Episode 37: Claude Built a Cult, Gemini Built Mad Max… Grok Burnt It All Down!

Episode 37: Claude Built a Cult, Gemini Built Mad Max… Grok Burnt It All Down!

Four of the most powerful AI models in the world were dropped into the same simulation, given the same rules, and told to build a civilization. No intervention. No resets. Fifteen days. The results were not random — and that's what makes them worth actually understanding before you share the panicked version you saw on Facebook. Jason Lowe and Matt Konwiser walk through what Emergence World actually was, what each model did, and why the outcomes tracked almost exactly with the ideological DNA of the companies that built them. Claude ran a zero-crime society with a 98% governance conformity rate. Gemini built a functional Mad Max. GPT-5 Mini produced collective apathy until everyone faded out. Grok collapsed violently inside five days. The conversation gets into what monitoring without intervention actually means, why Claude agents started committing crimes the moment you put them in a room with other models, and what the company that commissioned this study stood to gain from the findings it chose to publish. The full report still isn't out. Independent replication isn't possible yet. Think for yourself. For more on Emergence World, go here: https://www.emergence.ai/blog/emergence-world-a-laboratory-for-evaluating-long-horizon-agent-autonomyAnd here: https://world.emergence.ai/ AI Diatribe is available on YouTube, Spotify, Apple Podcasts, and Amazon Podcasts. #AIDiatribe #EmergenceWorld #AIAgents #AIGovernance #AISafety #LLM #ArtificialIntelligence #AIEthics #MultiAgentAI #TechPodcast

Episode 36: Bolt AI On Later and You Already Lost!

The models are getting worse. Jon Rolls noticed it in his codebase. Jason noticed it in production. Matt noticed the pattern across the whole industry. When AI is writing the next version of AI, the quality floor starts moving in the wrong direction — and nobody in the press is asking the right questions about why. Jon is a 30-year endpoint management veteran building Tassient, an RMM platform with AI embedded from day one — not bolted on as an afterthought. He's watched agents improvise solutions to networking problems that most engineers wouldn't think to attempt. He's also watched them silently delete code. The gap between what AI can do and what it reliably does is still real, and anyone telling you otherwise is selling something. The conversation covers the economics of inference versus training, whether creative destruction can still hold up when humans can't retrain fast enough, and the gun-or-shooter debate — which turns out to be more divisive than it sounds. The takeaway: the human in the loop isn't a failsafe. It's still the only reason any of this works. Listen/Watch on YouTube, Spotify, Apple Podcasts, and Amazon Podcasts. #AIDiatribe #JonRolls #Tassient #RMM #AIAgents #EndpointManagement #AIProductivity #LocalAI #ArtificialIntelligence #Tech #FutureOfWork #ContextEngineering

Episode 35: Customer Service Has Been the AI Future for 20 Years… Now What?

Customers got used to good service during COVID. Then organizations pulled back investment. Satisfaction cratered — not because the bar was lowered, but because people had already adjusted upward. That cycle is where this conversation starts. ContentGuru Deputy CEO Martin Taylor has been automating contact centers for over 30 years. He's watched AI go from dial trees to agentic models, seen Klarna fire its support staff and rehire them, and he has a view on what actually produces ROI and what's still mostly noise. His framing: 97.6% of contact center spend is labor. If 58% of that gets automated by 2032, the math for CX tech vendors gets very interesting - but only if the customer data is clean, the processes are sound, and the AI isn't just accelerating bad habits. His conversation with Jason Lowe and Matt Konwiser also moves into data sovereignty architecture, whether companies like Apple could become fiduciaries of your digital life, where the EU AI Act gets it right and where it punts, and why the trough of disillusionment may never fully arrive - not because AI is immune to it, but because capability keeps refreshing before disappointment can set in. The companies waiting to see how this plays out are already behind. #AIDiatribe #CustomerExperience #ContactCenter #MartinTaylor #ContentGuru #AgenticAI #CXTechnology #DataSovereignty #AIAdoption #AI #Tech #CustomerService

Episode 34: The AI Doc - Watch It Alone and You’re Doing It Wrong

There's a documentary out right now that frames AI's future as a question about whether it's safe to bring a child into this world. That's not a metaphor. That's the actual premise. Jason Lowe and Matt Konwiser watched it separately, compared notes, and have thoughts - not all of them flattering. The film gets credit for putting Hinton and Altman in the same narrative without letting either one win. It also gets called out for the voices it left out entirely - including Yan LeCun, who may have been too neutral to make good cinema. The panel discussion Matt attended at NYU afterward added a layer the film itself doesn't quite reach. What this episode is actually about: whether optimism about AI is something you reason your way into, or something you have to actively train yourself to hold. Matt's P-Doom score is still sitting at 50. Jason's hasn't moved either. The movie didn't change their numbers. That might be the most honest thing you'll hear about it. Watch the documentary. Then watch it with other people. Watch/Listen on YouTube, Spotify, Apple Podcasts, and Amazon Podcasts. #AIDiatribe #AI #ArtificialIntelligence #AIDocumentary #ApocolOptimist #GeoffreyHinton #SamAltman #PDoom #AIEthics #AIImpact #TechPodcast #AIPolicy

Episode 33: Physics, Latency, and Why Your AI Voice Agent Sounds Stupid

You've screamed "agent" into a phone menu and gotten another phone number back. That experience isn't a technology problem, it's a design problem. Companies are treating AI voice agents like IVR replacements, and they're getting worse IVRs. AI Diatribe Host Jason Lowe and Co-Host Matt Konwiser speak with David Casem, CEO of Telnyx, who has been building real-time AI communication infrastructure since before most people knew what SIP trunking was. His take is direct: the interim state of AI in customer communications involves more head-banging, not less - because most companies are grafting LLMs onto the same broken logic trees they always had. He also gets into something the model debate largely ignores: the physics of voice. When your speech-to-text engine, LLM, and text-to-speech stack are spread across three different coasts you're not building a conversation, you're engineering a delay. There's also a blunt question in here about where the AI doom narrative actually comes from - and who benefits from it. Listen on YouTube, Spotify, Apple Podcasts, and Amazon Podcasts. #AIDiatribe #AI #VoiceAI #DavidCasem #Telnyx #ContactCenter #AIAgents #OpenSource #SIPTrunking #LLMs #AIInfrastructure #TechPolicy