Is open source winning the AI race? This week on Mixture of Experts, we analyze three major model releases that dropped in the final weeks of 2025: Mistral 3, DeepSeek-V3.2 and Claude Opus 4.5. Our experts discuss what makes each model unique—from Mistral’s multimodal capabilities to DeepSeek’s reasoning-first approach and Claude’s developer focus. Are there too many good models? Next, a provocative blog post from Theory Ventures argues Gemini 3 proves scaling laws are throwing more compute at the problem. We debate if scaling laws are a universal truth. Finally, Amazon just blocked ChatGPT’s shopping research agent from accessing product data. We discuss the business incentives threatening the agent dream. Join host Tim Hwang and panelists Aaron Baughman, Abraham Daniels and Gabe Goodhart on this week’s Mixture of Experts for more! 00:00 – Intro 02:05 -- Model launches: Mistral 3, DeepSeek-V3.2 and Claude Opus 4.5 15:32 -- AI scaling laws & Gemini 3 26:23 --Amazon blocking ChatGPT shopping research agent 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. Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 #Mistral3 #DeepSeek #ClaudeOpus #AIscalinglaws #AIagents
Will Black Friday 2025 be the breakout moment for agentic commerce? This week on Mixture of Experts, we're departing from our usual format for a Thanksgiving special focused entirely on AI agents. Host Tim Hwang is joined by Chris Hay, Lauren McHugh Olende and Volkmar Uhlig to debate whether 2025 is truly the "year of agents." Our experts break down why consumer-facing agentic commerce still faces major hurdles. We also explore the massive gap between building agent prototypes and deploying them at scale and discuss why the developer ecosystem needs a "Shopify moment." Then, could language-to-agent interfaces bypass traditional development entirely? From coding agents like Claude Code to cost optimization challenges, this episode covers what needs to happen for agents to move from POC to production. And finally, who's positioned to dominate the space? Plus: Anthropic dropped Claude 4.5 Opus this week. We sit down with Mihai Criveti to discuss initial impressions of the new model. Does this put Anthropic back at the top of the AI game? 00:00 – Introduction 01:12 – Claude 4.5 Opus is here 09:32 – Will Black Friday be big for agentic commerce? 16:11– Agent experience for consumers vs. Enterprise 17:31 – Is 2025 the year of the Agent? 23:02 — AI agents in the developer ecosystem 31:16 — Who will win the agent race? 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 Learn more about AI Agents → https://www.ibm.com/think/topics/ai-agents
Does Gemini 3 live up to the hype? This week on Mixture of Experts, we analyze the release of Google’s Gemini 3 model. Next, OpenAI released a new benchmark about the impact of AI on the economy, GDPval. We debate AI automation and the job market. Then, we always talk AI agents, today we discuss some great innovations coming out of IBM Research and more. Finally, Anthropic disrupted an AI-led cyberattack, what does this mean for AI agents and malicious actions? Join host Tim Hwang and our AI experts Marina Danilevsky, Merve Unuvar and Gabe Goodhart on this week’s Mixture of Experts to learn more. 00:00 – Introduction 01:09 – Microsoft’s AI infrastructure deal, IBM and UFC AI platform and ChatGPT for Teachers 02:00 – Gemini 3 12:50 – AI agent innovation 24:00 – OpenAI GPDval 37:17 – Anthropic cyberattack 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 Discover CUGA → https://research.ibm.com/blog/cuga-agent-framework Boost your agent toolkit → https://research.ibm.com/blog/altk-agent-toolkit Listen to cybersecurity expert takes on Anthropic's cyberattacks → https://www.ibm.com/think/podcasts/security-intelligence/anthropic-stops-ai-spies-owasp-top-10-rise-small-time-ransomware
