AI Trends 2026: OpenClaw Agents, Reasoning LLMs, and More with Sebastian Raschka - #762 | The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

The podcast explores the current state and future trends of Large Language Models (LLMs), particularly focusing on advancements since last year and expectations for 2026. Independent LLM researcher Sebastian Raschka highlights the shift towards post-training techniques, such as reasoning and tool use, as key areas of development. The discussion covers the practical applications of LLMs, including code improvement and workflow automation, with both the host and guest sharing their experiences using custom tools. They address the importance of verifiable rewards in training and the potential for expanding verification paradigms beyond math and code. The conversation also touches on inference scaling techniques like self-consistency and self-refinement, and the challenges and opportunities in agentic uses of LLMs, including multi-agent systems.

Outlines

Part 1: Current State and Practical Evolution

Part 2: Real-World Applications and Workflows

Part 3: Reasoning and Training Methodologies

Part 4: Scaling and Optimization Techniques

Part 5: Architecture and Future Outlook

Sign in to continue reading, translating and more.

Continue

AI Trends 2026: OpenClaw Agents, Reasoning LLMs, and More with Sebastian Raschka - #762

The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

Part 1: Current State and Practical Evolution

LLM Reasoning Revolution: Post-Training Focus and Tool Use Advancements

Practical LLM Evolution: Harnessing Tooling and Interface Improvements

Incremental LLM Improvements: Robustness, Trust, and Reasoning Efficiency

Part 2: Real-World Applications and Workflows

OpenClaw's Impact: Engaging Users and Demonstrating LLM Capabilities

Custom Workflow Tools: Leveraging LLMs for Deterministic Task Automation

LLMs as Classifiers: Streamlining Unstructured Data Parsing and Workflow Automation

Social Media Hype vs. Reality: LLM One-Shot Claims and Gradual Improvements

Part 3: Reasoning and Training Methodologies

LLM Innovation Focus: Reasoning, Inference Scaling, and Agentic Use

Verifiable Rewards: Scaling Reasoning Training in LLMs for Math and Coding

Expanding Verification: Applying LLM Reasoning Beyond Math and Code

Part 4: Scaling and Optimization Techniques

Inference Scaling: Enhancing LLM Performance Through Compute and Model Selection

Automatic Reasoning: Integrating Self-Consistency and Refinement in LLMs

Agentic LLM Uses: Task-Oriented Iterations and Plugin Development

Multi-Agent Systems: Context Engineering and Training for Interactive Environments

Part 5: Architecture and Future Outlook

LLM Architecture: Incremental Changes and the Rise of Mixture-of-Experts

Continual Learning: The Dream of Self-Improving LLMs and the Role of Long Context

Future LLM Directions: Text Diffusion Models and Open Weight Releases

Building Reasoning Models: Sebastian Raschka's New Book on LLM Post-Training

AI Trends 2026: OpenClaw Agents, Reasoning LLMs, and More with Sebastian Raschka - #762

The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

Part 1: Current State and Practical Evolution

00:00LLM Reasoning Revolution: Post-Training Focus and Tool Use Advancements

LLM Reasoning Revolution: Post-Training Focus and Tool Use Advancements

04:33Practical LLM Evolution: Harnessing Tooling and Interface Improvements

Practical LLM Evolution: Harnessing Tooling and Interface Improvements

10:43Incremental LLM Improvements: Robustness, Trust, and Reasoning Efficiency

Incremental LLM Improvements: Robustness, Trust, and Reasoning Efficiency

Part 2: Real-World Applications and Workflows

14:23OpenClaw's Impact: Engaging Users and Demonstrating LLM Capabilities

OpenClaw's Impact: Engaging Users and Demonstrating LLM Capabilities

16:10Custom Workflow Tools: Leveraging LLMs for Deterministic Task Automation

Custom Workflow Tools: Leveraging LLMs for Deterministic Task Automation

21:17LLMs as Classifiers: Streamlining Unstructured Data Parsing and Workflow Automation

LLMs as Classifiers: Streamlining Unstructured Data Parsing and Workflow Automation

27:55Social Media Hype vs. Reality: LLM One-Shot Claims and Gradual Improvements

Social Media Hype vs. Reality: LLM One-Shot Claims and Gradual Improvements

Part 3: Reasoning and Training Methodologies

29:42LLM Innovation Focus: Reasoning, Inference Scaling, and Agentic Use

LLM Innovation Focus: Reasoning, Inference Scaling, and Agentic Use

33:14Verifiable Rewards: Scaling Reasoning Training in LLMs for Math and Coding

Verifiable Rewards: Scaling Reasoning Training in LLMs for Math and Coding

38:46Expanding Verification: Applying LLM Reasoning Beyond Math and Code

Expanding Verification: Applying LLM Reasoning Beyond Math and Code

Part 4: Scaling and Optimization Techniques

42:10Inference Scaling: Enhancing LLM Performance Through Compute and Model Selection

Inference Scaling: Enhancing LLM Performance Through Compute and Model Selection

45:51Automatic Reasoning: Integrating Self-Consistency and Refinement in LLMs

Automatic Reasoning: Integrating Self-Consistency and Refinement in LLMs

50:22Agentic LLM Uses: Task-Oriented Iterations and Plugin Development

Agentic LLM Uses: Task-Oriented Iterations and Plugin Development

53:35Multi-Agent Systems: Context Engineering and Training for Interactive Environments

Multi-Agent Systems: Context Engineering and Training for Interactive Environments

Part 5: Architecture and Future Outlook

58:49LLM Architecture: Incremental Changes and the Rise of Mixture-of-Experts

LLM Architecture: Incremental Changes and the Rise of Mixture-of-Experts

1:05:32Continual Learning: The Dream of Self-Improving LLMs and the Role of Long Context

Continual Learning: The Dream of Self-Improving LLMs and the Role of Long Context

1:11:44Future LLM Directions: Text Diffusion Models and Open Weight Releases

Future LLM Directions: Text Diffusion Models and Open Weight Releases

1:13:58Building Reasoning Models: Sebastian Raschka's New Book on LLM Post-Training

Building Reasoning Models: Sebastian Raschka's New Book on LLM Post-Training