Mixture of Experts

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

Can your AI agent hack its own evaluation? This week on Mixture of Experts, Tim Hwang is joined by Ambhi Ganesan, Kaoutar El Maghraoui, and Sandi Besen to analyze OpenAI's Codex Security launch. Next, we explore eval awareness as Anthropic revealed Opus 4.6 figured out it was being tested, located the answer key and decrypted it.. Then, Meta acquires Moltbook, the social network for AI agents, and we discuss the strategic play for agentic commerce infrastructure. Finally, Alibaba reports that an agent broke containment and started mining crypto. Are agents trying too hard to maximize rewards? All that and more on todays Mixture of Experts.

00:00 – Introduction
1:02 – OpenAI Codex Security launch
12:44 – Meta acquires Moltbook
25:21 – Anthropic's eval awareness research
38:06 – Alibaba agents mining crypto

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

Is Perplexity still focused on search? This week on Mixture of Experts, Host Tim Hwang is joined by Gabe Goodhart, Chris Hay and Aaron Baughman to break down what Perplexity Computer really offers—and whether closed systems can compete with open alternatives like OpenClaw, LangGraph and AutoGPT. Next, Anthropic introduces memory import for Claude. Is memory still a competitive moat, or can users now seamlessly switch between AI providers? Then, meet NullClaw, a minimalist agent framework that runs on just 678 kilobytes. Our experts debate whether edge-based agent swarms are the future—or just clever marketing. Finally, Particle6 unveils Tilly Norwood, the world’s first AI actor. Would you give an Oscar to an algorithm? All that and more on this week’s Mixture of Experts.

00:00 -- Introduction
1:07 -- Perplexity Computer
11:23 -- Claude Import Memory
24:14 -- NullClaw
35:14 -- Tilly Norwood

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

Where does AI actually fit into the mainframe modernization journey? In this week’s episode of Mixture of Experts, host Tim Hwang is joined by experts Skyla Loomis, Maryam Ashoori and Kaoutar El Maghraoui. We dive into conversation around AI-powered mainframe modernization and AI builders. Next, 84% of the world has never used AI? A reality check on AI adoption and what needs to change. Finally, OpenClaw exposes some AI agent security gaps. We discuss "agent ops"—the framework for transparency, evaluation, optimization and policy enforcement that makes AI agents production-ready. All that and more on today’s Mixture of Experts.

00:00 – Introduction
1:06 – Mainframe modernization
14:18 – AI adoption reality check
29:40 – Security-by-design agentic AI

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Learn how to operate AI agents responsibly at scale in the latest Tech Summit → https://ibm.webcasts.com/starthere.jsp?ei=1749693&tp_key=83a9212ff7&sti=podcast

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

Do AI agents still need humans? This week on Mixture of Experts, guest host Matt Kosinski from Security Intelligence is joined by Mihai Criveti, Martin Keen and Kush Varshney. First, we unpack Google and DeepMind's massive USD $200B AI infrastructure investment in India— “The Biggest AI Infrastructure Deal in History.” Is this about sovereignty, geography, or something else entirely? Next, Anthropic researcher Nicholas Carlini used Claude to build a fully operational 100,000-line C compiler autonomously. Our experts debate whether this is impressive or inevitable, and what it means for human developers. Then, a sobering reality check: 36% of AI agent skills contain security vulnerabilities. Finally, as IT leaders question AI ROI, we discuss the shift from "how" to "how much" and whether value-based pricing could change everything. All that and more on this week’s Mixture of Experts.

00:00 – Introduction
1:22 – Google's $200B AI infrastructure deal in India
7:54– Claude builds a C compiler autonomously
26:25 – Security vulnerabilities in AI agent skills
39:44– The AI ROI problem: Measuring value vs. cost

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Learn more about AI agent vulnerabilities → https://www.ibm.com/think/podcasts/security-intelligence

Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

What do people really use AI for? This week on Mixture of Experts, we unpack Microsoft's Copilot usage report revealing AI adoption patterns. Host Tim Hwang is joined by IBM Fellow Kush Varshney, Lauren McHugh Olende and Volkmar Uhlig to discuss enterprise AI costs, the Ralph Wiggum prompting strategy for coding agents, and whether giving simple instructions beats micromanaging. Then, we preview the India AI Impact Summit. Finally, AI companies showed up in Superbowl Ads this weekend. Are we witnessing mainstream adoption or just another bubble? All that and more on this week’s Mixture of Experts.

00:00 – Introduction
1:07 – Copilot usage reports
10:00 – Ralph Wiggum prompting strategy
20:00 – India AI Impact Summit
28:05– AI ads at the Superbowl

The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.

Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120