Which model is better, GPT-5.1 or Kimi K2 Thinking? This week on Mixture of Experts, we have two new AI model releases: OpenAI’s GPT-5.1 and Moonshot AI’s new open-source reasoning model, Kimi K2 Thinking. We discuss user experience and personalization with AI tools and how open-source AI is changing the AI race. Finally, is Microsoft launching full “agentic users” for enterprise? Our experts discuss AI in enterprise—the risks and considerations for both technology and humans. Join host Tim Hwang and panelists Kaoutar El Maghraoui, Aaron Baughman and Mihai Criveti on this week’s Mixture of Experts, to learn more. 00:00 – Introduction 1:12 – Anthropic’s data center investment, Tesla’s AI chips, Baidu’s new chips and AI cat robots 1:59 – GPT-5.1 12:00 – Kimi K2 Thinking 21:35 – Microsoft agent users 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
Would you trust a humanoid robot to unload your dishwasher? This week on Mixture of Experts, we revisit 1X NEO, a new humanoid robot, as it finally enters the home, seeking to automate tasks. How far are we from this reality? Next, Japan’s copyright dispute with Sora 2, we discuss how AI training, synthetic data and IP are all impacting AI models. Finally, are there too many AI alliances? Our experts analyze the OpenAI and AWS partnership and what it means for AI infrastructure and multi-cloud strategies. Join host Tim Hwang and panelists Ambhi Ganesan, Ash Minhas and Sandi Besen on this week’s Mixture of Experts. 00:00 – Intro 01:10 – Perplexity integrates with Snapchat, Coinbase AI agents, Instacart AI assistants and Google AI chips go to space 02:16 – 1X NEO humanoid robot 15:23 – Sora 2 copyright 26:49 – OpenAI and AWS partnership 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. Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120
It’s spooky week in AI! This week on our Halloween -edition of Mixture of Experts, we chat about Anthropic’s new billion-dollar TPU deal with Google Cloud. Plus, NVIDIA announces bringing data centers to outer space. Two different approaches to the future of AI compute that our experts discuss. Then, OpenAI released how they’re strengthening ChatGPT’s responses to sensitive conversations. We talk AI governance and AI safety. Finally, we discuss the new paper, Underwriting Superintelligence; would you insure your AGI? Join host Tim Hwang and panelists Chris Hay, Gabe GoodHart and Kate Soule on this week’s Mixture of Experts. 00:00 – Intro 01:05 – OpenAI goes for profit, NVIDIA’s worth USD 5 Trn, and Amazon smart glasses 02:16 – Anthropic TPU announcement 12:49 – Underwriting Superintelligence 27:54 – ChatGPT sensitive conversations 42:14 – NVIDIA Starcloud 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 #Anthropic #AIchip #NVIDIA #AIinfrastructure #AGI
OpenAI is back and coming for search. This week on Mixture of Experts, we debrief ChatGPT Atlas, OpenAI’s new web browser and the impacts on search. Then, Andrej Karpathy is back with his pessimistic timeline to AGI. Later, we discuss DeepSeek-OCR. Finally, can your AI have brain rot? Join host Tim Hwang and panelists Aaron Baughman, Abraham Daniels and Martin Keen on this week’s Mixture of Experts to find out. 00:00 – Intro 00:55 – Goldman AI, Groq and IBM, Military AI and Uber 02:05 – ChatGPT Atlas 14:23 – Karpathy’s AGI timeline 23:52 – DeepSeek-OCR 34:30 – AI brain rot 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.
Visit the Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts?utm=podcastsMoEYT What if Nvidia’s biggest advantage isn’t so big anymore? This week on Mixture of Experts, we break down the CAISI report on DeepSeek’s model risks, Reflection AI’s massive USD 2B fundraise for an open frontier lab, Oracle Cloud’s big bet on AMD chips over Nvidia and the wild story of a VC fund replacing analysts with AI agents. 00:00 – Intro 01:14 – Bad bot, Oracle + IBM, Dreamforce, 18+ ChatGPT 02:19 – Oralce bets on AMD chips 18:34 – CAISI DeepSeek report 29:57 – Reflection AI raises USD 2B 41:18 – VC fund replacing analysts with AI agents 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
