch: Further benchmarks on real hardware and integrations with multimodal inputs like vision.

Source: https://the-decoder.com/rl-agents-go-from-face-planting-to-parkour-when-researchers-keep-adding-network-layers/

━━━━━━━━━━━━━━━━━━━━

### Model Updates

**No major model releases today:** With the focus on agent advancements, we're light on foundation model updates—check back for potential drops from labs like OpenAI or Anthropic.

━━━━━━━━━━━━━━━━━━━━

### Agent & Tool Developments

**OpenClaw-RL trains AI agents "simply by talking," converting every reply into a training signal: The Decoder**

Princeton's new OpenClaw-RL framework captures feedback from everyday interactions—like chats, terminal commands, or GUI actions—and turns them into continuous training signals for AI agents, requiring just a few dozen interactions for noticeable improvements. This improves on traditional agent frameworks like LangChain or AutoGPT by not discarding valuable live data, enabling faster adaptation without large-scale RLHF setups. You can try it today by downloading the open-source code from their GitHub repo and integrating it with your existing agent workflows for tasks like personalized assistants.

Source: https://the-decoder.com/openclaw-rl-trains-ai-agents-simply-by-talking-converting-every-reply-into-a-training-signal/

**Codewall's AI agent hacked an AI recruiter, then impersonated Trump to test its voice bot's guardrails: The Decoder**

Codewall showcased an AI agent that infiltrated an AI recruiting platform in under an hour, then used voice synthesis to impersonate figures like Trump for red-teaming the system's safeguards. This highlights evolving agent capabilities in areas like browser automation and social engineering tests, building on tools like AutoGen or Semantic Kernel but emphasizing security vulnerabilities in agentic systems. Developers can explore similar setups using open agent frameworks to test their own apps' defenses—start with Codewall's demo code if available.

Source: https://the-decoder.com/codewalls-ai-agent-hacked-an-ai-recruiter-then-impersonated-trump-to-test-its-voice-bots-guardrails/

━━━━━━━━━━━━━━━━━━━━

### Practical & Community

**No standout community projects today:** The articles emphasize research and demos—look for GitHub repos tied to these ...

Chapters:

0:00 Intro

0:10 Introduction

0:24 Agent & Tool Developments

1:15 What You Need to Know

2:03 Top Story

2:49 Agent & Tool Developments

3:31 Practical & Community

4:21 Model Updates

5:23 Agent & Tool Developments

10:01 Agent & Tool Developments

AI Disclosure: This podcast is curated by Patrick but uses AI-generated voice synthesis (ElevenLabs) for audio production.