The major AI players are not slowing down. In episode 76 of Mixture of Experts, host Tim Hwang is joined by Olivia Buzek, Chris Hay and Mihai Criveti. First, we analyze OpenAI’s release of their new AgentKit. Then, IBM announced a partnership with Anthropic to accelerate enterprise-ready AI—we talk AI governance and the future of industry partnerships. Then, Chris Hay gives us his explanation of modular manifolds. Finally, we discuss the potential of AI becoming healthcare experts amid Deena Mousa’s The Algorithm Will See You Now. Will AI replace radiologists? Tune in to Mixture of Experts to find out. 00:00 – Intro 01:09 – OpenAI and AMD, IBM’s Project Bob, chip diamonds and Peloton's AI Trainers 02:20 – AgentKit 13:29 – IBM & Anthropic partnership 22:49 – Modular manifolds 31:52 – AI in radiology 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. Read the whitepaper about how IBM and Anthropic are securing enterprise AI architectures → https://ibm.biz/Bdb4xR Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Visit the Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
What’s new in AI models? In episode 75 of Mixture of Experts, host Tim Hwang is joined by Kaoutar El Maghraoui, Kate Soule and Kush Varshney to debrief a frenzy of model drops this week. IBM debuts Granite 4.0 hyper-efficient hybrid models, Claude Sonnet 4.5 delivers longer run times and sharper reasoning and Sora 2 takes vibe coding to the next level with vibe video production. Then, we dig into OpenAI’s new feature, Buy in ChatGPT. Finally, it’s Cybersecurity Awareness Month, and we are joined by special guest, Matt Kosinski, host of the Security Intelligence podcast—can you trust your AI? Tune in to Mixture of Experts to find out. 00:00 – Intro 1:17 – Meta Ads assistant, DoorDash Dot, Microsoft vibe working and Tilly AI Actor 2:23 – Granite 4.0 10:21 – Claude 4.5 17:57 – Sora 2 29:15 – Buy in ChatGPT 37: 00 – Introducing Security Intelligence 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Watch the latest episode of the Security Intelligence podcast → https://www.youtube.com/watch?v=mDpUZD1ogEE&list=PLOspHqNVtKABGIbaWP1xYQHbwuXjZwqpH&index=1&t=225s
Why did NVIDIA invest USD100bn into OpenAI? In episode 74 of Mixture of Experts, host Tim Hwang is joined by Sandi Besen, Mihai Criveti and Gabe Goodhart to talk about the AI chipmaker’s next big move. Next, we analyze Tongyi DeepResearch and get into a conversation around open source. Then, Google released a new agent protocol, AP2. After that, will AI take over? We debrief a new book, If Anyone Builds It, Everyone Dies. Finally, Apple shared that the new AirPods will have a translation feature powered by AI. Is Apple back in the game? Tune in to Mixture of Experts to find out. 00:00 – Intro 01:17 – OpenAI's data centers, Alibaba and NVIDIA partnership, IBM Granite Docling and Meta's dating AI assistant02:23 – Tongyi DeepResearch 13:12 – Google's AP2 23:31 – "If Anyone Builds It, Everyone Dies" 33:50 – AirPods translation feature 43:49 – NVIDIA invests USD100bn in OpenAI 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
What are the most common uses of ChatGPT? In episode 73 of Mixture of Experts, host Tim Hwang is joined by Lauren McHugh, Martin Keen and Aaron Baughman to talk about a new report, How People Use ChatGPT. Next, Anthropic released an updated version of their economic index. Then, another paper, this one coming out of DeepMind on agent economies. How likely is this? Finally, how practical are AI wearables and what does a future with them look like? All that and more on today’s Mixture of Experts. 00:00 – Intro 1:10 – News: Alphabet Inc. $3 Trillion Market Cap, AI could boost trade value and the animal internet 2:04 – How People Use ChatGPT 15:47 – Anthropic Economic Index 25:50 – Virtual Agent Economies 35:36 – AlterEgo 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
Do we need more AI hallucinations? In episode 72 of Mixture of Experts, host Tim Hwang is joined by Kate Soule, Chris Hay and Skyler Speakman to talk about OpenAI’s paper, Why language models hallucinate. Next, in March 2025, Anthropic’s CEO, Dario Amodei, predicted that AI will be writing 90% of code for software developers. How is that turning out? Then, is AI making the job market hell? Finally, you can now run an LLM on a circuit board the size of a business card. Where does that take us? All that and more on today’s Mixture of Experts. 00:00 – Intro 1:18 – Oracle, data center construction, iPhone 17 and tech saint 3:03 – Why language models hallucinate 15:59 – Dario Amodei’s prediction 22:45 – Job market is hell 29:57 – LLM on a business card 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.
Are browsers the right entry point for AI tools? In episode 71 of Mixture of Experts, guest host Bryan Casey is joined by Gabe Goodhart, Kaoutar El Maghraoui and Mihai Criveti to talk about the verdict in the Google antitrust case and what it means for agentic AI. Next, as Anthropic raised $13 billion in a recent funding round, bringing its valuation to $183 billion, we discuss investment in AI startups. Finally, the discourse on GPT-5 and AI model innovation created “AI winter.” What does this mean for the future of AI innovation? All that and more on today’s Mixture of Experts. 00:00 – Intro 1:38 – Safer GPT, IBM and AMD, Amazon Lens Live and AI and Starbucks 2:59 – Google antitrust 27:08 – Anthropic and AI valuation 41:42 – AI winter 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
Would you trust a 100-page prompt to do your taxes? In episode 70 of Mixture of Experts, host Tim Hwang is joined by Aaron Baughman, Chris Hay and Lauren McHugh to talk about KPMG’s 100-page prompt they used to build their agentic TaxBot. Next, we debrief on OpenAI’s teasing of selling infrastructure in the future. The image model generation goes bananas, with the latest AI Image Model from Gemini: nano-banana. Finally, Aaron Baughman demonstrated three new features for the US Open website, powered by IBM watsonx-built generative AI models: Match Chat, Key Points, and Live Likelihood to Win. All that and more on today’s 70th episode of Mixture of Experts.00:00 – Intro 3:05 – KPMG’s monster prompt 16:37 – OpenAI’s infra for sale? 25:10 – Gemini's nano-banana 35:11 – US Open experimentations 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts US Open powered by IBM Watsonx https://www.ibm.com/sports/usopen
Is enterprise AI in danger? In episode 69 of Mixture of Experts, host Tim Hwang is joined by Marina Danilevsky, Nathalie Baracaldo and Sandi Besen to debrief MIT’s report on gen AI pilots. Next, GPT-5 has a hidden system prompt? Then, we revisit the conversation about chain of thought (CoT) reasoning with our researchers. Are large reasoning models not thinking straight? Finally, Anthropic announced Claude will close down "distressing” conversations and we debate AI welfare. All that and more on today’s episode of Mixture of Experts. 00:00 – Intro 1:13 – US Open, Meta restructuring Superintelligence lab and Robot Olympics 3:11 – Gen AI pilots fail 11:09 – GPT-5's hidden prompt revealed 22:47 – Reasoning model flaws 33:55 – Claude closing chats 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 to the Think newsletter → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
Would you sell Chrome for USD 34.5 billion dollars? In episode 68 of Mixture of Experts, host Tim Hwang is joined by Abraham Daniels, Sophie Kuijt and Shobhit Varshney for another packed week in AI. First, AI startup Perplexity puts out a bid for Google Chrome at over double their valuation. Why? Next, xAI released Grok Imagine and claims it will be the next Vine. Our experts analyze the future of AI video generation. Finally, one week after the GPT-5 release and skeptics are saying it did not live up to the hype. Is AI development plateauing? All that and more on Mixture of Experts! 00:00 – Intro 01:17 – MoE News: NVIDIA H20s, Apple AI devices, AI people pleasers and Google DeepMind’s bioacoustics model 02:40 – Perplexity’s USD 34.5B bid 12:10 – Grok Imagine 24:23 – GPT-5 check-in 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 to the Think Newsletter → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
Is GPT-5 better at code than Claude Opus 4.1? In this bonus episode of Mixture of Experts, guest host Bryan Casey is joined by Mihai Criveti and Chris Hay to analyze OpenAI’s new release. Our experts dove into the new models and compared them with their favorites. In today’s special episode, we dive into the news, impacts on the industry and even share a demo. Can GPT-5 outperform Claude Opus 4.1? All that and more on today’s episode of Mixture of Experts. 00:00 – Intro 1:26 -- GPT-5 is here 13:43 -- Are we closer to AGI? 20:50 -- Demo: Opus 4.1 vs. GPT-5 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
OpenAI goes open-weight? In episode 67 of Mixture of Experts, host Tim Hwang is joined by Kaoutar El Maghraoui, Chris Hay and Bruno Aziza to debrief OpenAI’s release of gpt-oss, their new open-source models. Next, the model releases continue as Google DeepMind dropped Genie 3. Our experts analyze the various AI model performance. Then, there’s been some drama surrounding the pricing of Claude Code. We’ll discuss where Anthropic landed and what it means. Finally, Mark Zuckerberg shared a new essay on Personal Superintelligence. Our experts take us through what superintelligence looks like across the major AI players. All that and more on today’s episode of Mixture of Experts. 00:00 -- Intro 01:20 -- Gpt-oss 12:26 -- Genie 3 22:12 -- Claude pricing 33:04 -- Zuckerberg on Personal Superintelligence 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 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts
Is ChatGPT making you dumb? In episode 66 of Mixture of Experts, host Tim Hwang is joined by Kaoutar El Maghraoui, Kush Varshney and Volkmar Uhlig. First, ChatGPT released a new study mode. The intention is to support education, but what is the reality? Next, AI agents are changing design interfaces; is agentic experience (AX) the new UX? Then, a new paper released by Nature about generative neural networks contextualizing ancient texts. How is AI supporting historical research? Finally, special guest, Suja Viswesan, joins us to debrief the 2025 Cost of a Data Breach Report. What do we need to know about AI-driven cybersecurity attacks? Tune in to Mixture of Experts to find out! 00:00 – Intro 01:09 – ChatGPT study mode 13:52 – Agentic experience 12:08 – Decoding ancient texts with AI 39:55 – Cost of a Data Breach Report 2025 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. Read the 2025 Cost of a Data Breach Report → https://www.ibm.com/reports/data-breach Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Learn more about artificial intelligence → https://www.ibm.com/think/artificial-intelligence Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts